Jellyfish: The origin and distribution of extreme ram-pressure stripping events in massive galaxy clusters
Conor McPartland, Harald Ebeling, Elke Roediger, Kelly Blumenthal
MMon. Not. R. Astron. Soc. , 000–000 (0000) Printed 5 November 2015 (MN L A TEX style file v2.2)
Jellyfish: The origin and distribution of extreme ram-pressurestripping events in massive galaxy clusters
Conor McPartland , Harald Ebeling , Elke Roediger & Kelly Blumenthal Institute for Astronomy, University of Hawai’i at Manoa, 2680 Woodlawn Drive, Honolulu, HI, 96822, USA E.A. Milne Centre for Astrophysics, Department of Physics & Mathematics, University of Hull, Cottinton Road, Hull, HU6 7RX, United Kingdom
Draft version
ABSTRACT
We investigate the observational signatures and physical origin of ram-pressure stripping(RPS) in 63 massive galaxy clusters at z = 0 . − . , based on images obtained with theHubble Space Telescope. Using a training set of a dozen “jellyfish” galaxies identified earlierin the same imaging data, we define morphological criteria to select 211 additional, less ob-vious cases of RPS. Spectroscopic follow-up observations of 124 candidates so far confirmed53 as cluster members. For the brightest and most favourably aligned systems we visuallyderive estimates of the projected direction of motion based on the orientation of apparentcompression shocks and debris trails.Our findings suggest that the onset of these events occurs primarily at large distancesfrom the cluster core ( > kpc), and that the trajectories of the affected galaxies featurehigh impact parameters. Simple models show that such trajectories are highly improbable forgalaxy infall along filaments but common for infall at high velocities, even after observationalbiases are accounted for, provided the duration of the resulting RPS events is (cid:46)
500 Myr.We thus tentatively conclude that extreme RPS events are preferentially triggered by clustermergers, an interpretation that is supported by the disturbed dynamical state of many of thehost clusters. This hypothesis implies that extreme RPS might occur also near the cores ofmerging poor clusters or even merging groups of galaxies.Finally, we present nine additional “jellyfish” galaxies at z > Key words: galaxies: evolution - galaxies: clusters: intracluster medium - galaxies: structure
Evidence of accelerated galaxy evolution in galaxy clusters hasbeen presented as early as 1980, the most well known examplesbeing the increased occurrence of ellipticals in dense environ-ments (i.e., the morphology-density relation; Dressler 1980) andthe higher fraction of blue galaxies in clusters at higher redshift(i.e., the Butcher-Oemler effect, Butcher & Oemler 1984). Thephysical mechanisms responsible for these effects are, however,still very much debated. A variety of processes have been proposedin the literature, ranging form slow-acting gravitational interactionssuch as galaxy-galaxy harassment (Moore et al. 1996) to potentiallyextremely rapid galaxy transformations brought about by interac-tions with the gaseous intracluster medium (ICM).The latter process, ram-pressure stripping (RPS) is expectedto be especially efficient in massive galaxy clusters, as the pres-sure imparted on a galaxy is directly proportional to the local gasdensity of the ICM and to the square of the galaxy’s velocity withrespect to the ICM (Gunn & Gott 1972). The resulting removal ofthe galaxy’s interstellar medium (ISM) occurs in the direction ofmotion of the galaxy relative to the ICM, generating a trail of star-forming regions in the galaxy’s wake. For fortuitous viewing an- gles, this trail, or at least the associated deformation of the galacticdisk, is accessible to observation, thus creating a rare opportunityto constrain the motion of galaxies in the plane of the sky. Ob-servations of RPS events thus constitute a valuable complement tospectroscopic radial-velocity surveys and permit a detailed inves-tigation of the kinematics and spatial evolution of galaxies in thedense cluster environment.The physics and observational signature of RPS have beenthe subject of extensive numerical simulations which predict thatgradual stripping should be pervasive even in low-mass clusters(Vollmer et al. 2001). Indeed RPS events have been studied in greatdetail in the Virgo (Chung et al. 2007; Vollmer et al. 2012; Abram-son et al. 2011) and Coma clusters (Smith et al. 2010; Yagi et al.2010), as well as in other nearby systems, such as the Shapley Con-centration (Merluzzi et al. 2013) or Abell 3627 (Sun, Donahue &Voit 2007; Fumagalli et al. 2014). As expected, these events arerelatively modest though, with observations showing atomic hy-drogen to be displaced and only partially removed (Scott et al.2010), while the denser, more centrally located molecular gas isfound to be essentially unperturbed (Boselli et al. 1997; Vollmeret al. 2001). By contrast, in the most massive clusters the environ- c (cid:13) a r X i v : . [ a s t r o - ph . GA ] N ov C. McPartland et al. ment encountered by infalling galaxies can lead to their entire gasreservoir being stripped in a single pass through the cluster core(e.g. Takeda, Nulsen & Fabian 1984; Abadi, Moore & Bower 1999;Kapferer et al. 2009; Steinhauser et al. 2012). Observational evi-dence of extreme ram-pressure stripping is, however, sparse, dueto their reliance on favourable circumstances, such as suitable in-fall trajectory, gas mass, galaxy orientation, and high ICM density.Considering the small number and relatively low masses of nearbyclusters (except for Coma), these conditions are unlikely to be metin the local Universe.The extreme environment that is a prerequisite for extremeRPS is, however, routinely encountered by galaxies falling intomassive clusters where galaxy peculiar velocities in excess of 1000km s -1 are common and the ICM particle density easily exceeds10 -3 cm -3 . Since massive clusters are rare, larger volumes have tobe searched to efficiently probe such truly high-density environ-ments. Although their numbers are still small, striking examples ofextreme RPS events have been discovered in Hubble Space Tele-scope ( HST ) images of moderately distant ( z (cid:38) . ) massive clus-ters (Owen et al. 2006; Cortese et al. 2007; Owers et al. 2012) and,most recently, in X-ray selected massive clusters at z> . (Ebel-ing, Stephenson & Edge 2014, see Fig. 1). Importantly, these clus-ters are not only intrinsically more massive, they are also dynam-ically less evolved and more likely to be undergoing mergers thansystems in the local Universe (Mann & Ebeling 2012), a criticalrequirement if extreme RPS events are triggered by merger-drivenshocks, as suggested by Owers et al. (2012). Increasing the size ofthe still small sample of RPS examples clearly constitutes a crucialstep toward a meaningful statistical investigation of the physics ofaccelerated galaxy evolution.In this paper, we aim to compile a statistically significant sam-ple of galaxies that might be undergoing RPS in very massive clus-ters. We then use this sample to establish which galaxy trajectoriesare most conducive to creating extreme RPS, and thereby elucidatewhether the most dramatic RPS events are triggered by massivecluster mergers (Owers et al. 2012), rather than during regular infallof galaxies from the field or along filaments. In order to compile therequired sample, we develop morphological criteria to select RPScandidates from archival HST imaging data for a well defined sam-ple of massive clusters at z > . , and compare the spatial and dy-namical distribution of the selected objects with expectations fromnumerical and theoretical models.This paper is structured as follows: in Section 2 we introducethe cluster sample and present an overview of the observations anddata-reduction procedures; in Section 3 we discuss our morpho-logical criteria for the identification of galaxies experiencing ram-pressure stripping and present the sample of RPS candidates; inSection 4 we present the a simple model of clustre infall which weuse to interpret our data; in Section 5 we present our results forthe spatial distribution and dynamical properties of RPS events inmassive clusters; and in Section 6 we draw conclusions about theorigin, trajectories, and physics of extreme RPS. We present a sum-mary of our work in Section 7.Throughout this paper, we assume a concordance Λ CDM cos-mology with Ω M = 0.3, Ω Λ = 0.7, H = 70 km s -1 Mpc -1 . As theclusters in our sample span a range of redshifts of . < z < . ,the metric scale of our images varies from 4.45 to 7.15 kpc arcsec -1 . Our cluster sample is drawn from a master list of clusters identifiedin the course of the Massive Cluster Survey (MACS; Ebeling, Edge& Henry 2001; Ebeling et al. 2007, 2010; Mann & Ebeling 2012),designed to provide a large, statistically complete sample of X-rayluminous ( L X (cid:38) × erg s -1 , 0.1-2.4 keV) and moderatelydistant ( z (cid:38) . ) galaxy clusters. Covering over 22,000 sq.deg.,the MACS sample comprises the majority of massive galaxy clus-ters in the observable Universe, making it ideally suited for ourinvestigation. At redshifts z (cid:38) . , the sub-kiloparsec angular res-olution needed to identify the characteristic morphological traits ofRPS events can only be achieved with the Advanced Camera forSurveys (ACS) aboard HST . We thus limit our sample to MACSclusters with archival
HST /ACS images as described in more detailin the following section.
