ENISALA: II. Distinct Star Formation and Active Galactic Nucleus Activity in Merging and Relaxed Galaxy Clusters
DDraft version February 23, 2021
Typeset using L A TEX twocolumn style in AASTeX63
ENISALA: II. Distinct Star Formation and Active Galactic Nucleus Activity in Merging and RelaxedGalaxy Clusters
Andra Stroe ∗ and David Sobral Center for Astrophysics | Harvard & Smithsonian, 60 Garden St., Cambridge, MA 02138, USA Department of Physics, Lancaster University, Lancaster LA1 4YB, UK (Received November 2, 2020; Revised January 25, 2021; Accepted February 16, 2021)
Submitted to ApJABSTRACTThe growth of galaxy clusters is energetic and may trigger and/or quench star formation and blackhole activity. The ENISALA a) project is a collection of multiwavelength observations aimed at under-standing how large-scale structure drives galaxy and black hole evolution. Here, we introduce opticalspectroscopy of over 800 H α emission-line galaxies, selected in 14 z ∼ . − .
31 galaxy clusters,spanning a range of masses and dynamical states. We investigate the nature of the emission lines inrelation to the host galaxy properties, its location within the cluster, and the properties of the parentcluster. We uncover remarkable differences between mergers and relaxed clusters. The majority of H α emission-line galaxies in merging cluster fields are located within 3 Mpc of their center. A large fractionof these line-emitters in merging clusters are powered by star formation irrespective of cluster-centricradius, while the rest are powered by active galactic nuclei. Star-forming galaxies are rare within 3 Mpcof relaxed clusters and active galactic nuclei are most abundant at their outskirts ( ∼ . − Keywords:
Active galaxies (17), Early-type galaxies(429), Emission line galaxies (459), Galaxy clusters(584), Galaxy environments (2029), Galaxy evolution (594), Ionization (2068), Intraclustermedium (858), Spectroscopy (1558), Spiral galaxies (1560), Star formation (1569), Shocks(2086)
Corresponding author: Andra [email protected] a) The project is named as a tribute to the storied Enisala citadel(Dobrogea, Romania). Enisala (‘new settlement’, in Turkish andRomanian) sits on top of a windswept hill, at the crossroads ofthe Danube Delta and the Pontus Euxinus sea (‘hospitable sea’,Black Sea), forever shaped by forces of nature. It stands as ametaphor for the ever-evolving galaxy cluster environment and itsprofound influence on galaxy and black hole evolution. ENISALAcan also be understood to stand for ‘ENvironmental Influence onStar formation and AGN through Line Astrophysics’. ∗ Clay Fellow INTRODUCTIONLarge-scale structure plays a critical role in the evolu-tion of galaxies, accelerating the growth of star-formingspirals into passive ellipticals. Galaxies in evolved, re-laxed galaxy clusters have systematically redder colors,with more evolved, elliptical and lenticular morpholo-gies, lower star formation rates (SFR), and lower gasfractions compared to field counterparts (e.g. Gunn &Gott 1972; Kenney et al. 2004; Dressler 1980; Goto et al.2003; Melnick & Sargent 1977; Balogh et al. 2004; Chunget al. 2009a). One of the main pathways for galaxyevolution in clusters is the interaction between infallinggalaxies and the dense, hot intracluster medium (ICM). a r X i v : . [ a s t r o - ph . GA ] F e b The ICM can remove interstellar gas from the galaxyaltogether or prevent the accretion of new gas reser-voirs, thus depriving galaxies of the fuel needed for fu-ture star formation (SF) episodes (Gunn & Gott 1972;Larson et al. 1980; Bekki et al. 2002). Before fallinginto the cluster, galaxies can experience pre-processingin filaments (Darvish et al. 2017; Paulino-Afonso et al.2018; Connor et al. 2018). Especially at the outskirtsof relaxed clusters, high-speed close encounters betweenpairs of galaxies can lead to a truncation of the halo anda shut down of SF (Moore et al. 1996). Environmentaleffects are very efficient at quenching galaxies as evena single passage through the cluster can render galaxiessignificantly redder than the infalling population (e.g.Pimbblet 2011; Muriel & Coenda 2014).Through invaluable work spanning the entire electro-magnetic spectrum, complemented by simulations andtheoretical developments, galaxy evolution studies fromthe perspective of large scale structure have establishedthe critical role relaxed cluster environments have inshaping the evolutionary tracks of their member galax-ies. However, the bulk of research carried over the last50 years has mostly focused on contrasting field andrelaxed cluster environments (Dressler 1984; Boselli &Gavazzi 2006). Matter in the Universe is distributed ina non-uniform matter, spanning voids, filaments, sheets,groups, and large clusters. Therefore, a large fraction ofgalaxies do not reside in either a highly-evolved massivelocal cluster or in average density field environments. Todistinguish between galaxy evolution models, we need tofill in the missing information on the effect of intermedi-ate environments such as galaxy groups, filaments, dis-turbed clusters, and their outskirts (e.g. Kodama et al.2001; Darvish et al. 2017).The most transformational events in the lifetime ofa galaxy cluster are collisions and mergers with otherclusters. Mergers between clusters are the most en-ergetic events since the Big Bang and significantly al-ter the evolutionary pathways of the participating clus-ters. Simulations and observations find that mergersinject non-thermal components, in the form of mag-netic fields, relativistic particles, bulk motion, cluster-wide turbulence, and weak shocks. Mergers can alsoproduce Mpc-wide strong shocks, which heat the ICM,accelerate particles to relativistic speeds and provide5 −
25% of pressure support, with the highest impactat cluster outskirts (e.g. Eckert et al. 2019; Biffi et al.2016). Effects of turbulence can be measured throughthe broadening of X-ray emission lines or radio obser-vations revealing synchrotron emission associated withrelativistic electrons. Shock waves can be detected as X-ray discontinuities or as arc-like patches of diffuse radio emission from shock-accelerated particles (Markevitch& Vikhlinin 2007; Vazza et al. 2009; van Weeren et al.2019).Cosmological simulations unveiled that mergers werecommon at z > z <
1. However, in their pioneeringstudy, Jones & Forman (1999) discovered that 40% ofX-ray selected galaxy clusters at z < . −
70% ofmass-selected and/or X-ray volume-limited samples arenot relaxed (Andrade-Santos et al. 2017; Rossetti et al.2017; Lovisari et al. 2017; Chon & B¨ohringer 2017).Thus, a large fraction of SF in cluster galaxies actu-ally happens in merging, disturbed structures. If we areto understand the evolution of galaxies from the fieldto relaxed clusters, it is crucial to understand the in-termediary evolution stages. What happens to galax-ies when their parent cluster is undergoing a massivemerger? Studies have found striking differences betweenthe SF properties, morphologies, active galactic nucleus(AGN), and gas reservoir properties of galaxies in merg-ing clusters compared to those in relaxed clusters. Somedisturbed galaxy clusters have a higher density of H α -bright galaxies (Stroe et al. 2014, 2015b, 2017), a higherfraction of star-forming (e.g. Cohen et al. 2014; Yoon& Im 2020) and blue galaxies (e.g. Wang et al. 1997;Cortese et al. 2004; Hou et al. 2012; Cava et al. 2017)and are more gas-rich than counterparts in relaxed clus-ters (Stroe et al. 2015a; Jaff´e et al. 2012, 2016; Cairnset al. 2019). There is also compelling evidence frommultiwavelength data that cluster mergers trigger AGNactivity (Miller & Owen 2003; Owen et al. 1999; Sobralet al. 2015; Hwang & Lee 2009). Not only is the num-ber of star-forming galaxies increased in merging clus-ters compared to relaxed clusters, but the morphologicaland spectroscopic properties of the star formers also dif-ferent between merging clusters, relaxed environments,and the field. For example, Yoon et al. (2019) foundthat cluster-cluster interactions at z < .
06 trigger theformation of bars in galaxies at all stellar masses studied(10 − . M (cid:12) ), an effect attributed to strong asymmet-ric perturbations induced by the rapidly changing tidalfield in merging galaxy clusters. Mulroy et al. (2017)find that galaxy colors are standardized by a cluster-wide process, such as shock waves, in merging clusters.In a spectroscopic study, Sobral et al. (2015) find thatactive galaxies in merging clusters are metal-rich, haveevidence for outflows, and have very low electron den- Table 1.
List of galaxy clusters with spectroscopic follow-up of H α cluster member candidates.Cluster Nickname R.A. Decl. z L X − ray State hh mm ss ◦ (cid:48) (cid:48)(cid:48) (10 erg s − )(8) (9) (10) (11) (12) (13) (14)Abell 1689 A1689 13 11 29 −
01 20 17 0.183 14 relaxedAbell 963 A963 10 17 13 +39 01 31 0.206 6 relaxedAbell 2390 A2390 21 53 35 +17 41 12 0.228 13 relaxedZwicky 2089 Z2089 09 00 36 +20 53 39 0.2343 7 relaxedRX J2129+0005 RXJ2129 21 29 38 +00 05 39 0.235 12 relaxedRX J0437.1+0043 RXJ0437 04 37 10 +00 43 38 0.285 9 relaxedAbell 2254 A2254 17 17 40 +19 42 51 0.178 5 merger (+ turbulence)CIZA J2242.8+5301 Sausage 22 42 50 +53 06 30 0.188 7 merger (+ shocks)Abell 115 A115 00 55 59 +26 22 41 0.1971 9 merger (+ shocks)Abell 2163 A2163 16 15 34 −
06 07 26 0.203 38 merger (+ turbulence)Abell 773 A773 09 17 59 +51 42 23 0.217 6 merger (+ turbulence)1RXS J0603.3+4214 Toothbrush 06 03 30 +42 17 30 0.225 8 merger (+ shocks and turbulence)Abell 2219 A2219 16 40 21 +46 42 21 0.2256 12 merger (+ turbulence)Abell 2744 A2744 00 14 18 −
30 23 22 0.308 13 merger (+ shocks and turbulence) sities, which indicate high supernova rates (SNR) andsustained SF for timescales of 500 Myr. In additionto emission-line galaxies, some merging clusters containsignificant populations of post starburst (E+A) galaxiesin merging clusters, showing a possible correlation withcluster merger timescale (e.g. Ma et al. 2010; Prangeret al. 2014). By contrast, Deshev et al. (2017) found adepletion of star-forming galaxies in the core of the mas-sive merging Abell 520 cluster, with the bulk of SF hap-pening along infalling filaments. Chung et al. (2009b)and Shim et al. (2015) found that mid-infrared colors,a measure of specific SFR (sSFR), did not vary signifi-cantly across the shock fronts in the Bullet and the Abell2255 clusters. Simulations predict that the increasedpressure caused by the merger-induced traveling shockwaves can cause a temporary burst of SF, but ultimatelylead to a fast consumption of gas and a shut down on SF(Fujita et al. 1999; Roediger et al. 2014). Further, Ebel-ing & Kalita (2019) propose a scenario that brings someof these results into agreement, in which galaxies in thecore of merging clusters experience increased quenching,while late-type galaxies falling into the merging systemalong filaments can experience a burst of SF trigger bya cluster-wide, merger-induced shock wave.The drivers of SF and black hole (BH) activity in clus-ter galaxies seem to be closely linked to the merger his-tory of the host cluster and its associated filaments andinfalling groups, opening a new window into how galax-ies evolve. Despite the growing amount of evidence to their profound influence, some results are in disagree-ment, and the exact mechanisms through which clustermergers drive galaxy evolution are still poorly under-stood. ENISALA - AN EMISSION LINESPECTROSCOPIC SURVEY OF MERGING ANDRELAXED GALAXY CLUSTERSWe commenced the
ENISALA project, an ambitiousmultiwavelength photometric and spectroscopic observ-ing campaign to unveil the evolutionary pathways ofgalaxies in merging clusters and their large scale struc-ture, along with a comparison sample of relaxed clus-ters. In the first paper from the series (Stroe et al.2017), we presented the results from the first system-atic survey of SF activity in a statistically-significantset of 19 0 . < z < .
