Star Formation and Supercluster Environment of 107 Nearby Galaxy Clusters
S. A. Cohen, R. C. Hickox, G. A. Wegner, M. Einasto, J. Vennik
AA CCEPTED FOR PUBLICATION IN T HE A STROPHYSICAL J OURNAL
Preprint typeset using L A TEX style emulateapj v. 5/2/11
STAR FORMATION AND SUPERCLUSTER ENVIRONMENT OF 107 NEARBY GALAXY CLUSTERS S ETH
A. C
OHEN , R
YAN
C. H
ICKOX , G
ARY
A. W
EGNER
Department of Physics and Astronomy, Dartmouth College, 6127 Wilder Laboratory, Hanover, NH 03755, USA M ARET E INASTO , J
AAN V ENNIK
Tartu Observatory, 61602 T˜oravere, Estonia
Accepted for publication in The Astrophysical Journal
ABSTRACTWe analyze the relationship between star formation (SF), substructure, and supercluster environment in asample of 107 nearby galaxy clusters using data from the Sloan Digital Sky Survey. Previous works haveinvestigated the relationships between SF and cluster substructure, and cluster substructure and superclusterenvironment, but definitive conclusions relating all three of these variables has remained elusive. We find aninverse relationship between cluster SF fraction ( f SF ) and supercluster environment density, calculated usingthe galaxy luminosity density field at a smoothing length of 8 h − Mpc (D8). The slope of f SF vs. D8is − . ± . . The f SF of clusters located in low-density large-scale environments, . ± . , ishigher than for clusters located in high-density supercluster cores, . ± . . We also divide superclusters,according to their morphology, into filament- and spider-type systems. The inverse relationship between cluster f SF and large-scale density is dominated by filament- rather than spider-type superclusters. In high-densitycores of superclusters, we find a higher f SF in spider-type superclusters, . ± . , than in filament-type superclusters, . ± . . Using principal component analysis, we confirm these results and thedirect correlation between cluster substructure and SF. These results indicate that cluster SF is affected byboth the dynamical age of the cluster (younger systems exhibit higher amounts of SF); the large-scale densityof the supercluster environment (high-density core regions exhibit lower amounts of SF); and superclustermorphology (spider-type superclusters exhibit higher amounts of SF at high densities). Subject headings: large-scale structure of universe – galaxies: clusters: general – galaxies: star formation INTRODUCTION
The effects of galaxy cluster mergers on star formation (SF)have begun to be better understood in recent years, addingdepth to the relationships found in relaxed clusters betweenSF and clustercentric distance and local density (e.g., Dressler1980; Cohen et al. 2014, hereafter C14; Cohen et al. 2015;and many others). While some studies find no relationship be-tween cluster merger activity and SF in specific clusters (e.g.,Metevier et al. 2000; Ferrari et al. 2005; Braglia et al. 2009;Hwang & Lee 2009; Kleiner et al. 2014), many others reportsuch a relationship (e.g., Knebe & M¨uller 2000; Cortese et al.2004; Ferrari et al. 2005; Johnston-Hollitt et al. 2008; Bravo-Alfaro et al. 2009; Braglia et al. 2009; Hwang & Lee 2009;Ma et al. 2010; Wegner 2011; Wegner et al. 2015; Sobral et al.2015; Girardi et al. 2015; Stroe et al. 2015). Indeed, C14 andCohen et al. (2015) found that SF is statistically correlated tocluster substructure in studies of large numbers of clusters: ingeneral, clusters with more substructure exhibit greater levelsof SF.Recent studies have investigated the relationship betweencluster substructure and supercluster environment (Einasto etal. 2015; Krause et al. 2013; Einasto et al. 2012, hereafterE12b). In particular, E12b found that clusters in superclustersare more likely to have substructure than those that are iso-lated, though the correlation discussed in the paper is weak.Studies have also begun to probe the connection between su-percluster environment and SF (e.g. Costa-Duarte et al. 2013;Luparello et al. 2013; Lietzen et al. 2012). In voids, there isa general consensus that this lower-density large-scale envi-ronment only weakly affects galaxy properties, which dependmore strongly on local environment (Grogin & Geller 2000; Rojas et al. 2005; Patiri et al. 2006; Wegner & Grogin 2008;Hoyle et