As our primary observational diagnostics revolve around morpho-logical features traced by star-forming regions, we limit our studyto MACS clusters that have been observed in the
HST /ACS F606Wband. The F606W filter is well suited as it corresponds roughly tothe B band in the cluster rest frame and has been used in a largenumber of
HST observations of MACS clusters. We further re-quire clusters in our sample to also have imaging data in the ACSF814W passband, as the resulting F814W–F606W colours providea straightforward means to discriminate against the population ofpassively evolving cluster ellipticals.Of the entire MACS sample, 44 clusters were successfullyobserved in both the ACS F606W and F814W passbands as partof the
HST
SNAPshot programmes GO-10491, -10875, -12166,and -12884 (PI: Ebeling). These programmes use short exposures(1200 seconds for F606W and 1440 seconds for F814W) designedto reveal bright strong-lensing features and provide constraints onthe physical nature of galaxy-galaxy and galaxy-gas interactions incluster cores. Fundamental properties of this subset of the MACScluster sample are presented and discussed by Ebeling & Repp(in preparation). Supplementing these SNAPshots, we also includedata from observations of 17 additional MACS clusters obtained bythe Cluster Lensing and Supernova Survey with Hubble (CLASH;Postman et al. 2012), an
HST
Multi-Cycle Treasury Program em-ploying 16 filters from the UV to the NIR, including F606W andF814W. Exposure times for the CLASH observations are nomi-nally one and two orbits for all ACS filters, but vary substantiallybetween cluster fields around median exposure times of 4060 and8480 seconds for the F606W and F814W passbands, respectively(see Table A1 & A2 for a summary of the observations).In total, our sample thus comprises 63 MACS clusters. At theredshifts relevant to our study, the field of view of the ACS WideField Channel ( (cid:48)(cid:48) × (cid:48)(cid:48) ) covers an inscribed circle of radiusbetween 450 and 720 kpc and thus samples primarily the clus-ter core region. Charge-transfer-efficiency corrected images in thetwo passbands were registered using the astrometric solution of theF606W image as a reference, and source catalogs were created us-ing SExtractor (Bertin & Arnouts 1996) in dual-image mode, withF606W chosen as the detection band. We removed stars as well ascosmic rays and other artefacts as objects falling on or below the c (cid:13) , 000–000 am-Pressure Stripping in Massive Clusters . s . s . s . s . s . . . . h m . s − ◦ .
10 kpc . s . s . s . s . . . . h m . s ◦ .
10 kpc . s . s . s . s . s . . . . h m . s ◦ .
10 kpc . s . s . s . s . s . . . . h m . s ◦ .
10 kpc . s . s . s . s . s . s . . . . . h m . s ◦ .
10 kpc . s . s . s . s . s . s . . . . h m . s ◦ .
10 kpc . s . s . s . s . . . . . h m . s − ◦ .
10 kpc . s . s . s . s . s . s . s . s . . . . . h m . s ◦ .
10 kpc . s . s . s . s . s . s . . . . . h m . s ◦ .
10 kpc . s . s . s . s . s . . . . h m . s ◦ .
10 kpc . s . s . s . s . s . s . s . s . . . . h m . s ◦ .
10 kpc . s . s . s . s . s . . . . h m . s ◦ .
10 kpc
Figure 1.
The 12 galaxies deemed textbook examples of ram-pressure stripping and thus used as our training set; six of these (top two rows) were publishedpreviously by Ebeling, Stephenson & Edge (2014). Three members of our training set were recently found not to be cluster members (see Section 3.2.1) andare highlighted in the bottom row.c (cid:13) , 000–000
C. McPartland et al. A C M G Sk − S k − Figure 2.
The distribution of all galaxies in our target fields in various parameter spaces. Left: Concentration–Asymmetry; centre: Gini– M ; right: Sk − – Sk − . Our final sample of RPS candidates is marked by filled blue circles; the morphologically most compelling examples are shown as yellow asterisks.Members of our training set (see Fig. 1) are shown with open symbols. Squares show the six systems published in Ebeling, Stephenson & Edge (2014), andcircles show the six additional galaxies from their extended sample. Three members of our training set, all part of the extended ESE sample, were recentlyfound not to be cluster members (see Section 3.2.1) and are shown in red. The cuts defining our final morphological selection criteria are indicated by reddashed lines. star lines in both magnitude- µ max and magnitude- r , ell space .After removing spurious detections, we have a 5 σ
90% complete-ness limiting magnitude of 24.9 in F606W (here and in the fol-lowing the magnitudes quoted are measured within the Petrosianradius).As the quantitative morphological indicators we employ toidentify RPS candidates (see Section 3) require signal-to-noise ra-tios of (cid:104)
S/N (cid:105) > per pixel, we limit our galaxy sample to objectswith m F606W < , which leaves a total of 15,875 galaxies (11,550in the SNAPshot data and 4,325 in the CLASH data). We note that,due to the high density of objects in cluster cores and the presenceof objects of complex morphology, some of the objects in our mas-ter catalogue may in fact be blends of several objects, whereas oth-ers have suffered fragmentation, i.e., were broken up into multiplesources.To mitigate the effect of fragmentation in our master catalog,we enforce strict deblending criteria (DEBLEND NTHRESH=16,DEBLEND MINCONT=0.2). Due to the relatively shallow depth( ∼ Here, µ max and r , ell are the peak surface brightness and the ellipti-cal radius encircling 20% of the total flux, respectively. The sample of RPS candidates compiled in this work using mor-phological selection is expected to be heavily contaminated bygalaxies that are in fact not members of the respective MACS clus-ter and / or whose morphology is irregular for reasons other thanRPS (see Section 3 for details). In order to eliminate interlopers,we have embarked on a comprehensive spectroscopic survey ofour RPS candidate sample, aimed at (a) excluding fore- and back-ground galaxies from our sample of RPS candidates, and (b) obtain-ing peculiar radial velocities of those systems that are cluster mem-bers. We refer to a forthcoming paper (Blumenthal et al., in prepa-ration) for a more extensive report on these efforts, including a de-scription of the data-reduction procedure. We note here though thatall spectroscopic observations were conducted with the DEIMOSspectrograph on the Keck-II 10m-telescope on Maunakea, usingmulti-object spectroscopy with slits of 1mm width, the 600 l/mmZerodur grating set to a central wavelength of 6300 ˚A, the GG455blocking filter, and exposure times ranging from 3 ×
10 to 3 × − in radial velocity. A recent study by Ebeling, Stephenson & Edge (2014, here-after ESE) presented six textbook examples of “jellyfish” galax-ies (thought to be extreme RPS events ) discovered in HST imag-ing data for 36 of the 63 clusters used in this work. These objectswere visually identified, having to meet the following criteria: (1) astrongly disturbed morphology in optical images indicative of uni-lateral external forces; (2) a pronounced brightness and colour gra-dient suggesting extensive triggered star formation; (3) compellingevidence of a debris trail. Furthermore, the direction of motion im-plied by each of these features had to be consistent. We expandthe ESE sample by six additional, unpublished, jellyfish candidates, Although the observed morphology of these objects does not prove theoccurrence of RPS, in-depth follow-up studies of galaxies sharing the samestriking features unambiguously confirmed RPS to be at work (Sun et al.2010; Sivanandam, Rieke & Rieke 2010; Cort´es, Kenney & Hardy 2015).c (cid:13) , 000–000 am-Pressure Stripping in Massive Clusters C − A G − M Sk − Sk Figure 3.