31 clusters samples a range ofmasses, luminosities, and dynamical states. We em-ployed custom-made narrow-band (NB) filters to selectstar-forming and active galaxies through their H α emis-sion, over the entire 3D volume of their host clusters.Our method results in a very simple selection func-tion, which uniformly selected star-forming galaxies inand around clusters down to a well-understood star-formation rate (SFR) limit. We found striking differ-ences between relaxed and merging clusters, with merg-ing environments having over two times more H α galax-ies compared to merging clusters, especially those host-ing large scale shocks, which are overdense by a factor of Figure 1.
A multiwavelength view of Abell 2254, a massive 1 . − × M (cid:12) cluster undergoing a major merger that injectsturbulence in the ICM. Symbols mark the positions of H α candidates selected through our NB survey (small empty circles),those confirmed to be H α at the cluster redshift (large circles), and emitters confirmed to be at other redshifts (crosses). Filledsymbols show the confirmed H α candidates with spectroscopy available through our project. Left: Our NB image targeting H α at the cluster redshift is shown in grayscale, with X-ray emission from XMM-Newton tracing the ICM in pink contours andradio observations from NVSS in green contours, highlighting locations of radio galaxies and the diffuse emission associatedwith the ICM. Right: The same X-ray image is shown in the background with radio contours overlaid. α galaxies selected throughour NB observations in 14 merging and relaxed clus-ters and their immediate cosmic web (see Table 1 andFigure 1). The main drivers of our spectroscopic ob-servations are to confirm the cluster membership andconstrain the star-formation, ionization, metallicity, andelectron density properties of the galaxies. Our samplecontains over 800 galaxies with measurements of at leastone main optical emission line and over 300 galaxies withmeasurements of enough lines to classify them securelyin a Baldwin et al. (1981) (BPT) diagram.In this paper (Paper II) of the series, we will givea general introduction to our ENISALA spectroscopicsurvey, including the survey strategy, data acquisition,reduction, and the initial spectroscopic products, such asredshifts, line measurements, and ratios. We also discuss the first results from the ENISALA spectroscopic survey,focusing on the SF properties and ionization sources ofthe H α -selected galaxies, while future papers will focuson properties such as metallicity, ionization potential,temperatures, and electron density.The paper is structured as follows: Section 3 presentsthe parent NB sample and the observing strategy forthe follow-up spectroscopy, while in Section 4 we mea-sure redshifts from the spectra and present the red-shifts package. Section 5 presents a validation of theNB selection and the distribution of follow-up sourceswith respect to the parent sample. Section 6 describeshow measurements are derived from spectroscopy. Wepresent the final sample used in this paper in Section 7.In Section 8, we discuss velocity width and equivalentwidth (EW) properties of the sample as a function ofthe host galaxy color, ionization source, and locationwithin the cluster. Section 9 aims to embed our resultsinto the overall picture of environmentally driven evo-lution by focusing on viable galaxy and BH evolution-ary pathways in the context of galaxy cluster growth.We present our conclusions and outlook to the futurein Section 9. We assume a ΛCDM cosmology, with H = 70 km s − Mpc − , Ω M = 0 . Λ = 0 .
7. Wereport AB magnitudes throughout. The clusters in thesample range from 0.178 to 0.308 in redshift, which cor-responds to physical scales of 3 . − .
55 kpc arcsec − . Table 2.
Details of the spectroscopic observations. For each clus-ter, we list the telescope used for taking the data, the number ofindependent pointing/source setups, the spectral resolution of theinstrument, and the total exposure time for each pointing.Cluster Instrument Pointings ∆ λ Exp. Time(˚A) (hours)(6) (7) (8) (9) (10)A1689 Hectospec/MMT 6 6 0 . − . . − . . − . . − . . − . . − . . − . . − . SAMPLE, OBSERVING STRATEGY AND DATAREDUCTION3.1.
Parent NB H α sample In Stroe et al. (2017), using NB filters, we selectedover 3000 H α emitting candidates in the fields around 19galaxy clusters at 0 . < z < .
31. The sample includedrelaxed and merging clusters, with cluster-wide shocksand turbulence, and spanned masses of 5 − × M (cid:12) .The observations cover a field of view (FOV) of 0 . centered on each cluster, or about a maximum 3 − > . × Mpc . The limiting dust-uncorrectedH α luminosities are 10 . − . erg s − . The observa-tions reach equivalent dust-corrected SFRs at the levelof 0 . − . z ∼ .
2. 3.2.
Target Selection, Observations, and DataReduction
We carried out spectroscopic follow-up observationsof a subset of the emitters in 14 galaxy clusters, ofwhich six are relaxed, and eight are undergoing mergers(see Table 1 for details on the cluster properties). Weobtained new multi-object spectroscopy (MOS) usingthree different instruments (VLT/VIMOS, WHT/AF2,Keck/DEIMOS) described below. We also leveragedpublicly available spectroscopy obtained through theArizona Cluster Redshift Survey (ACReS ). In design-ing our observations, the main drivers were to cover anddetect all main rest-frame optical emission lines, includ-ing spectral coverage of H α , [N ii ] ( λλ , β ,[O iii ] ( λ ii ] ( λλ , α and[N ii ] , and the [S ii ] doublet, which enables us to accu-rately deblend lines, precisely measure redshifts, and, insome cases, resolve some lines in velocity. While notdesigned with absorption in mind, the Mg and Na ab-sorption lines, for example, are covered. We designedmasks and fiber configurations to prioritize NB H α can-didates with bright NB H α luminosities and ensure thedetection of emission lines with a small telescope timeinvestment. Any remaining space was allocated to otherlikely cluster members. Detecting continuum emissionfor individual sources was not part of the main aims ofour survey, but it is nevertheless detected in brightertargets (with typical i -band magnitudes brighter than21.5 mag). The FOV of all the MOS instruments wassmaller than the coverage of our NB observations, whichgenerally resulted in a denser sampling at low cluster-centric radii and a sparser sampling at larger radii whencompared to our NB selection.Figure 2 displays representative examples of H α NBcandidates confirmed to be SF galaxies and AGN-dominated sources at the cluster redshift, together withNB, red, green, and blue postage stamps of the galaxy.Figure 1 shows the distribution of candidate sources andthose confirmed to be H α cluster members for the merg-ing cluster A2254.In this paper, we rely on the spectroscopic observa-tions mainly for measuring redshifts to confirm clustermembership, for measuring line ratios to classify galax-ies as star-forming or AGN-dominated, and for obtain-ing line EWs. Further properties available from the http://herschel.as.arizona.edu/acres/acres.html Unless specified, when used alone, [N ii ] will refer to the λ (a) SF dominated spectrum taken with WHT/AF2.(b) AGN dominated spectrum taken with VLT/VIMOS. Figure 2.
Examples of cluster member spectra. We show 15 (cid:48)(cid:48) × (cid:48)(cid:48) cutouts in the NB, red ( i band), green ( r band), and blue( g band) light. We show illustrations of the fiber and slit sizes used for each instrument. data, such as metallicities, ionization parameters, elec-tron densities, will be discussed in a forthcoming paper.3.2.1. VLT/VIMOS
We employed the multi-slit capabilities of the VIMOS instrument mounted on the UT3 telescope at ParanalObservatory. Slits can be distributed over the 4 VIMOSdetectors, for a total FOV of 4 × (cid:48) × (cid:48) . With the MRgrating in combination with the GG475 blocking filterwith 1 (cid:48)(cid:48) slits, our observations covered the 500 − . − . (cid:48)(cid:48) seeing and thin cloud conditions during August 2017. The datawere reduced using the VIMOS pipeline implemented inEsoReflex (Freudling et al. 2013). The reduction firstcreates a set of final combined biases, darks, flat fields,and arcs. Each science exposure is debiased, flat-fielded,and corrected for spatial curvature and cosmic rays andhot pixels are flagged using a bad pixel table. The skycontribution is subtracted using a local sky model, anda set of sky lines are used to obtain the wavelengthsolution, calibrated to vacuum wavelengths. We alignand combine the individual exposures pertaining to acommon pointing/setup into a single image, on whichwe perform the detection and extraction of objects. Astandard star was observed for each observing block andcalibrated in the same fashion as the science data. Thestandard star is used to estimate the response curve. Fi- (c) Composite spectrum taken with Keck/DEIMOS.(d) A composite spectrum taken with Hectospec/MMT. Figure 2.
Continued. nally, the response curve is applied to the science datato flux-calibrate the extracted spectra.3.2.2.
WHT/AF2
We observed three clusters with the AF2 instrumentmounted of the William Herschel Telescope in La Palma,Spain. AF2 has configurable 1 . (cid:48)(cid:48) fibers, which can bedeployed over a FOV of about 30 (cid:48) × (cid:48) . Observations ofthe Sausage cluster were taken in two separate fiber con-figurations in July 2014, using the R600R grating andreaching a resolution of 4.4 ˚A and were presented in So-bral et al. (2015). The rest of the data on two clusterswere taken in January-February 2017. The data weretaken in good seeing conditions of 0 . − . (cid:48)(cid:48) for most of the run. We used the R316R grating, covering the4000 − ii ] ( λ Keck/DEIMOS
Two clusters in this sample were observed with theDEIMOS instrument on the Keck telescope as part ofthe Merging Cluster Collaboration (MC ) spectroscopicfollow-up efforts (Golovich et al. 2019). While the pri-mary targets for MC were candidate passive clustermembers, in this paper, we focus on the minority ofH α candidates included in the selection. The Sausagecluster data have been presented in detail in Sobral et al.(2015), while the entire Keck sample is presented in de-tail in Golovich et al. (2019). We refer the reader tothose two papers for full details on the target selectionand data reduction. In short, each cluster was coveredwith four slit masks for a total of about 0.75 h, underexcellent seeing conditions of about 0 . (cid:48)(cid:48) . With the 1200line mm grating, a resolution of 1 ˚A is achieved over the5400 − ±
500 ˚Afrom source to source, which means some very red emis-sion lines can be missed. The data are reduced using theDEEP2 version of the SPEC2D package (Newman et al.2013), which performs the same set of steps as describedfor VIMOS (see Section 3.2.1).3.2.4.
Hectospec/MMT (through ACReS)
We leverage publicly available spectroscopy for sevengalaxy clusters through the Arizona Cluster RedshiftSurvey (ACReS, Newman et al. 2013). ACReS followedup thousands of galaxies in the field of galaxy clusters,down to a limiting K-band magnitude. Unlike our otherdatasets, which included a clear selection for emission-line galaxies, the ACReS dataset, therefore, includes amultitude of stars and galaxies, of which some are inthe cluster, and some are foreground and backgroundgalaxies. In this paper, we use only galaxies with sig-nificant emission line detections. ACReS was conductedwith the Hectospec instrument (Fabricant et al. 2005)on the MMT telescope in Arizona, which can deploy 3001 . (cid:48)(cid:48) fibers over a FOV of 1 ◦ , in combination with the270 line grating, to cover the 3650 − ii ], and allthe other lines covered in our other observations. TheMMT observations are reduced through the Smithso-nian Astrophysical Observatory Optical/Infrared Tele-scope Data Center using the Hectospec pipeline . Sim-ilar steps are conducted to the other data sets. Takingadvantage of the stable fiber response, we correct theresponse curve by deriving the pixel response and fiberthroughput from the flat fields. Note that, while a rel- ative flux calibration is applied, bringing wavelengthson the same relative flux scale within each spectrum,no absolute flux calibration with a standard star is per-formed in the Hectospec pipeline. By comparing theseven pairs of galaxies with both an MMT and a VI-MOS spectrum, we conclude this has minimal effects onour results (see also Section 4 and 8.2). The wavelengthcalibration is done to air wavelengths, unlike the rest ofthe data, but this is inconsequential for the emission linemeasurements presented in our paper.3.3. Ancillary Imagining and Photometry
We use ancillary imaging and photometry to assessour selection of targets and interpret our results. Weparticularly rely on the NB observations centered on H α at the cluster redshift, the associated i broad-band ob-servations and derived products, such as the NB and i magnitudes and NB-derived H α luminosities, as pre-sented in Stroe et al. (2017).Throughout the paper, we choose to use H α luminosi-ties derived from the NB photometry and not from thespectroscopy because of two main reasons. Firstly, theH α luminosities were derived from the photometry us-ing large 5 (cid:48)(cid:48) apertures, which encompass all the emissioneven for the largest galaxies at z ∼ . − .