al. 2012; Kreckel et al. 2011, 2012).In superclusters, Einasto et al. (2014) recently showed thatsupercluster morphology is important in shaping the prop-erties of galaxies: higher levels of SF are found in galaxiesin spider-type superclusters than filament-type superclusters.Simulations by Aragon-Calvo et al. (2014) suggest that thequenching of SF in clusters depends on the geometry of thelarge-scale surrounding structure. This supports observationalwork by Einasto et al. (2014): spider-type superclusters havericher inner structure and larger numbers of filaments con-necting galaxy clusters than do filament-type superclusters.However, a definitive relationship between supercluster en-vironment and SF has yet to be shown. We seek to developa coherent picture connecting these four variables: clusterstar-forming fraction ( f SF ), amount of cluster substructure,supercluster environment density, and supercluster morphol-ogy. This paper considers the correlations between these pa-rameters, focusing in particular on the pairwise comparisonbetween supercluster environment and SF. Furthermore, weseek to confirm the pairwise comparisons involving clustersubstructure and cluster SF, and cluster substructure and su-percluster density. Finally, we investigate a potential multi-dimensional correlation among the three non-morphologicalvariables. In §
2, we introduce our cluster sample and discussmethods for determining substructure, SF, and superclusterproperties; § §
4. Throughout our analysis weassume a standard cosmology of H = 100 h km s − Mpc − , Ω m = 0 . , and Ω Λ = 0 . . a r X i v : . [ a s t r o - ph . GA ] D ec Cohen et al. DATA AND METHODS
In this section, we describe our cluster and superclustersamples. We also explain our methods for calculating SF andsubstructure properties of clusters. Finally, we introduce ouruse of principal component analysis (PCA) in determining re-lationships between SF, substructure, and large-scale environ-ment.
Cluster sample
We use the sample of rich clusters from the group cata-logue of Tempel et al. (2012), which is based on the SDSSDR8 spectroscopic data (Aihara et al. 2011). Using SDSSdata, Tempel et al. (2012) identified 77,858 groups and clus-ters using the friends-of-friends (FoF) method (Zeldovich etal. 1982; Huchra & Geller 1982). Einasto et al. 2012 (here-after E12a) used the subsample of rich clusters with at least50 members in the redshift interval . ≤ z ≤ . to deter-mine the substructure properties of the clusters. They foundthat 90 of these clusters contain substructure and 17 do not.In the present paper, we use this cluster sample, previouslyanalyzed by C14. We obtain galaxy stellar masses from theMax Planck Institute (MPA)/Johns Hopkins University (JHU)VAGC (Tremonti et al. 2004), and both observed and esti-mated total r -band luminosities ( L obs and L tot , respectively)from the catalogue of Tempel et al. (2012).As the SDSS data are flux-limited, the FoF method poten-tially suffers from the bias of fainter galaxies vanishing as dis-tance increases. This leads to differences in the luminositiesof member galaxies between nearby and more distant groups.Tempel et al. (2012) partly corrected for this effect by deter-mining a relationship between distance and the linking lengthused in their FoF algorithm, and then applying this relationwhen selecting groups at different distances. They note that,by applying this correction, their final group catalogue is quitehomogeneous in richness, size, and velocity dispersion, re-gardless of distance. However, we note that this does notcorrect for the fact that groups of a given richness at lowerredshift are less luminous than those of the same richnessat higher redshift. This is because, spectroscopically, faintergalaxies are more easily detected in the SDSS – and thus in-cluded as group members – at lower redshift than higher red-shift. Despite this bias, in § M . r < − . . The determination of this ab-solute magnitude cut follows the methods of Hwang & Lee(2009). A galaxy’s r -band absolute magnitude is calculatedfrom its apparent magnitude m r via M . r = m r − DM − K ( z ) − E ( z ) , (1)where m r is corrected for extinction; DM ≡ log ( D L / pc ) and D L is a luminosity distance; K ( z ) is a K -correction(Blanton & Roweis 2007) to a redshift of 0.1, denoted bythe superscript; and E ( z ) is an evolution correction definedby E ( z ) = 1 . z − . (Tegmark et al. 2004). Extinction-corrected magnitudes and K -corrections are collected fromthe NYU Value-Added Galaxy Catalogue (VAGC; Blantonet al. 2005; Padmanabhan et al. 2008). Star formation and substructure determinations