Venn diagram of the sets of galaxies selected by each of the mor-phological criteria shown in the three panels of Fig. 2. Although each typeof cut selects a similar number of galaxies (represented by the area of eachcircle), the modest overlap between these sets makes the final selection,achieved by requiring all criteria to be met, much more restrictive. identified by the same authors, that satisfy at least two of these cri-teria , and use the resulting superset of 12 objects (shown in Fig 1)as a training set for the identification of additional, less obviouscandidate objects.For each of the galaxies in our catalogue we compute severalnon-parametric galaxy morphology statistics defined previously inthe literature: concentration ( C ) and asymmetry ( A ) (Bershady,Jangren & Conselice 2000; Conselice 2003), Gini coefficient ( G )and M (Abraham, van den Bergh & Nair 2003; Lotz, Primack& Madau 2004). While these statistics were originally designedto identify the morphological features of galaxy mergers, we findthat they can be applied more widely to characterise and select ob-jects featuring disturbed morphologies. In addition to the aforemen-tioned four statistics, we introduce two “skeletal decomposition”parameters ( Sk − and Sk − ; see Appendix A).We compute values for each of these indicators using the el-lipticities, position angles, and locations provided by SExtractorbut note that the precise location of the centre of each object is it-eratively refined through minimisation procedures, as described inLotz, Primack & Madau (2004). Acknowledging the difficulty ofcleanly separating galaxies in crowded cluster cores, we resort tousing SExtractor’s segmentation maps to identify the pixels belong-ing to a given galaxy rather than relying on an isophotal definitionof a galaxy’s extent. We stress that, as a result, the morphologicalquantities measured here should not be directly compared to thosefrom other work. The fact that the extended ESE sample (Fig. 1) contains some ofthe most extreme examples of jellyfish galaxies known to date (i.e.,the brightest and most morphologically disturbed) makes it wellsuited as a training set for an iterative, semi-automated search foradditional RPS candidates. To this end, we examine the location ofthe training-set members in C – A , Gini– M , and Sk − – Sk − space, and define cuts in these parameter spaces that preserve thetraining set but eliminate the vast majority of other galaxies. Thephysical rationale behind these cuts is to discard extremely dif-fuse objects (achieved by a cut in C ), almost perfectly symmetricsources (cut in A ), morphologically undisturbed disk and elliptical Note that the inferred direction of motion for two candidates (leftmosttwo in the bottom row of Fig. 1) is largely aligned with our line of sight. galaxies (cut in G - M ), and, finally, objects with little substruc-ture (cuts in Sk − and Sk − ).We apply an initial set of morphological criteria (cuts in C – A and Gini– M ) to galaxies detected in the 10 cluster fields fromwhich the extended ESE sample originates. The ∼
650 candidateobjects thus selected are then visually scrutinised independently bytwo of us (CM and HE) and classified according to their plausibilityas RPS events. We attempt to reduce the subjectivity of this proce-dure by reviewing jointly, in a second iteration, all objects classifiedeither as compelling jellyfish galaxies or as plausible candidatesby one of the inspectors and assigning a consensus classification.From the resulting set of potential RPS events we select the mostcompelling candidates, add them to our original training set, andre-evaluate our initial morphological constraints. Cuts in colour-magnitude space were considered too during this process but ul-timately dismissed as largely redundant with the aforementionedmorphological cuts, which already remove the majority cluster el-lipticals and faint blue objects. The full set of morphological cri-teria (now also including cuts in Sk − – Sk − ) are then appliedto the remaining clusters, and the resulting subset is once againvisually screened. Fig. 2 shows the distribution of all galaxies invarious projections of our multi-dimensional morphology parame-ter space, as well as the applied selection criteria. Members of theextended training set and of our final sample of RPS candidates arehighlighted. Although the three sets of selection criteria shown inFig. 2 all select approximately the same fraction of galaxies (30-40%), their doing so largely non-redundantly leads to a much morerestrictive selection of merely 8% (1263 galaxies) when all criteriaare combined (Fig. 3).It is evident from Fig. 2 that the adopted selection criteria, al-though highly efficient in eliminating regular disk galaxies and el-lipticals, still select mostly galaxies that, although morphologicallydisturbed, are not necessarily undergoing RPS. In fact less than20% of the automatically selected systems are classified as RPScandidates in our visual screening process. The disturbed sourcesrejected after visual inspection can largely be assigned to one of thefollowing classes: strong-gravitational-lensing features (includingboth cluster-galaxy and galaxy-galaxy lensing events), foregroundirregular galaxies, close pairs of ellipticals, unclassifiable clumpyemission in low signal-to-noise areas, and artefacts due to sourceconfusion in crowded regions. We also note that, while colour in-formation was not directly included in our selection procedure, theavailability of images in both the F606W and F814W passbandsproved essential in our visual classification to distinguish betweenthe morphological disturbances caused by RPS and irregular ex-tinction due to dust (see Fig. 4). The process described in the previous section yielded 223 possibleram-pressure stripping events (including the training set). We con-sider 15 of these to be classical jellyfish galaxies (yellow symbolsin Fig. 2); an additional 115 objects show characteristic features ofRPS (albeit less extreme), and 93 are at least plausible candidates.While we cannot rule out that physical processes other than RPS(e.g., minor mergers or tidal interactions) contribute to, or in factcause, the observed morphology of our candidates, such alternativescenarios are likely to be relevant mainly for the fainter galaxies inour sample for which the most compelling sign of RPS (evidenceof a debris trail) cannot be discerned in the shallow imaging data inhand.As a complement to the first six ”jellyfish” galaxies discovered c (cid:13) , 000–000 C. McPartland et al.
Table 1.
Properties of the morphologically most compelling ”jellyfish” galaxies that constitute our training set. The projected radius r BCG is the projecteddistance to the (nearest) BCG; the listed angle of incidence is the mean of the values assigned by the three reviewers (see arrows in Fig. 5. The first six galaxiesform the jellyfish sample of ESE.).Name α [J2000] δ [J2000] m F606W m F814W r BCG [kpc] Incidence [deg.] z MACSJ0257-JFG1 02 57 41.4 −
22 09 53 18.75 18.22 166 10 0.3241MACSJ0451-JFG1 04 51 57.3 +
00 06 53 19.66 19.29 298 50 0.4362MACSJ0712-JFG1 07 12 18.9 +
59 32 06 19.10 18.39 87 107 0.3430MACSJ0947-JFG1 09 47 23.1 +
76 22 52 19.81 19.69 210 34 0.3417MACSJ1258-JFG1 12 57 59.6 +
47 02 46 19.10 18.70 133 45 0.3424MACSJ1752-JFG1 17 51 56.1 +
44 40 20 20.13 19.61 370 120 0.3739MACSJ0035-JFG1 00 35 27.3 −
20 16 18 19.49 19.02 182 103 0.3597MACSJ0257-JFG2 02 57 43.5 −
22 08 38 19.92 19.44 243 130 0.3297MACSJ0429-JFG1 04 29 33.3 −
02 53 02 20.97 20.64 203 113 0.4000MACSJ0429-JFG1 04 29 40.4 −
02 53 18 20.75 20.36 334 40 0.4049MACSJ0916-JFG1 09 16 12.9 −
00 25 01 20.43 19.97 334 81 0.3300MACSJ1142-JFG1 11 42 37.0 +
58 31 48 20.25 19.62 549 87 0.3267MACSJ1720-JFG1 17 20 13.6 +
35 37 17 20.05 19.52 309 30 0.3832MACSJ1752-JFG1 17 52 06.3 +
44 40 05 20.25 20.06 747 86 0.3527RXJ2248-JFG1 22 48 40.2 −
44 30 50 20.66 20.18 335 64 0.3515 . s . s . s . s . s . s . . . . . h m . s ◦ .
10 kpc . s . s . s . s . s . s . . . . . h m . s ◦ .
10 kpc
Figure 4.
Importance of colour information for our visual inspections.Viewed solely in the F606W passband (left) this object could be consid-ered a (remotely) plausible RPS candidate. A false-colour image includingdata in the F814W filter (right) strongly suggests a slightly disturbed dustydisk galaxy. in MACS clusters by Ebeling, Stephenson & Edge (2014), we showin Fig. 5 a second sample of nine compelling jellyfish galaxies;fundamental properties of these systems are further described inSection 3.3 and listed in Table 1.