3. The spec-troscopic follow-up included fibers and slits of varied sizeand shape, which can lead to slightly different estimatesof the H α luminosity. Secondly, the Hectospec/MMTobservations were not absolute flux calibrated. Overall,the H α luminosities derived from spectroscopy matchthose derived from the NB (see Figure 3), especiallyfor H α luminosity ranges of most interest in the pa-per (10 − erg s − ). After applying a single scalingfactor to the MMT luminosities (which lack flux calibra-tion), we find that spectroscopic H α luminosities have aGaussian spread with a standard deviation of 0 .
36 dexaround the expected 1:1 relation with the NB measure-ments. Hence, the NB-derived H α luminosities provideus with a clean, simple way to compare all the data,including the MMT observations. Note that all NB lu-minosities are corrected for [N ii ] contamination and fordust extinction within the host galaxy, as described inStroe et al. (2017). For a detailed description of the pro-cess of measuring properties, including H α luminositiesfrom spectroscopy, see Section 6.We also employ our own g and r data (Stroe et al.2017), and when not available, imaging and photom-etry from the Sloan Digital Sky Survey (SDSS DataRelease 12, Alam et al. 2015) or the Pan-STARRS sur- https://panstarrs.stsci.edu/ Figure 3.
The H α luminosity measured from spectroscopycorrelates well with the NB H α luminosity. The spectro-scopic luminosities follow a Gaussian distribution ( σ = 0 . − erg s). vey (PS1, Flewelling et al. 2016). In building our fig-ures, we also make use of publicly available X-ray ob-servations through the XMM-Newton and Chandra archives and 1.4 GHz radio data from the NRAO VLASky Survey (NVSS).3.4. redshifts Package and Ancillary Redshifts
While the overwhelming majority of the sources areexpected to be H α emitters at the cluster redshift, theNB filters are also sensitive to higher redshift emittersas [O ii ], [O iii ] and H β at z > . http://nxsa.esac.esa.int/nxsa-web https://cxc.harvard.edu/cda/ Figure 4.
The accuracy of our redshifts, compared to valuesfrom the literature. All available spectroscopy is included,not only H α candidates. Bottom: Our redshifts match wellwith those from the literature. Of note is the high densityof sources around the cluster redshifts, whose distributionis shown in the histogram. Top: The distribution of shiftsbetween the literature redshifts and our redshifts (note thelogarithmic scale). The bulk of the sources have matchingredshifts within 0.0005. us to further study the reliability of our NB selection ofH α candidates and to conduct a supplemental check ofthe redshifts derived from our spectroscopy.To aid in this endeavor, we introduce the redshifts package (Stroe 2020). redshifts is a Python pack-age that collects all unique spectroscopic redshifts fromthe VizieR and the NASA Extragalactic Database(NED ) online databases. With redshifts , the usercan perform a flexible search within a radius of a givenset of (R.A., Decl.) coordinates. redshifts leveragesAstroquery to use column names and descriptions (in-cluding UCD keywords) to identify columns containingspectroscopic redshifts or radial velocities. redshifts weeds out photometric redshifts and duplicates and re- https://vizier.u-strasbg.fr/ https://ned.ipac.caltech.edu/ https://astroquery.readthedocs.io/en/latest/ Figure 5. H α cluster members confirmed with spectroscopy with two independent observations from different telescopes. Theredshifts and emission line properties are in excellent agreement, despite the different telescope sizes (4, 6.5, 8, and 10-m),exposure times (0.5-2 h), observing conditions (dark time versus gray time with thin clouds), and instrument properties (1 (cid:48)(cid:48) slitsvs. 1.5 (cid:48)(cid:48) fibers, 1 − . turns a unique list of ‘best’ spectroscopic redshift mea-surements in FITS table format. redshifts uses aconfiguration file, written in YAML , a popular human-readable markup language, to specify the search radius,column names, and any VizieR ‘banned’ catalogs. Thesearch is performed from the specified (R.A., Decl.) po-sition out to the radius from the configuration file. Theuncertainty is used to evaluate whether redshifts couldbe photometric instead of spectroscopic. The ‘banned’catalogs encompass any VizieR catalogs one does notwant included in the search, for example, because theywere found to mix spectroscopic and photometric red-shifts in one column. One of the limitations of the pack-age is that it relies on the original authors to use theUCD and other column names correctly. For wide-areasearches (i.e. large radius), NED and VizieR sometimestime out. Additionally, the search requires a stable in-ternet connection and can fail if the connection is in-terrupted during the search. For details on installa- https://github.com/multiwavelength/redshifts https://yaml.org/ tion, usage, and full functionality, we refer the reader toStroe (2020) and https://github.com/multiwavelength/redshifts.For the ENISALA project, we used redshifts to col-lect available spectroscopic redshifts out to 0.7 ◦ from thecluster core, covering the FOVs of our NB observationsin their entirety. REDSHIFT MEASUREMENTSWe derive redshifts through visual inspection, primar-ily from emission line features, focusing on H α , [N ii ],H β , [O iii ] and [S ii ], and secondarily from absorptionfeatures, such as CaHK and G band absorption wherecovered, in the 1D spectra. In line with their respec-tive calibrations, vacuum wavelengths are used as a ref-erence for the VIMOS, Keck, and WHT samples, whileair wavelengths are used for the MMT sample. For com-pleteness and self-consistency, we derived redshifts forall 1D spectra available to us, including, for example,any passive galaxies selected as fillers in the VIMOS ob-servations and all galaxies in the MMT fields, many ofwhich are passive cluster galaxies and non-cluster mem-bers.1We cross-match the positions of sources in our sam-ple with the publicly available redshifts measurementsin the literature, using a 1 (cid:48)(cid:48) tolerance. Figure 4 showsthe distribution of redshifts in our sample, as well as acomparison to the literature values. We note that, interms of source numbers, the sample presented in Fig-ure 4 is dominated by the ACReS spectroscopy, whosebroad, magnitude limited sample, results in a highlycomplete cluster coverage, but also a significant contri-bution from foreground and background galaxies, andto a lesser degree, stars. However, the bottom panel ofthe figure highlights the high density of sources concen-trated around the cluster redshifts. The top panel showsa slight systematic median shift of 0 . < . SELECTION FUNCTION5.1.
How good was the NB selection?
With spectroscopy in hand, both from our work andthe literature, we can further quantify the robustness ofour NB selection for H α cluster candidates. Figure 6shows the distribution of redshifts for the NB sources.When we include spectroscopy from the literature, 65%of H α candidates are confirmed to be at the cluster red-shift. When focusing on our sample of sources for whichwe have direct access to spectroscopy, the fraction risesto 71% confirmed as H α candidates. This discrepancycan be explained as follows: when compared to our spec-troscopic campaign that prioritized bright emitters, theselection function of spectroscopy from the literature isbiased towards lower magnitudes and H α luminosities.The NB selection purity for H α is expected to drop asthe contamination from interlopers increases at fainterfluxes. This is because luminosity functions of higherredshift emission lines are steeper at fainter fluxes con-tributing more at faint fluxes than at brighter fluxes(e.g. Khostovan et al. 2015). The selection was par-ticularly poor for RXJ0437 and RXJ2129, where noneof the followed-up sources is confirmed to be a clustermember. The high fraction of M-type stars in RXJ0437 Figure 6.
Distribution of rest-frame wavelengths for thespectroscopically confirmed NB candidates. The top panelshows a zoomed-in view on the wavelength range aroundH α . We also show the transmission profiles of the nine NBfilters used to select the H α candidates, converted to the H α rest-frame (thin gray lines). We confirm the majority of NBcandidates ( > α emitters at the cluster redshifts,with the rest confirmed as either stars or galaxies at otherredshifts. (63% of candidates followed up) is most likely due to thecombination of central wavelength ( ∼ ∼
78% of galaxies areH α emitters. Overall, the spectroscopic follow-up purityin 7/13 clusters is over 80%.5.2. Distribution with cluster-centric distance
Figure 7 shows the distribution of the spectroscopicand parent samples with cluster-centric radius. Thedistribution of the underlying NB candidate sample isdriven by both physical factors (e.g. cluster influence onthe SF activity in infalling galaxies), as well as the FOVof the NB observations, which results in less coverageat large radii. In the figure, the bins are equally spacedin radius, which translates to increasing areas at largercluster-centric radii, which partially explains the largenumber of sources at 2000 − Figure 7.
Distribution of NB candidates and those withfollow-up spectroscopy with cluster-centric distance. Thespectroscopic follow-up for both relaxed and merging clus-ters more densely samples the parent population towards thecluster core, as expected. relaxed clusters, with higher rates at low cluster-centricdistances. For both merging and relaxed clusters, wefollowed-up sources within 2000 kpc at about twice therate of those at larger distances.5.3.
Distribution with NB magnitude and NB H α luminosity To maximize the chance of spectroscopic confirmation,we preferentially followed-up bright sources, with largeNB H α fluxes (typically > × − erg s cm − . Fig-ure 8 highlights this effect. Of note is the high frac-tion of sources confirmed as H α emitters at the clus-ter redshift, particularly those with bright NB magni-tudes ( < . α luminosities. Not un-expectedly, we find the most contamination from non-cluster sources, i.e. other line emitters at lower or higherredshift (see also Figure 6) and stars, happens in faintsources, especially those with small expected H α lumi-nosities. Overall, as shown in Figure 9, the spectroscopicfollow-up samples well the underlying NB population. GALAXY PROPERTIES FROMSPECTROSCOPYThe spectroscopic component of the ENISALAproject involves hundreds of cluster members of interest,which all required systematic measurements of emission and absorption lines. To this end, we used gleam ( G alaxy L ine E mission & A bsorption M odeling, Stroe& Savu 2020, 2021), a Python package for fitting Gaus-sian models to emission and absorption lines in largesamples of 1D galaxy spectra. For more details on theuse and capabilities of gleam , we refer the reader toStroe & Savu (2021), Stroe & Savu (2020) and https://github.com/multiwavelength/gleam.We use gleam to measure the continuum andemission-line fluxes, EWs, and luminosities from thespectroscopy. The data was processed, as much as pos-sible, in a uniform manner and without human interac-tion.For the bulk of the sources, we fit the same set ofemission lines, including [O ii ], H β , [O iii ], H α , [N ii ] andthe [S ii ] doublet. The fitting is done using the Leven-berg–Marquardt algorithm, as implemented by LMFIT(Newville et al. 2014), which has the benefit of beingwell-behaved, fast, and enabling an easy estimation ofuncertainties on the fit parameters (see LMFIT docu-mentation for more details). We included the uncertain-ties in the measured spectrum (the standard deviation)as weights in solving the minimization problem.We fit a Gaussian model plus a constant continuumfor each emission line. A total range of 70 ˚A on eitherside of the emission line is used for fitting the model.This range is large enough to encompass enough linefree spectrum to securely estimate the continuum with-out being affected by galaxy colors. The constant forthe continuum, as well as all Gaussian parameters (cen-tral wavelength, amplitude, and line width), are all freeparameters in the fit. However, the central wavelengthis bound to ± . α and [N ii ] and the [S ii ] doublet, re-spectively, are fit jointly. In a few sources with broadH α and strong [N ii ] ( λ α plus the [N ii ] doublet are jointly fit. Any other nearbylines outside of the 26 ˚A range are masked ( ±
20 ˚A acrossthe expected position) to not bias the continuum emis-sion estimation. Any emission lines outside the spectralcoverage are skipped. If necessary, we also mask the A( ∼ ∼ https://github.com/multiwavelength/gleam This range was 60 ˚A for the VIMOS and 40 ˚A for Keck, as thiswas found to produce more reliable results. Figure 8.