C14 determined which galaxies are star-forming using thedetection of H α emission, defined as the measurement of an equivalent width of at least 3 ˚A (a compromise between, e.g.,Ma et al. 2008; Balogh et al. 2004; Rines et al. 2005). Rel-evant equivalent width and flux measurements were retrievedfrom the MPA/JHU VAGC (Tremonti et al. 2004). When pos-sible, they also used the BPT diagram (Baldwin et al. 1981)that uses the emission line ratios log([ OIII ] λ / H β ) vs. log([ NII ] λ / H α ) to separate star-forming galaxies fromAGN and LINERs (Kauffmann et al. 2003; Kewley et al.2001); the latter two types we remove from our analysis. Acluster’s f SF is defined as the number of star-forming galax-ies divided by the total number of galaxies in the cluster.Cluster substructure properties were determined by E12ausing multidimensional normal mixture modelling via the Mclust package for classification and clustering (Fraley &Raftery 2006).
Mclust assigns each member galaxy to a com-ponent, thus determining the number of components in eachcluster. E12a also analyzed the substructure properties of ourclusters using the Dressler-Shectman (DS or ∆ ) test (Dressler& Shectman 1988). In short, for each cluster, this test mea-sures how each galaxy’s local kinematics differ from the kine-matics of the cluster as a whole. The results of the test arethen calibrated using Monte Carlo simulations to determine a p -value, the probability that any observed substructure is dueto chance. Thus, smaller p -values indicate higher probabili-ties of substructure. Please see § ∆ test and its calibration. Large-scale environment of clusters
Most clusters belong to a supercluster, and these superclus-ters are characterized by their total luminosity, richness, andmorphology (E12b). To demarcate superclusters, we use themethods of E12b, who calculated the galaxy luminosity den-sity field and determined the luminosity distribution of galax-ies. Supercluster membership was determined at the smooth-ing length of 8 h − Mpc (hereafter D8), and the density D (in units of mean density, (cid:96) mean = 1 . · − h − L (cid:12) ( h − Mpc ) ) isused to separate supercluster environments from the field (Li-ivam¨agi et al. 2012). Furthermore, as determined in Einasto etal. (2007), D ≈ separates the high-density cores of super-clusters from their outskirts. We direct the reader to AppendixB in E12b and references therein for more details on thesedensity calculations. We note that a correlation exists betweenD8 and redshift for our sample clusters. However, any evolu-tion in redshift should be minimal within the redshift rangewe study, and this bias should not affect our conclusions.Supercluster morphology is determined by the fourMinkowski functionals (E12b), which are proportional toan enclosed volume, the area of the surface surrounding it,the integrated mean curvature of this surface, and its inte-grated Gaussian curvature. The first three functionals describethe overall structure of a supercluster via two shapefinders(planarity and filamentarity) and their ratio (shape param-eter). The fourth functional describes a supercluster’s in-ner structure. This methodology divides superclusters intofour morphologies, based on the Minkowski functions and vi-sual appearance: spiders, multispiders, filaments, and multi-branching filaments (Einasto et al. 2011). For simplicity, inthis work, we combine these classifications into two maintypes: spiders, which exhibit one or more high-density clumpsof clusters connected by many galaxy chains; and filaments, inwhich high-density clumps or cores are connected by a smallnumber of galaxy chains. Please see Appendix C in E12band references therein for details on these morphology calcu-tar Formation and Superclusters 3
210 205 200 195RA (deg)-20246 D e c ( deg )
190 185 180RA (deg)2468 D e c ( deg ) Figure 1.