Impressive as the list of 223 RPS candidates may appear, we cau-tion again that most of these galaxies may not even be cluster mem-bers, and that, for those that are, the cause of the observed mor-phological features need not be RPS. In addition, our list is al-most certainly incomplete. Two primary observational biases areto blame: (a) our inability to reliably discriminate against non-RPSevents solely from morphological data (leading to contamination bynon-cluster galaxies) and (b) our inability to identify RPS events ingalaxies moving close to our line of sight (leading to incomplete-ness regarding true RPS events in our target clusters).First results from a comprehensive spectroscopic survey of allcandidates (Blumenthal et al., in preparation) indeed indicate thatmore than half of the objects we selected are in fact fore- or back-ground galaxies. The hazards of morphological selection alone areunderlined not just by this high percentage of projection effects,but also by the elimination of three members of our extended train-ing set (see bottom row Fig. 1): the edge-on disk with a stellar tail in MACSJ1236.9+6311 is in the foreground of the cluster, whilethe dramatically distorted face-on spiral galaxy near the core ofMACSJ1652.3+5534 was found to be a background object grav-itationally lensed by the massive MACS cluster. The bright blueface-on spiral in MACSJ1731.6+2252, finally, turned out to be amember of a foreground group of galaxies. Although the removalof these three objects from our training set has no effect on ourselection criteria, as can be seen from Fig. 2 in which these galax-ies are marked by red circles, the misidentification of galaxies weconsidered ”textbook” cases of RPS serves as a warning about therobustness of morphological selection and underlines the need forspectroscopic follow-up observations.The impact of the second observational bias cannot trivially bequantified by means of additional observations. Galaxies movingclose to our line of sight lack the tell-tale debris trail and bow-shockmorphology readily apparent for RPS proceeding in the plane of thesky (see Fig. 6) and are thus likely to be missed. We attempt to ac-count for the resulting systematic incompleteness when modellinggalaxy trajectories in Section 4.
Since one of the goals of our study is to distinguish between the dif-ferent geometric and kinematic scenarios associated with ”stream-fed” infall along filaments, and cluster mergers, we focus on twokey properties of cluster galaxies: the angle of incidence of their tra-jectory with respect to the gravitational centre of the cluster and thedistance from the cluster centre. To observationally constrain theformer, we consult the results of hydrodynamical modeling of RPS(e.g., Roediger & Br¨uggen 2006; Kronberger et al. 2008; Roedigeret al. 2014) for insights regarding the correlation between the mor-phological disturbances caused by RPS and the galaxy’s directionof motion. Figure 6 shows model predictions for the distribution ofgas and newly formed stars in galaxies undergoing RPS while mov-ing face-on through the ICM. As expected, identifying the directionof motion becomes challenging when a galaxy moves through theICM along our line of sight or is observed early in the strippingprocess.We attempt to assign projected directions of motion visuallyaccording to the following prescriptions: (1) if tails are discernible,the velocity vector is assumed to be parallel to the tail; (2) edge-on disks showing significant curvature are assigned velocity vec- c (cid:13) , 000–000 am-Pressure Stripping in Massive Clusters . s . s . s . s . . . . . h m . s − ◦ .
296 kpc10 kpc . s . s . s . s . s . s . s . s . . . . . h m . s ◦ .
548 kpc10 kpc . s . s . s . s . s . s . . . . . h m . s ◦ .
897 kpc10 kpc . s . s . s . s . s . . . . h m . s − ◦ .
218 kpc10 kpc . s . s . s . s . s . . . . h m . s − ◦ .
337 kpc10 kpc . s . s . s . s . . . . h m . s − ◦ .
204 kpc10 kpc . s . s . s . s . s . . . . . h m . s − ◦ .
408 kpc10 kpc . s . s . s . s . s . s . . . . h m . s ◦ .
308 kpc10 kpc . s . s . s . s . s . s . . . . . h m . s − ◦ .
334 kpc10 kpc
Figure 5.
Nine additional textbook examples of ram-pressure stripping discovered in this work; the first three of these were previously identified but notpublished by Ebeling, Stephenson & Edge (2014) (see also Fig. 1). The blue, green, and red arrows indicate the direction of motion assigned to the respectivegalaxy by the three reviewers; the yellow arrow and metric separation denote the direction and distance to the cluster centre (unknown to the reviewers). tors oriented perpendicular to said curvature and originating at itsapex; (3) if extended regions of star formation appear to be present,the velocity vector is placed perpendicular to the dominant elonga-tion of said regions; (4) if none of the previously mentioned indica-tors are present (or if they are contradictory), we attempt to makethe best physically motivated estimate. To avoid systematic biases,galaxies are inspected using small thumbnail images covering onlythe region immediately surrounding the galaxy with no indicationof the direction to the cluster centre. In recognition of the subjectivenature of our visual measurements (especially for galaxies movingpartly or largely along our line of sight, Roediger & Br¨uggen 2006),the process is performed independently by three reviewers to derivean approximate grade for the robustness of each estimated directionof motion. Figure 7 shows examples of objects falling into each ofour quality grades with uncertainty increasing top to bottom and left to right. We then define the angle of incidence as the anglebetween the apparent velocity vector and the position vector withrespect to the cluster centre (taken to be the location of the bright-est cluster galaxy, BCG), i.e., the angular deviation from a purelyradial infall trajectory (note again that all of these quantities aredefined and observed in projection).A second galaxy property that is critical to our efforts to de-duce trajectories is location within the cluster. For RPS candi-dates lacking radial-velocity measurements, we are unable to as-sess whether an object is located in front or behind the cluster cen-tre (defined by the redshift of the BCG), let alone further constrainits physical distance to the latter along our line of sight. Projecteddistances, however, measured in the plane of the sky and relative tothe location of the BCG, are trivially obtained for comparison withthe distribution expected for different geometries of galaxy infall. c (cid:13) , 000–000 C. McPartland et al.
Figure 6.
Distribution of gas (white) and newly formed stars (turquoise) fora simulated RPS event involving a spiral galaxy moving face-on throughthe ICM. A comparison with Fig. 1 shows that our morphological selectionis, unsurprisingly, most sensitive to features typical of mature RPS eventsin galaxies viewed edge-on. (Reproduced from Kronberger et al. 2008).
In order to understand which kind of galaxy trajectories are mostcompatible with the observed distributions of (projected) incidenceangle and cluster-centric distance, we compare our observationswith the results of a simple theoretical model. To this end, we calcu-late orbits in a canonical cluster representative of the MACS clus-ters in our sample and use simple prescriptions, described below, topredict the projected radii and incidence angles at which extremeRPS events are most likely to occur.As an infalling galaxy approaches the dense cluster core, theICM exerts an increasing ram-pressure, p ram = ρ ICM v , where ρ ICM is the ICM mass density and v gal is the relative velocity be-tween the galaxy and ICM (Gunn & Gott 1972, hereafter GG). Bycomparing p ram to the gravitational restoring force per unit area onthe gas within the galaxy, f grav ( R ) = Σ gas ( R ) ∂ Φ ∂Z ( R ) , (1)we find the critical radius where p ram = f grav ( R strip ) (Roediger &Br¨uggen 2007). Here Σ gas , Φ , and Z are the ISM mass surface den-sity, the gravitational potential of the galaxy, and its scale height, re-spectively. Beyond R strip , the galaxy potential is not strong enoughto retain the gas and stripping sets in. Vollmer et al. (2001) give ananalytic estimate for the GG criterion which determines the strip-ping radius: Σ gas v R − = p ram , (2)where v rot is the rotation speed of the galaxy. Although, in reality,the onset of RPS is likely to be a highly non-linear process, the sim-ple GG criterion has proven suitable for global characterisations of . s . s . s . s . . . . h m . s ◦ .
360 kpc10 kpc . s . s . s . s . s . . . . h m . s − ◦ .
337 kpc10 kpc . s . s . s . s . s . . . . h m . s − ◦ .
462 kpc10 kpc . s . s . s . s . s . s . . . . h m . s ◦ .
288 kpc10 kpc
Figure 7.