Distribution with respect to NB magnitude and H α luminosity derived from the NB. We show NB candidates andsources with spectroscopy, including confirmed H α emitters at the cluster redshift, foreground and background sources, andstars. The targets selected for spectroscopy are preferentially bright in the continuum and have bright emission lines. Onaverage, the fraction of NB candidates followed-up with spectroscopy and confirmed as H α cluster members increases for brightsources and those with large H α luminosities. gleam first attempts to fit the full model with theconstant and as many Gaussian components specified.If the fit does not converge, it simplifies the model byremoving combinations of one or more Gaussian com-ponents. The final model includes a constant for thecontinuum and all emission lines with significant de-tections, with reported upper limits for non-detections.The emission line and continuum measurements are usedto derive galaxy properties as described below in our re-sults section. As a testament to the quality of data andcalibration, we find good agreement between the spec-troscopically derived H α luminosities and those derivedfrom the NB (see Figure 3). FINAL SAMPLEThe entire sample of sources with spectroscopyamounts to over 4200 sources, of which 381 we obtainedthrough our targeted VLT/VIMOS, Keck/DEIMOS,and WHT/AF2 follow-up. The rest of the sample, fromthe ACReS program, included many stars, foregroundand background sources, as well as passive cluster mem-bers that do not lie at the focus of our current project.While our VLT/VIMOS, Keck/DEIMOS, andWHT/AF2 follow-up represents a strict sub-sampleof the H α NB candidates, the nature of the ACReSselection (effectively a mass-limited selection) included many sources that were not identified as H α candidatesin our NB survey. Such sources didn’t pass our origi-nal NB selection criteria, either because they were toofaint in the BB (i.e. faint continuum) or because theirNB-derived EW did not pass the 3 sigma threshold (i.e.faint emission line). For our study, we only includesources from Hectospec/MMT that pass the same se-lection criteria as our VLT/VIMOS, Keck/DEIMOS,and WHT/AF2 follow-up sources, i.e. have significantdetections of emission lines and have NB measurementsof the H α luminosity (both ensuring bright emissionlines).We include sources in the sample if their redshift cor-responds to the cluster redshift within a narrow range.We thus select sources whose redshift falls within theredshift range covered by the 2 times the full-width-half-maximum (FWHM) of the NB filter used for each clus-ter, transformed into the rest-frame of H α . For example,the central wavelength of the NB236 NB filter is 7839 ˚Awith a FWHM of 110 ˚A. Within 2 FWHMs, the filtertraces H α at redshifts between 0 . < z < . Figure 9.
The magnitude-H α luminosity distribution of confirmed cluster H α and non-cluster sources, compared to theunderlying NB candidate pool. We note that our spectroscopy samples well the underlying population. a second criterion of having strong emission lines, whichensures the removal of passive ellipticals and post-starburst galaxies from the final sample. The criteriafor selecting the final sample of sources are guided bydetections or upper and lower limits in H α , [N ii ], H β and [O iii ] which enable the classification of a source us-ing the BPT diagram. The selection criteria are detailedin Appendix A.To ensure comparable samples between our merg-ing and relaxed clusters, we also focus our analysis onsources with NB H α luminosities between 10 and 10 erg s − , located within 4500 kpc from the cluster center(see Section 8.3 and Figure 13).In summary, a source must pass a set of criteria tobe included in the final ENISALA sample used in ouranalysis. The criteria are as follows: • Source must be located at cluster redshift, • Source must have emission line detections, • Source must be classifiable in the BPT diagram, • Source must have NB H α luminosity in the 10 − erg s − range, • Source must be located within 4500 kpc of the clus-ter center.There are 818 sources with detections in at least oneemission line besides H α . Much of our analysis relieson classifying sources based on their ionization source,which restricts the sample to 640 galaxies (see Sec-tion 8.2). The total number of sources classified in theBPT diagram with luminosities > erg s − is 451.The full breakdown of the sample can be found in Ta-ble 3.5 Table 3.
The number of emission-line galaxies in the sample, in-cluding those classified as star-forming, AGN, and composite in theBPT diagram. We display the number of sources with NB H α lu-minosities greater than 10 erg s − , with total numbers of sourceslisted in parentheses. We also show the samples divided by clustertype and by the telescope of origin.Sample All SF AGN CompositeAll 451 (818) 275 (414) 43 (85) 87 (181)Merging clusters 233 (311) 143 (168) 23 (32) 43 (62)Relaxed clusters 218 (507) 132 (246) 20 (53) 44 (119)VLT/VIMOS 77 (84) 52 (55) 5 (6) 11 (12)WHT/AF2 65 (67) 42 (44) 7 (7) 9 (9)Keck/DEIMOS 34 (44) 20 (25) 4 (6) 8 (8)MMT/Hectospec 275 (623) 161 (290) 27 (66) 59 (152)8. RESULTSOur spectroscopic observations enable us to measureemission-line properties, which, in turn, unveil the phys-ical properties of the cluster galaxies, including their SFand ionization sources. In combination with the spec-troscopic observations, we will use, where relevant, NBmagnitudes and H α luminosities derived from the NBdata (see also Figure 3) to interpret our observations.While we focus our analysis on cluster members withH α luminosities in the 10 − erg s − range, locatedwithin 4500 kpc of the cluster center, and with BPT clas-sification in the BPT diagram, we tested the effect ofusing different selection cuts on our sample, such as im-posing tighter restrictions to the BPT classification (seeSection 8.2), including using a range of H α luminos-ity cuts (e.g. Section 8.3) and imposing minimum EWs(e.g. Section 8.5), and found that our results are robustagainst different selection criteria.8.1. Distribution with Magnitude and H α Luminosity
By focusing on our NB selected sample, we unveil thedistribution of sources in the magnitude-H α luminos-ity plane (Figure 9). We find that our NB selectionidentified H α candidates on two main tracks. The firstrepresents a sequence in which brighter galaxies have,on average, brighter emission lines. Simply, this can beinterpreted as a result of the stellar mass - SF relation.Above this relation, there is a second cloud of NB can-didates with continuum magnitudes between 19 and 22,which have, on average, NB-derived H α luminosities 3times larger than galaxies on the sequence. If located atthe target redshifts, galaxies in the cloud would displaylarge EWs, as a result of brighter H α luminosities ontop of a relatively faint continuum. The vast majority of sources located on the sequenceare confirmed H α galaxies at the cluster redshift. Bycontrast, half of the sources in the cloud are interlop-ers, either stars or other line emitters as lower or higherredshift. The other half of the sources are H α emitters,almost exclusively in merging galaxy clusters.Therefore, the spectroscopic follow-up reveals a fasci-nating dichotomy in the distribution of confirmed H α line emitters. Merging galaxy clusters host a pop-ulation of H α emitters with large EWs, that isvirtually absent in relaxed clusters. Figure 10.
We classify our galaxies as dominated by SF (bigpurple circle), dominated by AGN emission (pink squares),or composite with contributions from H ii regions and AGN(small orange circles), by using the Kauffmann et al. (2003)(dark gray line) and Kewley et al. (2001) (light gray line)relations. We classify 680 sources, of which pure star-forminggalaxies represent 61% of the sample, 26% are composite, and13% are dominated by AGN emission. SF versus AGN
We use the [N ii ]/H α and [O iii ]/H β ratios as diagnos-tics for placing H α emitting galaxies in the BPT diagram(Baldwin et al. 1981). We distinguish between differ-ent ionization mechanisms for nebular gas and classifysources as dominated by SF, AGN activity, or as com-posite, through the widely adopted Kauffmann et al.(2003) and Kewley et al. (2001) dividing lines. Kauff-mann et al. (2003) encompasses typical SF galaxies,while the bulk of sources above Kewley et al. (2001) areSeyfert galaxies with strong AGN contributions. Galax-6ies located between the two lines are composite, withcontributions from both H ii regions and AGN.No dust extinction was applied to the line ratios,which has the effect of slightly raising the [O iii ]/H β ra-tio (by ∼ .
06 dex, Sobral et al. 2015) and increasing thechance galaxies cross the Kewley et al. (2001) line intothe AGN regime. For two reasons, the lack of absoluteflux calibration in MMT observations does not signifi-cantly affect line ratios. Firstly, the line pairs enteringratios are near each other in wavelength. Secondly, wedouble-checked for any systematic biases by comparingthe line ratio measurements for the six galaxies withMMT and VIMOS observations and found them to beconsistent with each other. Any offset in line ratio doesnot change the classification of the galaxies in the BPTdiagram.Overall, the BPT classification enables us to includegalaxies with detections in all emission lines and use sig-nificant upper and lower limits on the line ratios wherethey enable the secure classification of sources. Basedon these criteria alone, we were able to classify a to-tal of 524 galaxies. For some sources, the classifica-tion could not be secured based on the [N ii ]/H α and[O iii ]/H β alone. Examples include sources where the[O iii ]/H β line ratio cannot be constrained (e.g. lack ofcoverage of H β , H β and [O iii ] not detected) or sourceswhere the upper or lower limits do not securely place thegalaxy in the AGN or SF quadrant of the BPT diagram.For such sources, we also explore the cloudy model-ing presented in Sobral et al. (2018) and Sobral et al.(2019), to use the [O iii ]/H α ratio in combination withthe [N ii ]/H α ratio to classify the galaxies as SF, AGN, orcomposite (see Figure B1). While the demarcation be-tween AGN, composite and star-forming sources is notas clear as in the BPT diagram, the [O iii ]/H α ratio de-pends on the ionization potential, and, as such, sourceswith low [O iii ]/H α and low [N ii ]/H α are typically pow-ered by SF, while a hard AGN ionizing spectrum wouldlead to high [O iii ]/H α and high [N ii ]/H α ratios. Addingthe [O iii ] criterion particularly helps in classifying star-forming galaxies with very low [O iii ]/H α ratio, where[O iii ] and H β are not detected. Including the [O iii ]/H α criterion enables us to classify an additional 156 sources.We note that the results presented in the paper are notaltered, but strengthened with the smaller error barsprovided by the increased number statistics.Figure 10 shows the distribution of our sample withinthe BPT diagram. With a total of 680 galaxies, 414are dominated by SF (61%), 85 are AGN (13%), andthe rest (181, 26%) have composite spectra. As we willfocus mostly on sources with H α luminosities greaterthan 10 erg s − , the distribution is as follows: 275 Figure 11.