Examples of a filament supercluster (top) and a spider supercluster(bottom). These are the richest superclusters of the Sloan Great Wall, SCl 027and SCl 019, respectively (see Einasto et al. 2014 for details). Black circlesdenote galaxies in clusters of at least 50 members, and gray circles representother galaxies in the supercluster. lations. Figure 1 shows an example of a filament supercluster(top) and a spider supercluster (bottom). Galaxies in clustersof at least 50 members are shown in black, and other galaxiesin the supercluster are shown in gray.
Principal component analysis methods
PCA has been widely used in astronomy for a number ofpurposes (see Einasto et al. 2011 for references). PCA trans-forms variables of interest to a new coordinate system whosenew variables are known as the principal components (PCs)of the data. These PCs are linear combinations of the originalparameters, and they illustrate the variable(s) along which theoriginal data has the most variance. The original data variesmost when projected along the first PC; the direction of thesecond PC indicates the direction of the second greatest vari-ance; etc. We normalize and centralize our parameters by di-viding each by its standard deviation and centering each onits mean. We use PCA to investigate how several variablesare potentially correlated with cluster SF. In particular, we fo-cus not only on the two variables discussed in this work –cluster substructure and supercluster environment – but alsoinclude total cluster halo mass via two proxies, cluster r -bandluminosity (both L obs and L tot ) and total cluster stellar mass( M ∗ ; e.g., Yang et al. 2007; Andreon 2010; Gonzalez et al.2013). RESULTS
Pairwise comparisons
In this section, we investigate how cluster f SF is relatedto both the density of the clusters’ surrounding environment,and the morphology of the superclusters in which the clus-ters reside. For convenience, all f SF values discussed can befound in Table 1, with the following columns: (1) morphol-ogy of supercluster; (2) environmental density (D8); (3) f SF ;(4) number of clusters; and (5) number of galaxies. Table 1
Supercluster EnvironmentalMorphology Density (D8) f SF N clust N gal (1) (2) (3) (4) (5)(1) All < . ± .
68 1948(2) All ≥ . ± .
38 2192(3) Filament ≤ D < . ± .
13 467(4) Filament ≥ . ± .
16 924(5) Spider ≤ D < . ± .
24 922(6) Spider ≥ . ± .
22 1268
In Figure 2, we plot f SF as a function of D8. Each bluepoint represents a cluster, and the best fit line is calculated viaa linear regression of the cluster values. The gray region rep-resents a σ error on the best fit, which is calculated by per-forming a bootstrap resampling of all clusters, recalculatingthe best fit line each time, and taking the standard deviationof the resulting slopes. The error bar represents the median ofthe standard deviations of each individual cluster’s f SF . Eachcluster’s f SF standard deviation is calculated by resamplingthe galaxies in the cluster, determining a new f SF each time,and taking the standard deviation of these f SF values.The slope of this relation, − . ± . , is negative at the99.9% confidence level with a significance of approximately . σ , indicating that a weak but significant inverse correlationexists between f SF and the density of the supercluster envi-ronment. We also calculate the average cluster f SF at lowerlarge-scale densities (D < ; row 1 of Table 1) and in high-density supercluster cores (D ≥ ; row 2). In low-densityareas, we find the f SF , . ± . , is higher than that inhigh-density cores, . ± . , a difference that is sig-nificant to 99% confidence (as determined through bootstrapresampling). These results suggest that, in general, there ex-ist higher values of f SF in clusters in low-density large-scaleenvironments than in high-density cores of superclusters. Wenote as an aside that, if we remove the highest-density clus-ter with D > from our analysis, the slope of our relationremains negative with a significance at the 99.7% confidencelevel.We test whether differences in cluster mass could be thecause of the observed correlation between SF and D8. Manystudies find decreasing SF with increasing cluster mass (e.g.,Finn et al. 2005; Homeier et al. 2005; Weinmann et al. 2006;Poggianti et al. 2006; Koyama et al. 2010), while others findno such correlation (e.g., Goto 2005; Popesso et al. 2007;Balogh & McGee 2010; Chung et al. 2011). As a proxy fortotal halo mass, C14 used the observed stellar mass of clustergalaxies, M ∗ , obtained from the