Examples of RPS candidate events illustrating our process to es-timate direction of motion and the associated error. The arrows are the sameas in Fig. 5.
RPS in in-depth numerical simulations (e.g., Roediger & Br¨uggen2007; Kronberger et al. 2008).
Since our simple model aims only to predict the distribution of RPSevents along galaxy orbits, but not the detailed properties of suchevents, we model all galaxies in our simulation as thin disks withradius R gal = 15 kpc and gas surface density Σ gas = atomsper cm moving face-on through the ICM.To account for galaxy-to-galaxy variation in f grav , we alsorun all models for a range of galaxy masses, parametrized by therotational velocity v rot (see Eq. 2). The explored range of v rot from150 to 350 km s − corresponds to dynamical masses, within 15kpc, of × , × , and × M (cid:12) . The adopted rangeof rotational velocities covers a spectrum of masses from sub- tosuper-Milky Way sized objects. We describe the gas and total mass distribution within the clusterusing a spherical β -model (Cavaliere & Fusco-Femiano 1976) ρ = ρ (cid:34) (cid:18) rr (cid:19) (cid:35) − β , (3)where ρ is the central mass (or gas) density, β and r are thepower-law index and core radius, respectively, and r is the cluster-centric radius. We adopt a total mass of . × M (cid:12) (the av-erage weak-lensing mass, M ( r < . Mpc ) , of MACS clusters at z > . according to Applegate et al. 2014). As the majority ( ∼ r and power-law slope β . We adopt r = 180 kpc and β = 0.59, the median of the values from the spatial X-ray analysis of c (cid:13)000
Since our simple model aims only to predict the distribution of RPSevents along galaxy orbits, but not the detailed properties of suchevents, we model all galaxies in our simulation as thin disks withradius R gal = 15 kpc and gas surface density Σ gas = atomsper cm moving face-on through the ICM.To account for galaxy-to-galaxy variation in f grav , we alsorun all models for a range of galaxy masses, parametrized by therotational velocity v rot (see Eq. 2). The explored range of v rot from150 to 350 km s − corresponds to dynamical masses, within 15kpc, of × , × , and × M (cid:12) . The adopted rangeof rotational velocities covers a spectrum of masses from sub- tosuper-Milky Way sized objects. We describe the gas and total mass distribution within the clusterusing a spherical β -model (Cavaliere & Fusco-Femiano 1976) ρ = ρ (cid:34) (cid:18) rr (cid:19) (cid:35) − β , (3)where ρ is the central mass (or gas) density, β and r are thepower-law index and core radius, respectively, and r is the cluster-centric radius. We adopt a total mass of . × M (cid:12) (the av-erage weak-lensing mass, M ( r < . Mpc ) , of MACS clusters at z > . according to Applegate et al. 2014). As the majority ( ∼ r and power-law slope β . We adopt r = 180 kpc and β = 0.59, the median of the values from the spatial X-ray analysis of c (cid:13)000 , 000–000 am-Pressure Stripping in Massive Clusters Table 2.
Model ParametersModel v (cid:107) [km s − ] σ v [km s − ] b max [Mpc]Stream-fed 200 100 1.5Slow Merger 1000 1000 2.5Fast Merger 3000 1000 2.5See Fig. 8 for a schematic illustration of v (cid:107) and b max ; σ v indi-cates the velocity dispersion of infalling galaxies. Mantz et al. (2010). Assuming a gas fraction f gas = 0 . (Mantzet al. 2014) and the model parameters above, our canonical clusterhas a central particle density n of . × − cm -3 . The orbits of test particles falling into our model cluster are com-puted for a wide range of initial orbital parameters that encompassexpectations for infall along connected filaments and from clustermergers. Orbit calculations begin at the end of a filament whichis assumed to be at a distance of 2.5 Mpc from the cluster core( ≈ R vir ). In Fig. 8, we show a schematic of the quantities thatcharacterise orbits in our model: the speed of a galaxy in the direc-tion of the filament axis v (cid:107) , the transverse velocity perpendicular tothe filament flow v ⊥ , and the impact parameter b .Radial profiles of filaments in cosmological simulations showa well defined edge at a radius of 1.0–2.0 h − Mpc ( ∼ b max .We populate these filaments with × galaxies using MonteCarlo sampling designed to provide constant density within b max and a normal distribution in v ⊥ to account for the velocity disper-sion of galaxies within the filament.In the following, we consider three infall scenarios that differprimarily in the approach velocity of galaxies at the cluster’s virialradius: 1) stream-fed infall along filaments; 2) a slow merger ; and3) a fast merger . Table 2 lists the model parameters that characteriseeach of these scenarios. For each infall scenario, we fix the initialvelocity v (cid:107) at one value for all orbits. For the stream-fed model, wechoose b max =1.5 Mpc and v (cid:107) = 200 km s − , the average filamentradius and the average velocity of matter at the cluster-filament in-terface, respectively (Colberg, Krughoff & Connolly 2005), as wella velocity dispersion characteristic of group environments ( ∼ − ). The slow and fast merger models are characterised byinitial velocities of 1000 and 3000 km s − , respectively, and a ve-locity dispersion of 1000 km s − and b max = 2.5 Mpc for eithermerger scenario. Since we know neither the number and orientationof connected filaments for our cluster sample, nor the orientation ofa putative merger axis, we place filaments/merging clusters at positions, sampled isotropically on a 2.5 Mpc sphere. In total, thisresults in × orbits per scenario which are each followed for 5Gyr ( ∼ t cross ) in time steps of 5 Myr.Defining the start of the RPS event as the time step in whichthe GG criterion is first satisfied, we explore a range of RPS eventdurations, from 50 Myr to 1 Gyr, during which the resulting event isassumed to remain observationally detectable. This choice is mo-tivated by numerical simulations: Roediger et al. (2014) find thatthe signature RPS morphology should be observable in galaxiesoverrun by an ICM shock for between ∼ several 10 Myr to a few100 Myr. Slightly longer durations are quoted by Kronberger et al.(2008) for a scenario similar to our stream-fed infall model (seealso Fig. 6). x y b max ~v~v i~r Figure 8.
Schematic diagram of the quantities that characterise the initialconditions and orbits of galaxies in our infall models: the maximal impactparameter b max , the initial velocities v (cid:107) and v ⊥ , the cluster-centric radius ˆ r , and the inclination angle i . For comparison with our observational results, segments of theorbits corresponding to an RPS event (under our definition) are pro-jected onto the plane of the sky, thus providing the projected angleof incidence (the projected angle between the galaxy’s velocity andposition vectors), i , and the projected radius from the cluster cen-tre. We then tabulate the amount of time spent in bins of projectedradius and inclination angle to construct simulated probability dis-tributions for each scenario. As mentioned in Section 3.2.1 and illustrated in Fig. 6, RPS eventsin galaxies moving along or close to our line of sight are likely tobe missed, as, for this particular geometry, the pronounced mor-phological features that our selection process is build upon are ob-scured by the galaxy being stripped. We examine the importance ofthis observational bias by imposing on our modeling results that allRPS events are undetectable that occur in galaxies moving along anaxis that is inclined to our line of sight by 0, 15, 30, or 45 degrees.As detailed in the following section, even the most severe imple-mentation of this line-of-sight bias results in only modest changesin the model predictions, suggesting that the effect does not signif-icantly affect the conclusions drawn from our comparison with thedata.
In order to reduce contamination by interlopers (fore- or back-ground galaxies), we restrict our analysis to the subset of candi-date RPS events with measured redshifts within ± − ofthe redshift of the host cluster; the 53 objects (of 124 with mea-sured redshifts) meeting this criterion are hereafter referred to asthe “spectroscopic sample”. Of these, we select a subset of the15 galaxies exhibiting the most compelling “jellyfish” morphologycomprised of the six systems presented by ESE and the nine shownin Fig. 5 (“jellyfish sample”). We further restrict the comparisonbetween data and model predictions to a projected radius of 415kpc from the cluster core, which leaves 23 and 11 galaxies in the c (cid:13) , 000–000 C. McPartland et al. C u m u l a t i v e F r a c t i o n Stream − fed Slow − Merger
All events
Fast − Merger i [ degrees ] LOS bias corrected
Figure 9.