The fraction of AGN increases, while the SFfraction decreases with velocity FWHM of the H α narrowcomponent. dominated by SF (68%), 43 AGN (11%), and 87 com-posite (21%), for a total of 405 sources. Beyond theline ratios, the AGN contribution can also be clearlyseen in the velocity FWHM of the narrow H α compo-nent of our sources. The fraction of sources with AGNcontribution (including AGN-dominated and compositesources) increases with H α velocity width (Figures 11and 12). This is in line with theoretical expectationsand large surveys, which find that the narrow lines forSeyfert 2 sources range between 200 and 900 km s − ,peaking at 350 −
400 km s − (e.g. Osterbrock & Mathews1986). We test for differences in the star-forming galaxyand AGN distributions with velocity width using a two-sample Kolmogorov-Smirnov (KS) test. For the mergingand relaxed cluster samples independently, we reject thenull hypothesis that there is no difference between theAGN and SF-dominated samples at the 95% confidencelevel . When combining data from all 14 clusters, wereject the null hypothesis at the 99.6% level . The sig-nificance increases when we focus on the 0 −
600 km s − range, where the bulk of the sample is located. The dis-tribution of star-formers and AGN differ at the 99.994%level for relaxed clusters, at the 97.0% level for merg- Provides the same confidence as a 2 σ significance level for a nor-mal distribution. Equivalent to a 2 . σ significance level for a normal distribution. for the combined sam-ples. Within the 0 −
600 km s − range, the distribution ofstar-forming galaxies in merging clusters is statisticallydifferent from relaxed clusters at the 99.7% confidencelevel . Figure 12.
Distribution with H α velocity FWHM, sep-arated by ionization source and galaxy cluster dynamicalstate. Samples are matched in H α luminosity and cluster-centric distance. The velocity width distribution of star-formers is statistically distinct from that of AGN. The bulk ofstar-forming galaxies, in both mergers and relaxed clusters,have narrow lines. On average, AGN have broader velocitywidths, extending above 400 km s − . Radial Distribution with H α Luminosity
We show the distribution of our ENISALA samplewith H α luminosity and cluster-centric distance in Fig-ure 13. Our survey covers four orders of magnitudein H α luminosities (10 − erg s − ) and stretchesout to cluster-centric radii of over 5 Mpc. The bulk ofsources have H α luminosities above 10 erg s − , with SFgalaxies, AGN, and composite sources spanning the en-tire range of luminosities shown. The AGN dominatedsources towards the cluster cores have the brightestemission lines, above 10 erg s − . Our sample includesAGN, star-forming galaxies, and composite sources atall luminosities, and cluster-centric radii, irrespective ofcluster dynamical state (merging or relaxed). Equivalent to a 4 . σ significance level for a normal distribution. Equivalent to a 2 . σ significance level for a normal distribution. Figure 13.
Distribution of sources powered by SF or AGN,as a function of NB H α luminosity and cluster-centric radius.Squares mark sources in merging clusters, and circles repre-sent sources in relaxed clusters. Star-forming galaxies aremarked with purple, AGN with pink and composite sourceswith orange. The luminosity and radius distribution for dif-ferent ionization types reflect the broad selection of sources inboth merging and relaxed clusters. We fully match the sam-ples for the analysis by selecting only sources within 4500 kpcand with H α luminosities in the 10 − erg s − range. The fraction of sources dominated by each ioniza-tion type strongly depends on the radial distance fromthe cluster core and the H α luminosity of the source(Figure 14). In building this plot, we focus on 10 − erg s − range and on radii < < σ , the difference isnot statistically significant. Interestingly, bright emis-sion lines are more often powered by SF at the outskirtsof relaxed clusters (3 − . . σ ) and 93% (1 . σ ) con-fidence levels, respectively. By combining the bins withsources located at < . σ . While a linear model mightnot provide the best description of the data, it is stillinstructive in comparing the radial dependence of the8 Figure 14.
The fraction of pure star-forming galaxies andsources with AGN contribution, as a function of cluster-centric distance. Samples are matched in H α luminosity be-tween 10 and 10 erg s − . In merging clusters, the fractionof line-emitters powered by SF is constant with radial dis-tance. In relaxed clusters, the fraction of purely star-forminggalaxies drops within 3 Mpc of the cluster core. The AGNand SF fractions at relaxed cluster outskirts match thosein the field (e.g. Kauffmann et al. 2003; Shioya et al. 2008;Hayashi et al. 2018). star-forming fractions in mergers and relaxed clusters.In merging clusters, the relationship is consistent withno radial dependence of the star-forming fraction, witha slope of − . ± . − and an intercept of0 . ± .
02. We find that the star-forming fraction mildlycorrelates with radial distance in relaxed clusters, witha slope of 0 . ± . − and an intercept of0 . ± .
13. A Pearson correlation analysis also indi-cates that relaxed clusters (Pearson r = 0 .
81) presenta slightly stronger linear correlation between the frac-tion of star-forming galaxies and radial distance thanmergers (Pearson r = − . . σ in all but the outermost binin relaxed clusters. Therefore, the outskirts of relaxedclusters have AGN and SF fractions similar to averagecosmic fields (80 : 20, as per Kauffmann et al. 2003; Sh-ioya et al. 2008). For mergers, the chance of a galaxy being powered by SF is constant at all cluster-centricradii. The star-forming fraction drops in relaxed clus-ters, from a value similar to an average field beyond3 Mpc to about 65% (or factor of 1.2) within 3 Mpc. Figure 15.
Normalized probability distribution withcluster-centric distance. Only sources with H α luminositiesbetween 10 and 10 erg s − are included. We separatesources into four classes, based on their parent cluster typeand the powering source of the H α emission (SF or AGN). The radial distribution of star-forming galaxies andAGN is different between relaxed and merging clus-ters.
The bulk of SF and AGN activity in merging clusters ishappening within 3 Mpc of the cluster core. By contrast, thebulk of emission-line galaxies in relaxed clusters are locatedoutside the core ( > . Another way to look at the radial distribution of ion-ization types is to study the probability distribution de-pendence on cluster-centric radius. We split the sampleinto four categories as a function of ionization source(pure SF versus AGN contribution) and galaxy clusterdynamical state (merging versus relaxed). In Figure 15,we show the distribution of each population as a functionof cluster-centric radius (i.e. what fraction of each pop-ulation is located in each radial bin). We employ a KSand a two-sample Z-score methodology to test whetherthe radial distribution of sources is different as a func-tion of ionization source and cluster relaxation. Witha KS test, we reject the null hypothesis that the radialdistribution of star-forming galaxies is the same betweenrelaxed and merging clusters at the 99.99% confidence9level . A two-sample Z-test on the binned data givesa similar statistical significance of 3 . σ when combiningthe significance of each pair of radial bins using Fisher’smethod. We also find mild evidence that the AGNdistribution differs between mergers and relaxed clus-ters, at the 93.2% confidence level with a KS test and2 . σ level using the Z-test in combination with Fisher’smethod.For mergers, about 40% of star-forming galaxies arelocated within 1.5 Mpc of the cluster core, another 40%are located between 1 . − < . σ ) of the star-forming pop-ulation in merging clusters. The lower fraction of AGNand star-forming galaxies beyond 3 Mpc is statisticallydifferent from the two innermost bins at a significancelevel of 2 . . σ , for AGN and star-forming galaxies,respectively.Our data paint a different picture for emission-linegalaxies in relaxed clusters. The radial distribution ofstar-forming galaxies and AGN in relaxed clusters is dif-ferent, albeit with a relatively low confidence level of91.3% . Only 20% of purely star-forming galaxies inrelaxed clusters are located within 1.5 Mpc of the core,and over 30% are outside 3 Mpc. Using a one-sampleZ-test, the increased fraction of star-forming galaxiesbetween 1 . − . . σ level. Over 60% of line-emitters powered, at least inpart, by AGN are located at the outskirts of relaxedclusters (between 1.5 and 3 Mpc), a fraction that is sta-tistically higher than other cluster regions (at 3 and4 . σ , respectively). The bulk of star-forming galax-ies and AGN in the fields of merging clustersare distributed towards the core of merging clus-ters (within 3 Mpc), while SF and AGN activitymost likely occurs at relaxed cluster outskirts( . − Mpc).
Distribution with Galaxy Color
We investigate how the ionization source of the H α emission relates to the properties of the host galaxy (see Equivalent to a significance of 3 . σ for a normal distribution. Equivalent to a significance of 1 . σ for a normal distribution. Equivalent to 1 . σ Figure 16). Specifically, we employ the observed g − r color, which traces the rest-frame UV–optical color forthe redshift of our sources. This range encompasses theBalmer and 4000 ˚A breaks and enables a broad separa-tion of star-forming and passive galaxies. By focusing ongalaxies with i -band magnitudes between 17 and 22 andmatching samples in H α luminosity (10 − erg s − )and cluster-centric distance ( < Figure 16.
Normalized distribution with g − r galaxy color.We compare the distribution of sources in our sample, sep-arated by ionization source (purple for SF versus pink forAGN) and galaxy cluster type (filled histogram for merg-ers versus hatched histograms for relaxed clusters). Samplesare matched in magnitude, luminosity, and cluster-centricdistance. AGN hosts have redder colors, while star-forming galaxies are on average bluer, especially inmerging clusters.
We split our sample by cluster dynamical state(merger versus relaxed) and ionization type (pure SFversus AGN-dominated or composite). A KS test con-firms that the color distributions of AGN and star-forming galaxies are different at a high significance level.We reject the null hypothesis at a very high significancefor the relaxed, merging, and combined sample (greater0than 99.99999% ). On average, line emitters poweredby SF are by, an overwhelming margin, blue: over 80% ofstar-forming galaxies in both relaxed and merging clus-ters have colors bluer than 0.9. By contrast, sourceswith AGN and composite spectral features are redderthan SF galaxies.Overall, according to a KS test, the color distribu-tion of star-forming galaxies in merging clusters is dif-ferent from relaxed counterparts at the 96.3% confidencelevel . Further, a higher fraction (40 . ± < ± . σ level. By con-trast, the color distribution of AGN is consistent be-tween relaxed and merging clusters.8.5. H α Equivalent Widths
As mentioned in Section 3, we cannot measure H α luminosities, and hence SFRs, directly from our spec-troscopy for the entire sample. We can, however, mea-sure the rest-frame EW of the H α line. The EW tracesthe strength of the emission line on top of the continuumand thus broadly relates to the sSFR of the host galaxy,as long as the H α emission is powered by SF and dustextinction does not vary significantly between galaxies.The average EW drops with increasing (redder) rest-frame UV–optical color for all ionization sources, includ-ing pure star-forming galaxies, those with AGN contri-butions, and those whose H α is dominated by AGN.This effect is exhibited in Figure 17, where we matchthe samples in broad-band magnitude and H α luminos-ity. On average, at all rest-frame colors, the EW ofstar-forming galaxies is larger than AGN and compos-ite sources by a factor of 1 . −
2. We tested this hy-pothesis using a two-sample KS test, a two-dimensionaltwo-sample KS test (also known as Peacock’s test, Pea-cock 1983) , and a two-sample Student’s t -test. Therest-frame EWs distribution of SF and AGN sources isdifferent for both the relaxed and the merging clustersample. A KS test indicates that the null hypothesiscan be rejected at > . confidence level for bothrelaxed and merging clusters. The two-dimensional KStest in the rest-frame EW–color space yields the sameconclusion at a similar confidence level. The mean EW Equivalent to a 5.7, 6.2, and 7 . σ significance level for the merg-ing, relaxed and entire sample, respectively, when using a normaldistribution. Equivalent to 2 . σ for a normal distribution. Using the 2DKS Python implemention at https://github.com/Gabinou/2DKS. Equivalent to > σ for a normal distribution. Figure 17.
