THE ENVIRONMENTS OF STARBURST AND POST-STARBURST GALAXIES AT z = 0.4–0.8 ^*

"Post-starburst" galaxies, also known as "E+A" or "k+a" galaxies¹⁸, are identified on the basis of their optical spectra for the exceptionally strong Balmer lines in absorption and the lack of emission lines (Dressler & Gunn 1983; Couch & Sharples 1987).

Extensive spectrophotometric modeling has demonstrated that this spectral combination arises if the star formation (SF) activity has ended abruptly sometime between ∼5 × 10⁷ and ∼1.5 × 10⁹ years prior to observations (Couch & Sharples 1987; Newberry et al. 1990; Abraham et al. 1996; Barger et al. 1996; Poggianti & Barbaro 1996, 1997; Bekki et al. 2001; Dressler et al. 2009; see Poggianti 2004 for a modeling review). Only those spectra with the strongest Balmer lines require a recent burst, while k+a spectra with moderate line strengths can be modeled as post-star-forming galaxies in which the truncation follows a regular SF activity. In the latter cases, the commonly adopted naming of "post-starburst" is in fact misleading. The exact duration, time elapsed, and strength of the most recent SF episode remains hard to constrain, in the absence of high signal-to-noise ratio (S/N) and high resolution spectra (Leonardi & Rose 1996). Keeping this in mind, in the following we will refer interchangeably to "k+a" or "post-starburst" spectra for convenience.

First discovered in distant galaxy clusters over 25 years ago (Dressler & Gunn 1983), k+a galaxies have seen a renewed interest in recent years for a number of reasons. Being galaxies in rapid transition from a blue, star-forming phase to a red, passively evolving phase, they were soon recognized to play an important role in the evolution of the star-forming fraction in galaxy clusters (the so-called Butcher–Oemler effect; Butcher & Oemler 1984), related to the transformation of spirals in distant clusters into the numerous S0 galaxies that dominate clusters today (Poggianti et al. 1999; Tran et al. 2003). More recently, it has been realized that the presence of k+a galaxies in environments other than clusters can contribute to the evolution of the red and blue galaxy fractions in any environment (Norton et al. 2001; Tran et al. 2004; Yan et al. 2008).¹⁹ At redshifts above 1, k+a spectra may be a common feature of massive galaxies that formed most of their stars at high redshift (Doherty et al. 2005; van Dokkum & Ellis 2003; van Dokkum & Stanford 2003).

Post-starburst spectra are observed both in clusters and in the general field at all redshifts, but their incidence varies with redshift, environment, and galaxy mass.

At intermediate redshifts (z = 0.3–0.6), several spectroscopic surveys of galaxy clusters have reported large fractions of k+a spectra reaching up to 1/4 of the entire galaxy population in distant massive clusters (Dressler & Gunn 1983; Couch & Sharples 1987; Henry & Lavery 1987; Fabricant et al. 1991; Dressler & Gunn 1992; Barger et al. 1996; Dressler et al. 1999; Tran et al. 2003, 2004, 2007; Swinbank et al. 2007). As is obvious, the computed k+a fraction depends on the adopted criteria for Balmer and O ii strengths, line measurement method, and spectral quality. As a consequence, more significant than the absolute k+a fraction, is the relative fraction in different environments within the same study. In general, these studies have found that the frequency of k+a galaxies is significantly higher in clusters than in the field at intermediate redshifts (e.g., Dressler et al. 1999; Tran et al. 2004), although this is not a universal conclusion: for example, Balogh et al. (1999) argued for a similarly low k+a proportion in clusters and in the field at z ∼ 0.2–0.3. Low post-starburst fractions are found both in the field and in groups at z ∼ 0.8 from Yan et al. (2008), who argued for a 2.5σ higher post-starburst fraction in the field than in groups. A cluster-related origin for a significant population of k+a's at intermediate redshifts is also supported by the analysis of the spatial distribution of galaxies with k+a spectra in a supercluster region at z ∼ 0.5 by Ma et al. (2008), who found that k+a's occur within the cluster ram pressure stripping radius and are essentially absent in the supercluster filamentary structure.

In the local universe, k+a spectra are rare among bright galaxies in all environments. As a result, since clusters are rare objects, the majority of k+a galaxies at low-z, in number, reside in the field (Zabludoff et al. 1996; Blake et al. 2004; Hogg et al. 2006; Goto 2007). Zabludoff et al. (1996) were the first to note the rather common association of this type of spectra in the low-z field with signs of galaxy–galaxy interactions and mergers. Other works have confirmed these findings, suggesting that the majority of field k+a's at low-z result from mergers that evolve into early-type galaxies (Yang et al. 2004, 2008; Goto 2005; Nolan et al. 2007). Their high merger rate is consistent with the high incidence of spectroscopically confirmed companions (Yamauchi et al. 2008) and with the fact that they statistically prefer high local density regions at the scale of <100 kpc (Goto 2005).