Cumulative distribution of the incidence angles of the jellyfish and cluster-member samples (asterisks and squares, respectively). The left, center,and right panels show predictions for the stream-fed, slow-merger, and fast-merger models (see Table 2), respectively. The dotted, dashed, and solid linescorrespond to event durations of 50 Myr, 300 Myr, and 1 Gyr, respectively. Colors denote the mass of the infalling galaxy: blue (thin), green (medium), andred (thick) correspond to dynamical masses of × , × , and × M (cid:12) , respectively. Model predictions shown in the top row assume that RPSevents are identifiable as such regardless of the inclination of the galaxy’s direction of motion with respect to our line of sight; results shown in the bottom rowmimic the observational bias discussed in Sections 3.2.1 and 4.4 by excluding all events triggered in galaxies with velocity vectors within 30 degrees of ourline of sight. C u m u l a t i v e F r a c t i o n Stream − fed Slow − Merger
All events
Fast − Merger r proj [ kpc ]
100 200 300 350 400
LOS bias corrected
Figure 10.
As Fig. 9 but for the projected radius, r proj . c (cid:13)000
As Fig. 9 but for the projected radius, r proj . c (cid:13)000 , 000–000 am-Pressure Stripping in Massive Clusters r p r o j K S P r o b a b ili t y τ event [ Gyr ] i K S P r o b a b ili t y Stream − fed × M fl × M fl × M fl Slow − Merger Fast − Merger
Figure 11.
KS model probabilities for the projected incidence angle i (top row) and the projected radius r pro (bottom row) for the sample of cluster members(green squares in Figs. 9 and 10), shown as a function of the duration of the RPS event, τ event , and the mass of the respective galaxy (see legend). Infall alongfilaments (leftmost panels) is clearly disfavoured. spectroscopic and jellyfish samples respectively. This radial cutoffminimises systematic incompleteness introduced at larger cluster-centric radii, which are covered only by images of the most distantclusters in our sample.Fig. 9 shows the cumulative distributions of the incidence an-gle for our two RPS subsamples plotted against predictions fromour infall model, with (bottom row) and without (top row) correc-tion for the bias discussed in Section 3.2.1 and 4.4. The bottomThe left, center, and right columns of Fig. 9 show predictions forthe stream-fed, slow-merger, and fast-merger models, respectively(see Table 2 for the parameters characterising these models).Visual comparison suggests that the observations are bestmatched by the model predictions for the slow-merger scenario,provided that the duration of the stripping process is less than aGyr . Contrary to the traditional picture of RPS being driven purelyby infall from the low-density field, preferably along filaments, wefind poor agreement between the data and stream-fed models whichover-predict events at extreme incidence angles (at (cid:46) ◦ for almostall combinations of model parameters explored by us, and at (cid:38) ◦ for low-mass galaxies experiencing long RPS events). In this sce-nario, the motion of galaxies is dominated by the cluster poten- Note that, in the top panel of Fig. 9, all of the solid lines, as well as thered dashed line, fall on top of each other and are thus indistinguishable byeye. tial, which leads to a preferential alignment of trajectories towardthe cluster core (at least in our projected view) and thus a highlyanisotropic distribution of incidence angles.Fig. 10 shows the cumulative distributions of the number ofRPS events within a given projected cluster-centric radius. To pro-vide more natural, equal-area sampling, we bin the data in equalsteps of r ; a uniform areal distribution thus appears as a straightline from zero to one. We find that both of the cluster merger mod-els predict a nearly uniform areal distribution of events in agree-ment with our observations. Stream-fed models with the most mas-sive galaxies and/or the longest event timescales predict an excessat small projected radii which is not supported by our data. Note,that this comparison also effectively rules out the stream-fed modelwith a Milky-Way sized galaxy and 300 Myr timescale that at leastmarginally matched the observed distribution of incidence anglesand is shown as the green dashed line in Fig. 9.A more quantitative assessment of the significance of the dis-crepancies between the observed and predicted distributions canbe obtained with Kolmogorov-Smirnov (KS) tests. In Fig. 11, weshow KS probabilities for the null hypothesis that the observed dis-tributions are drawn from the same parent population as the pre-dictions of a given model. Correcting all models for the aforemen-tioned line-of-sight bias (Sections 3.2.1 and 4.4) does not changeour conclusions significantly. For simplicity, we therefore ignorethe bias due to motion along the line of sight in the KS tests. Tomaximize the number of objects in the comparison, we show re- c (cid:13) , 000–000 C. McPartland et al. sults for the spectroscopic sample only. However, considering thesmaller jellyfish subsample does not significantly alter our conclu-sions.Consistent with our qualitative assessments above, we find noagreement with the observed distribution of incidence angles forany model assuming infall along filaments, although the distribu-tion of projected radii does not rule out such models (at least notfor low-mass galaxies, see bottom panel of Fig. 11). By contrast,practically all of the models for the two merger scenarios providean acceptable (or good) description of the data, with the exceptionof those involving the most massive galaxies, for which models as-suming long RPS durations of τ event (cid:38)
300 Myr are ruled out atmore than σ confidence. Since our models are intrinsically three dimensional, the compar-isons presented above, although involving solely parameters mea-sured in projection, allow us to distinguish between distinctly dif-ferent three-dimensional scenarios.In the merger scenarios, RPS events are triggered in fast-moving galaxies near the outskirts of the cluster and, due to the rel-atively short duration of ∼
500 Myr required by our incidence angledata (see top row of Fig. 11), remain confined to a shell well out-side a (three-dimensional) cluster-centric radius of 400 kpc. On theother hand, the projected radius data favour event durations longerthan ∼
100 Myr to explain the uniform areal distribution (Fig. 10).The RPS candidates detected by us are thus the projection of the es-sentially uniform distribution of much more distant RPS events inthe fore- and background segments of this shell. In principle, galax-ies of all masses may contribute to the observed RPS distribution;however, the majority are likely to be systems of low to interme-diate mass, since models for extremely massive galaxies generallyrequire finely tuned, short RPS lifetimes of about 100 Myr approx-imately to match the observations (red lines in Fig. 11).By contrast, galaxies falling into the cluster along filaments doso at much lower peculiar velocities and thus require higher ICMdensities for the GG criterion to be met; as a result, RPS events aretriggered only much closer to the cluster core. To match the ob-served, broad distribution of incidence angles, these galaxies needtime to enter our field of view from all sides, which mandates thatthe associated RPS events remain observable for 300 Myr or longer(Fig. 9). Such long life-times, however, lead in turn to an excess inthe number of events close to the cluster core that is not observed(Fig. 10).We therefore tentatively conclude that extreme RPS events inmassive clusters are generally short-lived ( (cid:46)
500 Myr) and trig-gered far from the cluster core, likely driven by cluster mergers.Interestingly this preference of our analysis for RPS events beingmost readily observed in galaxies moving at high speed through anonly modestly dense ICM suggests that textbook cases of “jelly-fish galaxies” might also be observed near the cores of less massiveclusters (or even groups of galaxies, see also Poggianti et al. 2015)provided a cluster or group merger event ensures sufficiently highpeculiar initial velocity. Note also that, while our data disfavour in-fall along filaments as the primary trigger, they do not rule out acontribution from such a scenario. Wide-field imaging surveys thatare able to detect RPS events out to the virial radius are needed todetermine the relative contributions of stream-fed infall and clustermergers.