Distribution of sources with EW and observed g − r color (tracing approximately UV–optical rest-frame),separated by ionization source and cluster dynamical state.The samples are matched in i -band magnitude and H α lu-minosity. Star-forming galaxies have higher EW than AGN,irrespective of galaxy color. Star-forming galaxies inmerging clusters have higher EW, or sSFRs, com-pared to counterparts in relaxed clusters. is greater for SF-dominated sources compared to thosewith AGN contributions, for both relaxed clusters andmerging clusters, with a t -score of 6 . .
3, respec-tively, and a high significance level of > . > . . The two-dimensional distribution ofstar-forming galaxies in the EW–color space is statisti-cally different between relaxed and merging clusters ata 99.61% confidence level . A t -test confirms this find-ing and indicates that the mean EW for star-forminggalaxies in merging clusters is higher than in relaxedclusters ( t -score of 3.3). By contrast, in the case ofAGN, a one-dimensional KS test on EW distribution, Equivalent to > . σ for a normal distribution. Equivalent to 2 . σ for a normal distribution.
1a two-dimensional KS test in the EW–color space, anda t -test confirm that the distribution of EWs, as well asthe mean EW, do not differ between relaxed and merg-ing clusters.The ENISALA sample has a wide diversity of rest-frame H α EW properties. We note that the depth ofeach observation, the strength of the emission line incombination with the strength of the continuum, de-termine the limiting EW measurement. EW inverselycorrelates with stellar mass, and, in field samples at red-shifts similar to our cluster z ∼ . − .
3, star-forminggalaxies with stellar masses > M (cid:12) have rest-frameH α EWs of 10 −
40 ˚A (Fumagalli et al. 2012). In linewith massive field galaxies, the bulk of our sources haverest-frame EWs under 50 ˚A, with AGN measuring thesmallest EWs. The distribution of pure star-forminggalaxies extends to higher EW, over 100 ˚A, and up to250 ˚A, indicative of galaxies with masses of ∼ M (cid:12) and < M (cid:12) , respectively.By matching the samples in NB H α luminosity, wecan observe that the fraction of source powered by SFincreases with increasing EW, especially for the sourcesin relaxed cluster fields (see Figure 18). In merging clus-ters, over 60% of sources with low EW below 10 ˚A arepowered in part by AGN. This fraction drops sharplyto 20 −
30% for EW >
10 ˚A. Conversely, for relaxed clus-ters, the fraction of AGN-powered sources drops from ∼
50% to 0 with increasing EW. In bins of EW, the frac-tion of sources powered by SF or AGN as a functionis consistent between cluster dynamical states. How-ever, it is noteworthy that the distribution of EWs inthe two samples are vastly different in our luminosity-matched samples. Relaxed clusters in our ENISALAprojects contain more low EW sources. By contrast,merging clusters have an overabundance of sources withlarge EWs, reflecting the preponderance of highly star-forming galaxies.To further identify the nature of galaxies with largeH α EW, we plot the distribution of sources with i -bandmagnitude versus the H α luminosity measured from theNB (Figure 19). Naively, sources with large EW likelyhave bright emission lines (i.e. large NB H α luminosity)on top of a faint continuum (i.e. large i magnitude). Wethus expect large EW sources to reside in the top-rightquadrant of the plot.In Figure 20, we plot the rest-frame H α EW distri-bution of emisison-line sources with faint continuum(i.e. we select sources with i -band magnitudes fainterthan 20 mag, with 10 − − erg s − H α luminosi-ties, which are located within 4.5 Mpc from the clustercenter). Overall, for a similar total number of emissionline galaxies in each sample, merging clusters are > Figure 18.
Fraction of targets powered by each ionizationsource, binned by EW. The samples are matched in H α lu-minosity. At the lowest EWs <
10 ˚A, the emission is poweredby AGN, at a higher rate in merging clusters compared torelaxed clusters. Beyond 10 ˚A, the emission is powered bypure SF in 70 −
80% of cases. times more numerous in faint-continuum emitters thanrelaxed counterparts (66 versus 21 sources). This effectis caused by a genuine paucity of sources with large EWsin relaxed clusters and not because of a lack of spectro-scopic follow-up in parts of the magnitude-luminosityplane. As shown in Figure 9, many galaxies were se-lected for follow-up from the space populated by opti-cally faint, H α bright sources. However, spectroscopyconfirmed that many of these sources were not at thecluster redshift, including a majority of sources hostedby relaxed clusters.We confirm that a large fraction of the faint-continuum population consists of star-forming galaxiesand a few AGN with large H α EWs measured from thespectroscopic observations, exceeding 80 −
90 ˚A (Fig-ure 20). Within the population of faint-continuumsources, merging clusters have 13 sources with EWs >
90 ˚A, while we find a single such source in relaxedcounterparts. An E-test confirms that the occurencerate of high EW emission-line sources is higher in merg-ing clusters than in relaxed clusters at the 99.93% con-fidence level . Galaxy clusters undergoing mas-sive mergers contain a population of highly star-forming galaxies with high sSFR, which is absentfrom relaxed clusters. Equivalent to 3 . σ for a normal distribution. DISCUSSIONWe specifically designed the ENISALA project tounveil the physical mechanisms through which clustergrowth drives galaxy and BH evolution. With an exten-sive sample of over 800 high S/N spectra of H α -selectedgalaxies, of which we securely classify 680 as SF, AGN,or composite, we find striking differences between star-forming galaxy populations in merging and relaxed clus-ters.The distribution of emission-line properties withcluster-centric distance reveals some interesting trends,which differ between relaxed and merging clusters. In asample of emission-line galaxies matched in H α luminos-ity, the bulk of emission-line galaxies are located outsideof the 1.5 Mpc radius from the cores of relaxed clusters.The fraction of AGN drops outside 3 Mpc for relaxedclusters, and, conversely, the fraction of sources poweredby SF drops inside the cluster, encoding tell-tale signsof environmental quenching of SF. Overall, as shown inPaper I (Stroe et al. 2017), the density of H α line emit-ters for the relaxed clusters in the ENISALA sampleis lower, resulting in an overall smaller number of star-forming galaxies towards their cores, compared to merg-ing clusters. This effect could be reproduced if burstsof SF are triggered in gas-rich galaxies by ram pressureat the infall region of relaxed clusters, which, in turn,power bright emission lines, which rapidly fade as thegalaxy further approaches the cluster core. Any brightemission lines close to the cluster core would then onlybe sustained through AGN activity, hinting at an almostcomplete lack of vigorous SF in relaxed cluster cores. Interms of relative abundance, AGN are most abundantbetween 1 . − z < . z ∼ g − r redder than 1.2 mag.While this present study does not focus on morphol-ogy, we speculate this population might be related tothe population of red spirals found at intermediate localdensities nearby galaxy groups and clusters (e.g. Bam-ford et al. 2009). In explaining their origin, Bamfordet al. (2009) suggests that red spirals might be low-masssources, recently accreted onto the cluster, which aftera fast removal of their gas supply will eventually evolvein S0s.Surprisingly, a higher fraction of star-forming galaxiesin merging clusters have very blue colors compared tocounterparts in relaxed clusters. Invoking older stellarpopulations for star-forming galaxies in relaxed clusterswould translate into redder average colors. An alter-native interpretation comes from Sobral et al. (2016),who find that star-forming galaxies in a z ∼ . α EWs, irrespective of galaxy color. More generally, thebulk of star-forming galaxies with large EW ( >
50 ˚A) arefound in merging clusters, surprisingly, within 3 Mpcof the cluster center and likely embedded in hot ICMplasma. This readily supports a scenario in which star-forming galaxies in merging clusters have large sSFR.Other studies have unsuccessfully searched for variationsin sSFRs across the shock fronts in merging clusters(Chung et al. 2009b; Shim et al. 2015). These studiestargeted individual clusters, did not benefit from spec-troscopic observations, and relied on mid-infrared colors3
Figure 19.
Distribution of the ENISALA sample with H α luminosity and i -band magnitude, highlighting in color the spec-troscopic H α EW. Note the logarithmic scaling of the color bar.
Merging clusters host a population of continuum faintstar-forming galaxies with bright H α emission lines, confirmed with large spectroscopic EW or sSFRs, whichdoes not exist in relaxed clusters. to estimate sSFR, which are more susceptible to con-tamination from foreground and background interlopers,and AGN. By contrast, the ENISALA project benefitsfrom a large sample and contains hundreds of confirmedcluster members in different clusters, with EWs (andthus sSFRs) measured from spectroscopic observations,facilitating more conclusive results.The most consequential finding of this paper mightbe the existence of a star-forming population uniqueto merging clusters. We discover star-forming galax-ies with large sSFR (EW ∼ − i ∼ − α EW from Belfiore et al. (2018), the H α EWsimply sSFRs ranging between 0 . − . − . Assum-ing stellar masses anywhere between 10 and 10 M (cid:12) would place these high EW galaxies 0 . − . z ∼ . . − . − . < z < . − . M (cid:12) , up to − . M (cid:12) (Erfa-nianfar et al. 2016). Furthermore, the faint star-forminggalaxies with large EWs also display lower [N ii ]/H α ratios (0 . ± .
01) compared to lower EW galaxieswith similar i -band and H α luminosity sources, implyingmetal-poor gas consistent with low-mass, highly star-forming galaxies. This effect is illustrated in Figure 21,where we show the average stacked spectrum for differ-ent EW bins. The 68% confidence intervals plotted inFigure 21 are obtained through a bootstrapping method.To sustain this high level of SF and AGN activity, alarge supply of gas is necessary. Even in relaxed clus-ters, infalling jellyfish star-forming galaxies have higherSFRs fueled by large molecular gas reservoirs (Morettiet al. 2020). The authors attribute the increased molec-ular gas fractions to the efficient conversion of neutralinto molecular gas under ram pressure. Unlike relaxedclusters, where cluster star-forming galaxies become in-4 Figure 20.
EW distribution of faint-continuum line-emitters. We show sources with H α luminosities between10 and 10 erg s − , i -band magnitudes >
20, locatedwithin 4500 kpc from the cluster center. Merging clus-ters host faint-continuum line-emitters, especially those withbright emission lines, at a higher rate than relaxed clusters. creasing deficient in neutral hydrogen towards the clus-ter core (e.g. Chung et al. 2009a), in Stroe et al. (2015a),we discovered that star-forming galaxies in the massive,binary merging ‘Sausage’ cluster (part of the ENISALAsample) have large reservoirs of neutral hydrogen, com-parable to counterparts in the field around the cluster.Jaff´e et al. (2012, 2016) find a strong correlation be-tween substructure and the presence of neutral gas-richgalaxies, supporting a post-processing scenario in whichram pressure, possibly increased by shock waves, cantrigger SF. Cairns et al. (2019) uncover a large popu-lation of galaxies rich in molecular gas in a disturbedlow-redshift cluster, reminiscent of gas-rich galaxies inyoung clusters at z ∼ . α line emitters, comparedto relaxed counterparts. The results are even more pro-nounced in individual clusters. In a pilot study usingthe NB technique to uniformly select star-forming galax-ies inside and around merging clusters through their strong H α emission, Stroe et al. (2014, 2015b) founda spectacular increase of a factor of 20 in the densityof star-forming galaxies a massive merging galaxy clus-ter (which is part of the ENISALA project), attributedto cluster-wide shock-induced SF or collapsed filamentsand groups rich in star-forming galaxies. The clusterunderwent a very recent massive merger about 0.5 Gyrago, which induced a cluster-wide, large-scale shock-wave, which passed through the cluster galaxies andpossibly triggered SF, thus effectively elevating their SFefficiency in the last 0.5 Gyr. In a study of 105 clusters,Yoon & Im (2020) find that the fraction of star-forminggalaxies is enhanced by a factor of 1.2 in interactingclusters compared to relaxed clusters, with the mostprominent effect happening in possibly gas-rich galax-ies with stellar masses < . M (cid:12) . The harsh ICM inrelaxed clusters affects all infalling galaxies to a certaindegree, resulting in a complete shutdown on SF withinone crossing of the cluster. In a study of over 100 SDSSnearby clusters, Cohen et al. (2014) corroborate these re-sults. Cohen et al. (2014) attribute the weak correlationbetween cluster substructure and SF fraction to eithercluster mergers enhancing the SF in cluster galaxies or tothe less evolved nature of mergers compared to relaxedclusters. This interpretation is echoed by other authors,who find a clear overabundance of blue and star-forminggalaxies in clusters exhibiting substructure, with mostsources tracing infalling sub-clusters (Wang et al. 1997;Cortese et al. 2004; Hou et al. 2012; Cava et al. 2017).These studies might support a scenario in which only afraction of galaxies lose their gas supply upon infallinginto the ICM of a merging cluster, while a number man-age to hold onto their gas reservoirs. The results fromthe spectroscopic analysis of the ENISALA sample tella different story, which suggests SF is triggered in merg-ing clusters, with sSFRs in few dozen galaxies exceedingvalues expected from the main sequence relation. Sev-eral scenarios that partly explain the observations fromthe literature: • Scenario 1 : Cluster shocks/turbulence trigger ac-tivity in gas-rich cluster galaxies, which would im-ply cluster-wide effects and marked differences be-tween merging and relaxed clusters. • Scenario 2 : Active accretion of groups/filamentsrich in star-forming galaxies would show an in-crease in SF activity in merging/young clusters,as they are located in active cosmic web nodes. • Scenario 3 : Fast and localized processes such asram pressure, with pronounced effects at the out-skirts of all clusters, irrespective of merger state.5
Figure 21.