MPA/JHU VAGC (Tremontiet al. 2004). We use a similar metric, but multiply M ∗ by theratio of L tot to L obs to obtain an estimated total cluster stellarmass, M tot ∗ = M ∗ × ( L tot /L obs ) .Our method to test the effect of cluster mass is as follows.In short, in each bin of D8 we weight the clusters to havethe same M tot ∗ distribution as the sample as a whole, and usethese weights to calculate measurement errors for our linearregression. This effectively removes any effect of cluster masson our f SF measurements. First, we calculate each cluster’s M tot ∗ and determine the normalized distribution of these clus-ter masses. Next, for each bin of D8, we weight the bin’s M tot ∗ values so their normalized distribution matches that ofour entire sample. Each cluster is assigned the weight of its M tot ∗ bin. Finally, we apply these weights to the galaxies inour linear regression analysis. We find that the slope of our Cohen et al. S F F r a c t i on Median std. dev. ofcluster SF fraction
Figure 2. f SF versus D8. Blue points represent individual clusters, and thegray region represents a σ error on the best fit solid line. The error bar isthe median standard deviation of each individual cluster’s f SF . In general,clusters in lower-density environments exhibit higher values of f SF . relation actually becomes slightly more negative, decreasingto − . ± . , when controlling for cluster mass. Further-more, the significance of the correlation increases slightly to . σ . This suggests that a relation between large-scale densityand cluster mass is not the cause of the observed correlationbetween SF and D8. We also perform the same weightingprocedure using r -band luminosity (both L obs and L tot ) as aproxy for halo mass, and the results remain the same. Wefurther note that, when plotting f SF as a function of clus-ter mass, we observe no correlation. Finally, we perform thesame weighting procedure using number of cluster galaxiesinstead of M tot ∗ . In this case, the significance of the correla-tion drops slightly, but still remains above σ .We now test this relationship in superclusters of spider andfilament morphology separately. Note that we only includeclusters within superclusters, i.e., with D > . Figure 3shows f SF as a function of D8 for clusters in spider (blue,right-hatched) and filament (red, left-hatched) superclusters.The hatched regions represent σ errors on the best fit lines,determined, as in Figure 2, by bootstrapping over clusters ofeach type. Interestingly, we observe the same inverse corre-lation between f SF and D8 only for filament superclusters:the slope of this relation is negative at the 99.8% confidencelevel, and f SF at lower densities, . ± . , is higher thanthat at higher densities, . ± . , with greater than 99%confidence (rows 3 and 4 of Table 1, respectively). In spidersuperclusters, there is no significant correlation between f SF and D8.We also examine f SF in clusters in spider and filament su-perclusters at high densities (rows 4 and 6 of Table 1, respec-tively). The value of f SF in spider superclusters with D > , . ± . , is higher than that in filament superclusterswith D > , . ± . , with greater than 99% confi-dence. This difference is apparent in Figure 3. In low densityoutskirts (rows 3 and 5 of Table 1), there is no difference in f SF between clusters in spider and filament superclusters.The specifics of the FoF algorithm used by Tempel et al.(2012) introduces a complication into our analysis. As theFoF algorithm builds a given cluster, the higher density ofgalaxies within superclusters makes it easier for the algorithm S F F r a c t i on SpiderFilament
Figure 3. f SF versus D8 in spider (blue, right-hatched) and filament (red,left-hatched) superclusters. Points represent individual clusters, and thehatched regions represent σ errors on the best fit lines. The inverse cor-relation between f SF and D8 is predominantly due to filament superclusters.Also, we observe higher f SF values in spider superclusters than filamentsuperclusters at high environmental densities. to include galaxies at larger cluster radii. Since galaxies onthe outskirts of clusters typically exhibit more SF, this couldartificially enhance cluster f SF values at higher superclusterdensities. We test for this possibility in two ways:1. We first measure f SF against the ratio of virial radius( r vir , derived as the projected harmonic mean radiusby Tempel et al. 2012) to L obs , serving as a proxy fora measurement of cluster radius based solely