We have conducted a systematic search for galaxies experiencingram-pressure stripping (RPS) in 63 MACS clusters at z = ∼ Hubble Space Telescope images of these systemswe identify 211 potential cases of RPS that complement a trainingset of 12 “jellyfish” galaxies used to define our selection criteria.Where possible, the direction of motion in the plane of the sky isestimated for these systems based on morphological indicators suchas the curvature and orientation of the apparent galaxy-ICM inter-face region or a visible debris trail. Several systematic biases areinherent to our approach: (a) the classification of galaxies accord-ing to their likelihood of undergoing RPS is partly based on visualinspection and thus to some extent subjective, (b) the small field offield of view our observations prevents us from sampling the galaxypopulation in the outer regions of our cluster targets (except in pro-jection) where RPS events might be initially triggered, and (c) ourselection process is fundamentally unable to robustly identify RPSevents in galaxies moving along, or close to, our line of sight.We attempt to address the first of these biases by obtainingspectroscopic redshifts of all our RPS candidates. While the re-sulting spectra do not immediately confirm or refute an RPS event,they allow us to establish whether or not a morphologically selectedcandidate is in fact a cluster member and whether its spectral char-acteristics are consistent with ongoing or recent star formation. Sofar, 53 of 124 systems targeted in spectroscopic follow-up obser-vations were confirmed as cluster members. A detailed analysis ofthese galaxies’ spectral properties will be presented in a forthcom-ing paper (Blumenthal et al., in preparation).The remaining two observational biases mentioned above canbe accounted for by three-dimensional modelling of the trajectoriesand environment of galaxies falling into a massive cluster. Specif-ically, we compare the distributions of the observed projected in-cidence angle and distance from the BCG with predictions fromsimple models of galaxy orbits in a MACS-like cluster. We investi-gate two scenarios: accretion of galaxies from an attached filament,and a cluster merger event.We find significantly better agreement for the merger scenario,provided the duration of RPS events is (cid:46)
500 Myr. We thus ten-tative conclude that extreme ram-pressure stripping events is pri-marily triggered in massive cluster mergers (rather than by infallalone) where relative velocities between galaxies and the ICM arelarge enough to initiate RPS far from the cluster core ( (cid:29)
400 kpc).Although our study is, by design, limited to relatively massive clus-ters, we note that this result implies that extreme RPS events mayalso occur in mergers of poorer clusters and even groups of galax-ies, where the required ingredients (high peculiar velocity and mod-erately high ICM density) are both met by galaxies close to corepassage. We also find that galaxies of mass similar to, or less than,our Milky Way are likely to dominate the set of observable RPSevents in massive clusters, although more massive galaxies maycontribute too at a lower level. Although models assuming infallalong a filament were found to yield predictions that are largelyin conflict with our data, both processes (accretion along filamentsand via cluster mergers) can be expected to contribute. The extent towhich the two mechanisms are responsible for the observed popu-lation of RPS events in our sample is difficult to quantify but couldbe tested by imaging surveys that probe the distribution of RPSevents to larger cluster-centric radii.In-depth studies of the X-ray properties of RPS host clustersalong with spectroscopic investigations of the star-formation rates c (cid:13) , 000–000 am-Pressure Stripping in Massive Clusters and histories of the candidates identified in this study will be crit-ical to test our conclusions and allow a quantitative comparison ofobservational diagnostics with predictions of numerical models ofram-pressure stripping. ACKNOWLEDGEMENTS
We thank the anonymous referee for their helpful comments, ques-tions, and suggestions on revising the manuscript. CM thanks J.Lotz for providing the galaxy morphology source code which wasadapted for this work. HE gratefully acknowledges financial sup-port from STScI grants GO-10495, -10875, -12166, and -12884.This research made use of Astropy, a community-developed corePython package for Astronomy (Astropy Collaboration, 2013).
APPENDIX A: SKELETAL DECOMPOSITIONPARAMETERS
The morphological indicators discussed in Section 3 were gener-ally defined to identify characteristic morphological traits of galaxymergers (e.g. Lotz et al. 2011). We introduce a new metric basedon the concept of the morphological skeleton (Maragos & Schafer1986) to both quantify the amount of substructure in a galaxy whileconcurrently identifying arm/tail-like structures. Conceived in thecontext of mathematical morphology (see Serra 1988) and orig-inally introduced as a means for binary image compression, themorphological skeleton (or medial axis transform) reduces a shapeto a line that maintains the topological structure of the full image,thus allowing exact reconstruction.We here generalise the definition of the morphological skele-ton to images with non-binary, continuous greyscale pixel values.However, we must be cautious as noise in relatively short expo-sures used in this survey ( ∼ Sk i . We perform skeletal decompositions under threesmoothing scales corresponding to the Petrosian radius r p , the halflight radius r , and the 10% light radius r which define Sk , Sk , and Sk . Note that due to the cleaning process we apply here,exact reconstruction of the original image is not possible.To further reduce erroneous signal due to residual noise, wedefine Sk x + y (where y = x + 1 ) as comprising all pixels in thehigher-order skeleton (i.e. under a smaller smoothing kernel) con-nected to that of the lower-order skeleton (larger smoothing kernel).To generate a common reference point and to avoid bias due to im-age size and lower order structure, we then subtract the length of thelower-order skeleton from Sk x + y and normalise by the length ofthe lower-order skeleton (e.g. [ | Sk | − | Sk | ] / | Sk | ) defining afinal numerical measure Sk x − y which quantifies the excess in sub-structure under smoothing scale y with respect to x (see Fig. A1).A simple way to understand this qualitatively is to considera case where Sk − or Sk − is equal to zero. This would im-ply that image smoothed on a finer scale (smaller kernel) does notreveal any more substructure or that the galaxy’s light profile is es-sentially smooth below the upper smoothing scale. However, as afull interpretation of the meaning and reliability of these indicatorsis beyond of the scope of this paper, we here characterise Sk − only to be a measure of bending in the galaxy or the deviation from Sk Sk Sk − Sk − = 1.18 Sk Sk Sk − Sk − = 2.03 Figure A1.
An example of the results of the greyscale skeletonization pro-cess which we use to define the skeletal decomposition parameters Sk − and Sk − . a symmetric object (somewhat correlated with asymmetry), while Sk − quantifies the amount of clumpy substructure connected tothe brighter regions of the galaxy. REFERENCES
Abadi M. G., Moore B., Bower R. G., 1999, MNRAS, 308, 947Abraham R. G., van den Bergh S., Nair P., 2003, ApJ, 588, 218Abramson A., Kenney J. D. P., Crowl H. H., Chung A., vanGorkom J. H., Vollmer B., Schiminovich D., 2011, AJ, 141, 164Applegate D. E. et al., 2014, MNRAS, 439, 48Bershady M. A., Jangren A., Conselice C. J., 2000, AJ, 119, 2645Bertin E., Arnouts S., 1996, A&AS, 117, 393Boselli A., Gavazzi G., Lequeux J., Buat V., Casoli F., Dickey J.,Donas J., 1997, A&A, 327, 522Butcher H., Oemler, Jr. A., 1984, ApJ, 285, 426Cavaliere A., Fusco-Femiano R., 1976, A&A, 49, 137Chung A., van Gorkom J. H., Kenney J. D. P., Vollmer B., 2007,ApJ, 659, L115Colberg J. M., Krughoff K. S., Connolly A. J., 2005, MNRAS,359, 272Conselice C. J., 2003, ApJS, 147, 1Cort´es J. R., Kenney J. D. P., Hardy E., 2015, ApJS, 216, 9Cortese L. et al., 2007, MNRAS, 376, 157Dressler A., 1980, ApJ, 236, 351Ebeling H., Barrett E., Donovan D., Ma C.-J., Edge A. C., vanSpeybroeck L., 2007, ApJ, 661, L33Ebeling H., Edge A. C., Henry J. P., 2001, ApJ, 553, 668Ebeling H., Edge A. C., Mantz A., Barrett E., Henry J. P., MaC. J., van Speybroeck L., 2010, MNRAS, 407, 83Ebeling H., Stephenson L. N., Edge A. C., 2014, ApJ, 781, L40Fumagalli M., Fossati M., Hau G. K. T., Gavazzi G., Bower R.,Sun M., Boselli A., 2014, MNRAS, 445, 4335Gunn J. E., Gott, III J. R., 1972, ApJ, 176, 1Kapferer W., Sluka C., Schindler S., Ferrari C., Ziegler B., 2009,A&A, 499, 87Kronberger T., Kapferer W., Ferrari C., Unterguggenberger S.,Schindler S., 2008, A&A, 481, 337Lotz J. M., Jonsson P., Cox T. J., Croton D., Primack J. R.,Somerville R. S., Stewart K., 2011, ApJ, 742, 103Lotz J. M., Primack J., Madau P., 2004, AJ, 128, 163 c (cid:13) , 000–000 C. McPartland et al.