Stacked average spectra for faint star-forming galaxies in merging clusters, binned by rest-frame H α EW. The68% confidence intervals are plotted. We only show sources with i -band magnitudes >
20 and H α luminosities between 10 and 10 erg s − . The average [N ii ]/H α ratio decreases with increasing H α EW. From the low [N ii ]/H α ratios, we can infer thatthe unique population of faint galaxies with high EW (sSFRs) in merging clusters have very low metallicities. • Scenario 4 : Slow, cluster-wide process, where SFslowly quenches because of a lack of new gas sup-ply, with the strongest effects in relaxed clusters.Overall, the detailed spectroscopic analysis of theENISALA survey enables us to break degeneracies be-tween the models and lend support for Scenarios 1 and 2.We unveil that the overwhelming majority of emission-line galaxies with H α luminosities above 10 − powered by either pure SF or AGN activity populatethe central 3 Mpc region of merging clusters. Merg-ing clusters also host highly star-forming galaxies deepwithin their hot ICM plasma. Our results are readilyexplained by the model proposed by Ebeling & Kalita(2019), in which gas-rich galaxies infall along filamentsin the rich cosmic web around merging clusters, followedby triggered SF. Merging clusters are permeated bycluster-wide shocks and turbulence traveling at speeds of1000 − − (Stroe et al. 2014; van Weeren et al.2019), which can act as a catalyst for triggering SF ingas clouds, as well as funnel gas into the galaxy core,causing accretion onto the BH and promoting AGN ac-tivity (as per Poggianti et al. 2017; Ricarte et al. 2020).Stroe et al. (2015a) and Roediger et al. (2014), for exam-ple, posit that a shock passing through a gas-rich galaxyshould lead to a sharp rise of SF for up to 100-500 Myr,which is perfectly compatible with the high sSFR galax-ies we find exclusively in merging galaxy clusters. Bycontrast, we find evidence for mild enhancement of AGNand SF at the outskirts of relaxed clusters, followed bya dearth of SF towards their core. Our observationscorroborate the extensive literature in the field, whichinvokes mild ram pressure triggering activity at large radii and rapid quenching taking over as the infallinggalaxy crosses the cluster core. SUMMARYThe ENISALA project is a multiwavelength observingcampaign exploring the evolution of galaxies in merg-ing and relaxed clusters. Here, in Paper II of the se-ries, we introduce the spectroscopic follow-up surveyof star-forming galaxies and AGN in 14 relaxed andmerging massive (0 . − . × M (cid:12) ) galaxy clusters at0 . < z < . α line-emitters in merg-ing and relaxed cluster environments and constrain evo-lutionary pathways from the perspective of large scalestructure growth. Our main findings are: • Our cluster emission-line sample comprises about16% AGN-dominated sources, the majority ofwhich dominated by narrow-line (hence classify-ing them as Seyfert 2 sources), with an additional20% of source with composite spectra, in line withstudies of H α selected emission-line galaxies in thefield. The fraction of emitters powered by AGNincreases sharply with the velocity FWHM of H α ,with over 80% of star-forming galaxies measuringprofiles under 200 km s − . • Pure star-forming galaxies in merging clusters,have on average, bluer UV–optical colors thancounterparts in relaxed clusters. The bulk of theAGN have flat UV–optical colors, firmly classify-ing them as Seyfert 2 type source.6 • H α line emitters in merging clusters are poweredby SF at a higher rate than in relaxed clusters.Further, the bulk of emission-line galaxies in merg-ing cluster fields are located within 3 Mpc from thecenter. By contrast, AGN peak at the outskirts( ∼ . − • Galaxies powered by SF have larger EWs thanthose with AGN contribution. The star-formingpopulation in merging clusters have higher H α EW, or sSFR, than relaxed clusters. • We measure H α EWs exceeding 90 ˚A, poweredalmost exclusively by SF. The sources with thelargest EWs are found within 3 Mpc of the centersof merging clusters. • We discover a population of continuum-faint H α emitters with bright emission lines, with large EW,or large sSFR, exclusively found in merging clus-ters. These galaxies are likely located well abovethe field main-sequence. • We introduced redshifts , which enables the userto obtain all spectroscopic redshifts from the liter-ature, publicly available on VizieR and NED.In conclusion, we find significant populations of star-forming galaxies and Seyfert 2 type AGN in merginggalaxy clusters. The emission-line galaxy populationin merging clusters permeates the inner parts of theICM, which suggests that these galaxies are survivingthe strong environmental effects typically seen in relaxedclusters. Their blue colors, in combination with high sS-FRs, imply that they are undergoing episodes of vigor-ous star formation, contrary to expectations from mod-els of galaxy evolution in relaxed clusters, which predictan exponential decline of SFRs over 0 . − Facilities:
VLT:Melipal (VIMOS spectroscopy),ING:Herschel (AF2 spectroscopy), Keck:II (DEIMOSspectroscopy), MMT (Hectospec spectroscopy),ING:Newton (WFT imaging, photometry), MaxPlanck:2.2m (WFI imaging, photometry), Subaru(Suprimecam imaging, photometry), CFHT (Megacamimaging, photometry), Sloan (SDSS survey imaging,7photometry), PS1 (PS1 survey imaging, photometry),CXO (ACIS-I imaging), XMM (EPIC imaging), VLA(NVSS survey imaging)
Software: gleam (Stroe & Savu 2020), redshifts (Stroe 2020), Matplotlib (Hunter 2007), Astropy (As-tropy Collaboration et al. 2013), Astroquery (Ginsburg et al. 2019), APLpy (Robitaille & Bressert 2012), LMFIT(Newville et al. 2014), TOPCAT (Taylor 2005), STILTS(Taylor 2006), DS9 (Joye & Mandel 2003), AstrOmaticSoftware (Bertin & Arnouts 1996), EsoReflex (Freudlinget al. 2013), 2DKS (Taillon et al. 2019)APPENDIX A. SELECTION CRITERIA FOR THE FINALSAMPLESources located at the cluster redshift were includedin the final sample if they meet any of the followingcriteria: • Detection in [N ii ] and H α , or • Upper limit in, lower limit in, or unconstrained[N ii ]/H α and detection in [O iii ]/H β , or • No coverage in [N ii ]/H α or [O iii ]/H β , as long asthe other ratio is detected securely, or • Lower limit in [N ii ]/H α and upper limit in[O iii ]/H β , or • Upper limit in [N ii ]/H α and lower limit in[O iii ]/H β , or • Upper limit in [N ii ]/H α and upper limit in[O iii ]/H β , or • Unconstrained [N ii ]/H α and upper limit in[O iii ]/H β , or • Upper limit [N ii ]/H α and unconstrained in[O iii ]/H β . B. [O iii ]/H α RATIOS TO HELP AGN VS. SFCLASSIFICATIONFor sources that could not be securely classified basedon the BPT diagram alone, we employed the [O iii ]/H α ratio in relation to the [N ii ]/H α and the [O iii ]/H β ra-tios to classify the sources, only when the classificationwas unambiguous using the combination of the three ra-tios. We classified sources based on the spaces occupiedby model sources from Sobral et al. (2018) and (Sobralet al. 2019), classified as AGN, composite, or SF dom-inated using the BPT criteria, as well as sources fromour sample, which we securely classified in the BPT di-agram. We note that our [O iii ]/H α ratios are on aver-age lower than those predicted by the models because we they are not corrected for dust extinction. This ef-fect was taken into account when classifying new sourcesbased on the [O iii ]/H α ratio. For sources securely clas-sified based on the BPT diagram alone, we do not alterthe classification. We show the distribution of sourcesin the [O iii ]/H α vs. [N ii ]/H α space in Figure B1. Figure B1.