on clus-ter mass (e.g., Yang et al. 2007). Clusters with high r vir for their mass-derived radii (via L obs ) could haveenhanced values of f SF due to galaxies in outskirts in-cluded by the FoF algorithm. We find no correlationbetween these quantities.2. Second, we measure f SF against the surface density ofcluster galaxies ( L obs /πr vir ). Clusters with lower sur-face densities may be artificially expanded by the FoFalgorithm, including galaxies in outskirts with higherSF and thus exhibiting enhanced values of f SF . Again,we find no correlation between these quantities.We perform these tests not only with L obs as a proxy forcluster radius and mass, but also with L tot and M tot ∗ . Theresults of the tests suggest that any extended tails of galaxiesincluded in clusters due to the FoF algorithm are not artifi-cially enhancing cluster f SF .All of these results suggest that 1) there is a significant in-verse correlation between f SF and D8, dominated by clustersin filament superclusters; and 2) in high-density cores of su-perclusters, spider superclusters exhibit higher values of f SF than filament superclusters. Principal component analysis
As discussed in § f SF , amount of cluster substructure,density of supercluster environment, and total cluster mass.We use two measurements of amount of substructure fromC14: number of components; and the results from the ∆ test,which in this case is the negative of log( p ∆ ) . We also use twoproxies for total cluster mass: r -band luminosity (both L obs tar Formation and Superclusters 5 Table 2
Variable PC1 PC2 PC3 PC4 f SF .
173 0 . − . − . − log( p ∆ ) − .
302 0 .
664 0 . − . D8 − . − . − . − . L obs [10 h − L (cid:12) ] − .
675 0 . − .
139 0 . Std. dev. .
403 1 .
161 0 .
714 0 . Prop. of var. .
492 0 .
337 0 .
127 0 . Cum. prop. .
492 0 .
830 0 .
957 1 . and L tot ) and M tot ∗ .Our PCA results are consistent whether we use L obs , L tot ,or M tot ∗ as a proxy for total cluster mass. Furthermore, ourresults remain the same whether we use − log( p ∆ ) or numberof components as a measure of amount of substructure. Thus,we present and discuss only the results when using L obs and − log( p ∆ ) . Table 2 displays the results of our analysis. Itshows the values of the four PCs for our four variables; andthe standard deviation, proportion of variance, and cumulativeproportion for these PCs.The cumulative proportion shows that the first two PCs ac-count for 83% of the variance in these cluster properties, witheach PC being equally important. Thus, we will focus pri-marily on the first two PCs. The PC1 values of D8 and L obs are close in magnitude and of the same sign, confirming thecorrelation between these two variables. Since the value of f SF is of opposite sign, this suggests that f SF is weakly anti-correlated with D8 and L obs . Furthermore, the PC2 value ofD8 – also of opposite sign to f SF – is approximately fourtimes larger than that of L obs . This suggests that D8 ratherthan L obs is more strongly related to f SF . This agrees withour analysis in § M tot ∗ (i.e., cluster r -band luminosity) are not the cause of theobserved correlation between f SF and D8.The PC2 values of f SF and amount of substructure are ofsimilar magnitude and of the same sign, confirming the directcorrelation between these variables found in C14. While thePC1 values of these variables are of opposite sign, the corre-lation suggested by the PC2 values is more robust: the valuesof these variables along PC2 are closer in magnitude to eachother than those along PC1; and the D8 and L obs values alongPC2 are much lower than those along PC1. These values,compared to the others discussed, suggest that f SF is proba-bly most strongly related to amount of substructure than theother variables discussed.Finally, the PC1 values of D8 and amount of substructuresuggest a direct correlation between these variables, whichagrees with the findings in E12a (though they admit thatthis correlation is weak). Furthermore, PC1 also shows thatamount of substructure is also correlated with L obs . Intu-itively, this is expected: a richer cluster will have a higherluminosity and more opportunity for substructure to be de-tectable. This effect does act counter to the main result of thispaper – the inverse correlation between D8 and f SF – and theresult of C14 – the direct correlation between substructure and f SF (see Figures 5 and 6 in that paper). In other words, as D8and luminosity increase, the results from this paper and C14suggest that f SF (and thus amount of substructure) shoulddecrease, not increase. This, however, bolsters the correlationwe find between D8 and f SF – it must be significant enoughto counter the weak correlation between D8 and amount ofsubstructure. DISCUSSION