Mann A. W., Ebeling H., 2012, MNRAS, 420, 2120Mantz A., Allen S. W., Ebeling H., Rapetti D., Drlica-Wagner A.,2010, MNRAS, 406, 1773Mantz A. B., Allen S. W., Morris R. G., Rapetti D. A., Apple-gate D. E., Kelly P. L., von der Linden A., Schmidt R. W., 2014,MNRAS, 440, 2077Maragos P., Schafer R. W., 1986, Acoustics, Speech and SignalProcessing, IEEE Transactions on, 34, 1228Merluzzi P. et al., 2013, MNRAS, 429, 1747Moore B., Katz N., Lake G., Dressler A., Oemler A., 1996, Na-ture, 379, 613Owen F. N., Keel W. C., Wang Q. D., Ledlow M. J., MorrisonG. E., 2006, AJ, 131, 1974Owers M. S., Couch W. J., Nulsen P. E. J., Randall S. W., 2012,ApJ, 750, L23Poggianti B. M. et al., 2015, ArXiv e-printsPostman M. et al., 2012, ApJS, 199, 25Roediger E., Br¨uggen M., 2006, MNRAS, 369, 567Roediger E., Br¨uggen M., 2007, MNRAS, 380, 1399Roediger E., Br¨uggen M., Owers M. S., Ebeling H., Sun M., 2014,MNRAS, 443, L114Scott T. C. et al., 2010, MNRAS, 403, 1175Serra J., 1988, Image analysis and mathematical morphology:Theoretical advances, Image Analysis and Mathematical Mor-phology. Academic PressSivanandam S., Rieke M. J., Rieke G. H., 2010, ApJ, 717, 147Smith R. J. et al., 2010, MNRAS, 408, 1417Steinhauser D., Haider M., Kapferer W., Schindler S., 2012,A&A, 544, A54Sun M., Donahue M., Roediger E., Nulsen P. E. J., Voit G. M.,Sarazin C., Forman W., Jones C., 2010, ApJ, 708, 946Sun M., Donahue M., Voit G. M., 2007, ApJ, 671, 190Takeda H., Nulsen P. E. J., Fabian A. C., 1984, MNRAS, 208, 261Vollmer B., Cayatte V., Balkowski C., Duschl W. J., 2001, ApJ,561, 708Vollmer B., Wong O. I., Braine J., Chung A., Kenney J. D. P.,2012, A&A, 543, A33Yagi M. et al., 2010, AJ, 140, 1814 c (cid:13) , 000–000 am-Pressure Stripping in Massive Clusters Table A1.
SNAPS Observations Name α [J2000] δ [J2000] t exp [s] GO Prop. IDEMACSJ1057.5+5759 10:57:31.680 +57:59:33.72 1200 12884MACSJ0032.1+1808 00:32:11.344 +18:07:49.37 1200 12166MACSJ0035.4-2015 00:35:26.957 -20:15:50.66 1200 10491MACSJ0140.0-0555 01:40:01.626 -05:55:06.71 1200 10491MACSJ0152.5-2852 01:52:35.361 -28:53:39.88 1200 10491MACSJ0257.6-2209 02:57:40.596 -22:09:27.80 1200 10875MACSJ0308.9+2645 03:08:56.839 +26:45:43.91 1200 12166MACSJ0451.9+0006 04:51:55.443 +00:06:11.66 1200 10491MACSJ0521.4-2754 05:21:25.808 -27:55:06.91 1200 10491MACSJ0547.0-3904 05:47:01.796 -39:04:13.24 1200 12166MACSJ0553.4-3342 05:53:23.850 -33:42:42.21 2092 12362MACSJ0712.3+5931 07:12:21.985 +59:32:24.82 1200 10491MACSJ0845.4+0327 08:45:28.224 +03:27:28.46 1200 10491MACSJ0916.1-0023 09:16:12.344 -00:23:47.00 1200 10491MACSJ0947.2+7623 09:47:10.744 +76:23:21.62 1200 10491MACSJ0949.8+1708 09:49:52.655 +17:07:06.38 1200 10491MACSJ1006.9+3200 10:06:55.632 +32:01:33.91 1200 10491MACSJ1115.2+5320 11:15:15.968 +53:19:47.47 1200 10491MACSJ1124.5+4351 11:24:29.365 +43:51:32.97 1200 12166MACSJ1133.2+5008 11:33:14.109 +50:08:29.50 1200 10491MACSJ1142.4+5831 11:42:26.434 +58:32:01.30 1200 12166MACSJ1226.8+2153C 12:26:41.421 +21:53:07.58 1200 12166MACSJ1236.9+6311 12:36:59.868 +63:11:02.26 1200 10491MACSJ1258.0+4702 12:58:02.708 +47:02:42.87 1200 10491MACSJ1319.9+7003 13:20:09.685 +70:04:28.16 1200 10491MACSJ1354.6+7715 13:54:31.253 +77:15:08.71 1200 10491MACSJ1447.4+0827 14:47:26.289 +08:28:37.08 1200 12166MACSJ1452.9+5802 14:52:57.957 +58:02:43.28 1200 12166MACSJ1526.7+1647 15:26:42.342 +16:47:48.83 1200 12166MACSJ1621.3+3810 16:21:23.928 +38:10:16.28 1200 12166MACSJ1644.9+0139 16:45:01.729 +01:40:09.83 1200 12166MACSJ1652.3+5534 16:52:19.726 +55:34:46.63 1200 10491MACSJ1731.6+2252 17:31:39.268 +22:52:05.09 1200 12166MACSJ1738.1+6006 17:38:05.383 +60:06:14.92 1200 12166MACSJ1752.0+4440 17:51:57.961 +44:39:45.45 1200 12166MACSJ1806.8+2931 18:06:51.898 +29:30:23.03 1200 12166MACSJ2050.7+0123 20:50:42.381 +01:23:24.69 1200 12166MACSJ2051.1+0215 20:51:10.058 +02:16:00.72 1200 12166MACSJ2135.2-0102 21:35:12.822 -01:02:51.52 1200 10491MACSJ2241.8+1732 22:41:56.386 +17:32:47.33 1200 12166SMACSJ0234.7-5831 02:34:43.512 -58:31:16.51 1200 12166SMACSJ0549.3-6205 05:49:18.358 -62:05:07.88 1200 12166SMACSJ0600.2-4353 06:00:12.915 -43:53:19.33 1200 12166SMACSJ0723.3-7327 07:23:18.709 -73:27:06.01 1200 12166SMACSJ2031.8-4036 20:31:46.993 -40:37:03.68 1200 12166SMACSJ2131.1-4019 21:31:05.693 -40:19:12.22 1200 12166c (cid:13) , 000–000 C. McPartland et al.
Table A2.
CLASH Observations Name α [J2000] δ [J2000] t exp [s] GO Prop. IDMACSJ0329-0211 03:29:41.560 -02:11:46.10 4104 12452MACSJ0416-2403 04:16:08.380 -24:04:20.79 4036 12459MACSJ0429-0253 04:29:36.049 -02:53:06.10 3938 12788MACSJ0647+7015 06:47:50.269 +70:14:54.99 4128 12101MACSJ0717.5+3745-POS5 07:17:32.629 +37:44:59.70 7920 10420MACSJ0744+3927 07:44:52.819 +39:27:26.89 4128 12067MACSJ1115+0129 11:15:51.900 +01:29:55.10 3870 12453MACSJ1149+2223 11:49:34.704 +22:24:04.75 4128 12068MACSJJ1206.2-0847 12:06:12.055 -08:47:59.44 6608 10491MACSJ1311-0310 13:11:01.800 -03:10:39.79 4158 12789RXJ1347-1145 13:47:32.110 -11:45:11.36 3878 12104MACS1423+2404 14:23:47.88 +24:04:42.49 4240 12790RXJ1532+3021 15:32:53.779 +30:20:59.39 4060 12454MACSJ1720+3536 17:20:16.780 +35:36:26.49 4040 12455MACSJ1931-2635 19:31:49.62 -26:34:32.90 3850 12456MACSJ2129-0741 21:29:26.059 -07:41:28.79 3728 12100RXJ2248-4431 22:48:43.960 -44:31:51.30 3976 12458c (cid:13)000