Distribution of our galaxies in the [O iii ]/H α versus [N ii ]/H α space, color-coded by AGN, composite, andSF emission. We show sources securely classified by using theKauffmann et al. (2003) and Kewley et al. (2001) criteria, aswell as sources classified with the help of the [O iii ]/H α ratio.We show the distribution of sources from cloudy modelingin the background (from Sobral et al. 2018, 2019). Alam, S., Albareti, F. D., Prieto, C. A., et al. 2015, ApJS,219, 12, doi: 10.1088/0067-0049/219/1/12Alberts, S., Pope, A., Brodwin, M., et al. 2016, ApJ, 825,72, doi: 10.3847/0004-637X/825/1/72Andrade-Santos, F., Jones, C., Forman, W. R., et al. 2017,ApJ, 843, 76, doi: 10.3847/1538-4357/aa7461Astropy Collaboration, Robitaille, T. P., Tollerud, E. J.,et al. 2013, A&A, 558, A33,doi: 10.1051/0004-6361/201322068Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981,PASP, 93, 5, doi: 10.1086/130766Balogh, M., Eke, V., Miller, C., et al. 2004, MNRAS, 348,1355, doi: 10.1111/j.1365-2966.2004.07453.xBamford, S. P., Nichol, R. C., Baldry, I. K., et al. 2009,MNRAS, 393, 1324,doi: 10.1111/j.1365-2966.2008.14252.xBekki, K., Couch, W. J., & Shioya, Y. 2002, ApJ, 577, 651,doi: 10.1086/342221Belfiore, F., Maiolino, R., Bundy, K., et al. 2018, MNRAS,477, 3014, doi: 10.1093/mnras/sty768Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393,doi: 10.1051/aas:1996164Biffi, V., Borgani, S., Murante, G., et al. 2016, ApJ, 827,112, doi: 10.3847/0004-637X/827/2/112Boselli, A., & Gavazzi, G. 2006, PASP, 118, 517,doi: 10.1086/500691Boylan-Kolchin, M., Springel, V., White, S. D. M., Jenkins,A., & Lemson, G. 2009, MNRAS, 398, 1150,doi: 10.1111/j.1365-2966.2009.15191.xCairns, J., Stroe, A., Breuck, C. D., Mroczkowski, T., &Clements, D. 2019, ApJ, 882, 132,doi: 10.3847/1538-4357/ab3392Cava, A., Biviano, A., Mamon, G. A., et al. 2017, A&A,606, A108, doi: 10.1051/0004-6361/201730785Chon, G., & B¨ohringer, H. 2017, A&A, 606, L4,doi: 10.1051/0004-6361/201731854Chung, A., van Gorkom, J. H., Kenney, J. D. P., Crowl, H.,& Vollmer, B. 2009a, AJ, 138, 1741,doi: 10.1088/0004-6256/138/6/1741Chung, S. M., Gonzalez, A. H., Clowe, D., et al. 2009b,ApJ, 691, 963, doi: 10.1088/0004-637X/691/2/963Cohen, S. A., Hickox, R. C., Wegner, G. A., Einasto, M., &Vennik, J. 2014, ApJ, 783, 136,doi: 10.1088/0004-637X/783/2/136Cohn, J. D., & White, M. 2005, Astroparticle Physics, 24,316, doi: 10.1016/j.astropartphys.2005.07.006Connor, T., Kelson, D. D., Mulchaey, J., et al. 2018, ApJ,867, 25, doi: 10.3847/1538-4357/aae38b Cooke, E. A., Hatch, N. A., Muldrew, S. I., Rigby, E. E., &Kurk, J. D. 2014, MNRAS, 440, 3262,doi: 10.1093/mnras/stu522Cortese, L., Gavazzi, G., Boselli, A., Iglesias-Paramo, J., &Carrasco, L. 2004, A&A, 425, 429,doi: 10.1051/0004-6361:20040381Darvish, B., Mobasher, B., Martin, D. C., et al. 2017, ApJ,837, 16, doi: 10.3847/1538-4357/837/1/16Darvish, B., Scoville, N. Z., Martin, C., et al. 2020, ApJ,892, 8, doi: 10.3847/1538-4357/ab75c3de Souza, R. S., Dantas, M. L. L., Krone-Martins, A., et al.2016, MNRAS, 461, 2115, doi: 10.1093/mnras/stw1459Delgado, R. M. G., Heckman, T., Leitherer, C., et al. 1998,ApJ, 505, 174, doi: 10.1086/306154Deshev, B., Finoguenov, A., Verdugo, M., et al. 2017,A&A, 607, A131, doi: 10.1051/0004-6361/201731235Dressler, A. 1980, ApJ, 236, 351, doi: 10.1086/157753Dressler, A. 1984, ARA&A, 22, 185,doi: 10.1146/annurev.astro.22.1.185Ebeling, H., & Kalita, B. S. 2019, ApJ, 882, 127,doi: 10.3847/1538-4357/ab35d6Eckert, D., Ghirardini, V., Ettori, S., et al. 2019, A&A,621, A40, doi: 10.1051/0004-6361/201833324Erfanianfar, G., Popesso, P., Finoguenov, A., et al. 2016,MNRAS, 455, 2839, doi: 10.1093/mnras/stv2485Fabricant, D., Fata, R., Roll, J., et al. 2005, PASP, 117,1411, doi: 10.1086/497385Flewelling, H. A., Magnier, E. A., Chambers, K. C., et al.2016, arXiv e-prints, 1612, arXiv:1612.05243.http://adsabs.harvard.edu/abs/2016arXiv161205243FFreudling, W., Romaniello, M., Bramich, D. M., et al. 2013,A&A, 559, A96, doi: 10.1051/0004-6361/201322494Fujita, Y., Takizawa, M., Nagashima, M., & Enoki, M.1999, PASJ, 51, L1, doi: 10.1093/pasj/51.3.L1Fumagalli, M., Patel, S. G., Franx, M., et al. 2012, ApJL,757, L22, doi: 10.1088/2041-8205/757/2/L22Ginsburg, A., Sip¨ocz, B. M., Brasseur, C. E., et al. 2019,AJ, 157, 98, doi: 10.3847/1538-3881/aafc33Golovich, N., Dawson, W. A., Wittman, D. M., et al. 2019,ApJS, 240, 39, doi: 10.3847/1538-4365/aaf88bGoto, T., Yamauchi, C., Fujita, Y., et al. 2003, MNRAS,346, 601, doi: 10.1046/j.1365-2966.2003.07114.xGunn, J. E., & Gott, III, J. R. 1972, ApJ, 176, 1,doi: 10.1086/151605Hatch, N. A., Kurk, J. D., Pentericci, L., et al. 2011,MNRAS, 415, 2993,doi: 10.1111/j.1365-2966.2011.18735.xHayashi, M., Tanaka, M., Shimakawa, R., et al. 2018,PASJ, 70, S17, doi: 10.1093/pasj/psx088 Hou, A., Parker, L. C., Wilman, D. J., et al. 2012, MNRAS,421, 3594, doi: 10.1111/j.1365-2966.2012.20586.xHunter, J. D. 2007, Computing in Science & Engineering, 9,90, doi: 10.1109/MCSE.2007.55Hwang, H. S., & Lee, M. G. 2009, MNRAS, 397, 2111,doi: 10.1111/j.1365-2966.2009.15100.xJaff´e, Y. L., Poggianti, B. M., Verheijen, M. A. W., Deshev,B. Z., & van Gorkom, J. H. 2012, ApJL, 756, L28,doi: 10.1088/2041-8205/756/2/L28Jaff´e, Y. L., Verheijen, M. A. W., Haines, C. P., et al. 2016,MNRAS, 461, 1202, doi: 10.1093/mnras/stw984Jones, C., & Forman, W. 1999, ApJ, 511, 65,doi: 10.1086/306646Joye, W. A., & Mandel, E. 2003, in Astronomical Society ofthe Pacific Conference Series, Vol. 295, AstronomicalData Analysis Software and Systems XII, ed. H. E.Payne, R. I. Jedrzejewski, & R. N. Hook, 489Kauffmann, G., Heckman, T. M., Tremonti, C., et al. 2003,MNRAS, 346, 1055,doi: 10.1111/j.1365-2966.2003.07154.xKenney, J. D. P., van Gorkom, J. H., & Vollmer, B. 2004,AJ, 127, 3361, doi: 10.1086/420805Kewley, L. J., Dopita, M. A., Sutherland, R. S., Heisler,C. A., & Trevena, J. 2001, ApJ, 556, 121,doi: 10.1086/321545Khostovan, A. A., Sobral, D., Mobasher, B., et al. 2015,MNRAS, 452, 3948, doi: 10.1093/mnras/stv1474Kinney, A. L., Calzetti, D., Bohlin, R. C., et al. 1996, ApJ,467, 38, doi: 10.1086/177583Kodama, T., Smail, I., Nakata, F., Okamura, S., & Bower,R. G. 2001, ApJ, 562, L9, doi: 10.1086/338100Koyama, Y., Kodama, T., Tadaki, K.-i., et al. 2013,MNRAS, 428, 1551, doi: 10.1093/mnras/sts133Larson, R. B., Tinsley, B. M., & Caldwell, C. N. 1980, ApJ,237, 692, doi: 10.1086/157917Lovisari, L., Forman, W. R., Jones, C., et al. 2017, ApJ,846, 51, doi: 10.3847/1538-4357/aa855fMa, C.-J., Ebeling, H., Marshall, P., & Schrabback, T.2010, MNRAS, 406, 121,doi: 10.1111/j.1365-2966.2010.16673.xMarkevitch, M., & Vikhlinin, A. 2007, PhR, 443, 1,doi: 10.1016/j.physrep.2007.01.001Melnick, J., & Sargent, W. L. W. 1977, ApJ, 215, 401,doi: 10.1086/155369Miller, N. A., & Owen, F. N. 2003, AJ, 125, 2427,doi: 10.1086/374767Moore, B., Katz, N., Lake, G., Dressler, A., & Oemler, A.1996, Nature, 379, 613, doi: 10.1038/379613a0Moravec, E., Gonzalez, A. H., Dicker, S., et al. 2020, ApJ,898, 145, doi: 10.3847/1538-4357/aba0b2 Moretti, A., Paladino, R., Poggianti, B. M., et al. 2020,ApJL, 897, L30, doi: 10.3847/2041-8213/ab9f3bMulroy, S. L., McGee, S. L., Gillman, S., et al. 2017,MNRAS, 472, 3246, doi: 10.1093/mnras/stx2108Muriel, H., & Coenda, V. 2014, A&A, 564, A85,doi: 10.1051/0004-6361/201322033Newman, J. A., Cooper, M. C., Davis, M., et al. 2013,ApJS, 208, 5, doi: 10.1088/0067-0049/208/1/5Newville, M., Stensitzki, T., Allen, D. B., & Ingargiola, A.2014, LMFIT: Non-Linear Least-Square Minimizationand Curve-Fitting for Python, 0.8.0, Zenodo,doi: 10.5281/zenodo.11813Noble, A. G., McDonald, M., Muzzin, A., et al. 2017,ApJL, 842, L21, doi: 10.3847/2041-8213/aa77f3Ochsenbein, F., Bauer, P., & Marcout, J. 2000, A&AS, 143,23, doi: 10.1051/aas:2000169Osterbrock, D. E., & Mathews, W. G. 1986, ARA&A, 24,171, doi: 10.1146/annurev.aa.24.090186.001131Owen, F. N., Ledlow, M. J., Keel, W. C., & Morrison,G. E. 1999, AJ, 118, 633, doi: 10.1086/300974Paulino-Afonso, A., Sobral, D., Darvish, B., et al. 2018,A&A, 620, A186, doi: 10.1051/0004-6361/201832688Peacock, J. A. 1983, MNRAS, 202, 615,doi: 10.1093/mnras/202.3.615Pimbblet, K. A. 2011, MNRAS, 411, 2637,doi: 10.1111/j.1365-2966.2010.17869.xPimbblet, K. A., Shabala, S. S., Haines, C. P.,Fraser-McKelvie, A., & Floyd, D. J. E. 2013, MNRAS,429, 1827, doi: 10.1093/mnras/sts470Poggianti, B. M., Jaff´e, Y. L., Moretti, A., et al. 2017,Nature, 548, 304, doi: 10.1038/nature23462Pranger, F., B¨ohm, A., Ferrari, C., et al. 2014, A&A, 570,A40, doi: 10.1051/0004-6361/201424727Ricarte, A., Tremmel, M., Natarajan, P., & Quinn, T. 2020,ApJL, 895, L8, doi: 10.3847/2041-8213/ab9022Robitaille, T., & Bressert, E. 2012, APLpy: AstronomicalPlotting Library in Python. http://ascl.net/1208.017Roediger, E., Br¨uggen, M., Owers, M. S., Ebeling, H., &Sun, M. 2014, MNRAS, 443, L114,doi: 10.1093/mnrasl/slu087Rossetti, M., Gastaldello, F., Eckert, D., et al. 2017,MNRAS, 468, 1917, doi: 10.1093/mnras/stx493Shim, H., Im, M., Lee, H. M., et al. 2015, ApJ, 727, 14,doi: 10.1088/0004-637X/727/1/14Shioya, Y., Taniguchi, Y., Sasaki, S. S., et al. 2008, ApJS,175, 128, doi: 10.1086/523703Sobral, D., Stroe, A., Dawson, W. A., et al. 2015, MNRAS,450, 630, doi: 10.1093/mnras/stv521Sobral, D., Stroe, A., Koyama, Y., et al. 2016, MNRAS,458, 3443, doi: 10.1093/mnras/stw534 Sobral, D., Matthee, J., Darvish, B., et al. 2018, MNRAS,477, 2817, doi: 10.1093/mnras/sty782Sobral, D., Matthee, J., Brammer, G., et al. 2019, MNRAS,482, 2422, doi: 10.1093/mnras/sty2779Stroe, A. 2020, redshiftsredshifts