We find a significant inverse correlation between the den-sity of supercluster environment and the amount of SF withingalaxy clusters. While this could in principle be an indirectresult of a correlation between supercluster density and clus-ter substructure, we find this not to be the case. Rather, bothcluster substructure and supercluster environment are inde-pendently related to a cluster’s SF, and, while these effectsoppose each other, the influence of cluster substructure ap-pears stronger than that of supercluster environment. Theseresults are not simply due to the correlation between D8 andcluster mass, luminosity, or richness.We also find that supercluster morphology is important inaffecting cluster SF: the relation between supercluster densityand SF is observed only in filament rather than spider super-clusters. Furthermore, SF in spider superclusters is higherat high densities compared to filament superclusters. Whenwe consider these differences between filament and spider su-perclusters, from the complexity of effects explained aboveemerges a coherent picture. Spider superclusters have richerinner structure, and are dynamically younger, than filamentsuperclusters (e.g., E12b). We expect to find more SF in dy-namically younger systems, and we indeed see this in thehigh-density cores of spider superclusters.In galaxy clusters, more structure indicates a less relaxed,younger system (e.g., Bird & Beers 1993; Knebe & M¨uller2000; C14; Cohen et al. 2015). Such clusters are more likelyto live in superclusters with richer inner structure where groupmergers occur more easily than in superclusters with simpleinner structure. Thus, combining our results from this workwith those of C14, we can explain in more detail the effects ofcluster substructure and supercluster environment on SF. Asclusters form hierarchically from smaller groups, the dynam-ically younger systems exhibit more SF, since the SF in thesesystems has had less time to be quenched by various gravita-tional and hydrodynamical processes (see Boselli & Gavazzi2006 for a review of such mechanisms). Thus, it is more likelyto find high SF in clusters that find themselves in the high-density environments of spider superclusters than filament su-perclusters. Additionally, this shows that high-density coresof superclusters are a special environment for clusters. For in-stance, they may be collapsing (Einasto et al. 2015; Gramannet al. 2015; Einasto et al. 2016), possibly affecting the prop-erties of galaxy clusters and their galaxy populations. Thisinteresting result of our study emphasizes the role of super-cluster morphology in shaping the properties of galaxies andgroups/clusters in them.The main result of this work agrees with Lietzen et al.(2012), who found that more elliptical galaxies are foundin the higher-density environments of superclusters than atlower densities. Furthermore, Luparello et al. (2013) usedthe galaxy spectra parameter D n to show that galaxies ingroups in superclusters are systematically older than those inlower-density environments. They found that this result holdseven though the groups themselves have higher velocity dis-persions and are therefore dynamically younger than groupselsewhere. These results agree well with the interpretationfrom this work explained above.We note that Costa-Duarte et al. (2013) found no correla-tion between the mean stellar ages of superclusters and theshape parameter of superclusters. This is not in conflict withour results, since we used information about the inner struc-ture of superclusters to divide them into two morphological Cohen et al.classes, while Costa-Duarte et al. (2013) only used the shapeparameter to characterize the outer shape of superclusters.Einasto et al. (2014) showed, in agreement with Costa-Duarteet al. (2013), that galaxy content of superclusters dependsonly weakly on the overall shape of superclusters.Our results, while significant, still exhibit substantial scat-ter. This owes to the complicated dynamics affecting cluster f SF , many aspects of which are discussed here. One variablewe have not taken into account is the stage of formation of acluster or supercluster. Studies have shown that clusters withsimilar degrees of apparent substructure can exhibit different f SF