In nearby clusters, "Balmer-enhanced" spectra were identified in several galaxies of the Coma cluster (Caldwell et al. 1993). However, in the majority of these cases, the Balmer lines were far weaker than those in distant clusters and Poggianti et al. (2004) concluded that luminous galaxies with post-starburst/post-starforming spectra are infrequent in Coma. The latter study revealed a numerous population of fainter k+a galaxies (M_B > − 17), comprising ∼15% of the cluster dwarfs. If Coma (the only local cluster studied in detail to date) is representative of massive clusters locally, these results would imply that the evolution of the k+a luminosity distribution proceeds in a downsizing fashion (Poggianti et al. 2004). The same mechanism may still give rise to cluster k+a's both at high- and low-z, if the maximum luminosity/mass of actively star-forming galaxies infalling onto clusters decreases at lower redshift. A striking correlation between the position of the youngest k+a galaxies and the substructure in the Coma hot intracluster medium (ICM) strongly suggests that the quenching of SF is due to the interaction of galaxies in infalling groups with the shock fronts in the ICM.

Several works based on the Sloan Digital Sky Survey (SDSS) have used the local galaxy density to investigate the environment of k+a galaxies at low-z (Balogh et al. 2005; Goto 2005; Hogg et al. 2006; Nolan et al. 2007; Yan et al. 2008). They have unanimously found that low-z k+a's do not preferentially reside in high-density regions, and inferred from this that they do not preferentially belong to low-z clusters. However, as we will show, a low- and intermediate-density preference of k+a's does not necessarily imply residing outside of clusters. Again from SDSS, Yan et al. (2008) found a possibly double peaked density distribution of post-starburst galaxies, the strongest peak at the lowest densities sampled and a second peak at densities above average, suggesting two possible formation channels. Still at low-z, Hogg et al. (2006) found only a small k+a excess inside the virial radii of clusters, and a slight excess with near neighbors, supporting the hypothesis that some k+a spectra are due to interactions with the ICM.

Overall, although a detailed study of a large sample of nearby clusters is still lacking, the conclusion emerging from all low-z studies is that there is a paucity of bright, massive k+a galaxies in all environments, and a suggestive abundance of k+a faint galaxies in the one studied cluster. In contrast, at intermediate redshifts, most studies have measured significant fractions of bright k+a galaxies in clusters, and a higher fraction in clusters than in the field.

It is fundamental to realize that several different physical processes can produce a k+a spectrum. Any time SF is truncated on a short timescale, a galaxy will experience a k+a phase. In different environments, different mechanisms may be, and very probably are, responsible for truncating SF. The results obtained to date suggest that mergers may be the predominant mechanism in the low-z field, while the environmental dependence of k+a's at higher redshifts and the results in Coma strongly point to cluster-related phenomena. Other effects have been proposed, such as active galactic nuclei (AGNs) and supernovae quenching (e.g., Bekki et al. 2001; Kaviraj et al. 2007) and massive dark matter halo formation (Birnboim et al. 2007). In clusters, besides the aforementioned impact of infalling galaxies with the ICM shock fronts (Poggianti et al. 2004; Crowl & Kenney 2006), other quenching mechanisms include ram pressure stripping (Gunn & Gott 1972; Ma et al. 2008; Vollmer 2008), galaxy harassment (Moore et al. 1996, 1999), and tidal interactions (Koopmann & Kenney 2004; Oemler et al. 2009; see also Treu et al. 2003; Boselli & Gavazzi 2006 for reviews of cluster processes).²⁰

k+a's are not the only Balmer-strong type of spectra. Emission-line spectra with exceptionally strong Balmer lines in absorption (e(a) spectra) were identified in large numbers at intermediate redshifts, both in clusters and in the field (Dressler et al. 1999; Poggianti et al. 1999). At low-z, this type of spectra are associated with ongoing starbursts whose young stellar populations are strongly obscured by dust. In fact, they are rare among local normal spirals, but common among dusty starbursts and luminous infrared galaxies (Liu & Kennicutt 1995; Poggianti & Wu 2000).²¹ An e(a) spectrum can be produced if the O and B stars causing the emission lines are highly obscured by dust and the older A-type star—responsible for the strong Balmer absorption—have had time to free themselves or drift away from their dusty cocoons (Poggianti et al. 1999). On average, younger stellar generations are more strongly obscured than older ones, and this phenomenon is often referred to as "selective extinction." As a result of this effect, the strongest starbursts locally have only weak to moderate emission lines (due to high extinction), and have usually e(a) spectra, not spectra with the strongest emission lines.

Spectrophotometric modeling of e(a) galaxies has confirmed both the adequacy and the necessity of selective extinction and of an ongoing starburst to explain simultaneously the optical spectral characteristics and the FIR fluxes of local e(a) galaxies (Poggianti et al. 2001; Shioya et al. 2001; Bekki et al. 2001). Mid-IR ISOCAM (Duc et al. 2002) and Spitzer (Dressler et al. 2009) studies have confirmed that statistically e(a) galaxies at z ⩾ 0.2 have the highest SFRs, and that they are experiencing an increase in SFR compared to their past average (Dressler et al. 2009).

From the optical point of view, the only difference between k+a and e(a) spectra is the presence of emission lines. If the scenario outlined above is correct, however, the two spectral classes correspond to fundamentally different phases of SF activity: post-starburst the k+a's, and ongoing dusty starburst the e(a)'s. An evolutionary link between the two types of galaxies can occur, with e(a)'s being in some cases the progenitors of some of the k+a galaxies (Poggianti et al. 1999; Balogh et al. 2005).

The association between spectral class and SF history in these two classes works well in a statistical sense, but the demarcation between the two histories is not straightforward in all galaxies and several caveats need to be kept in mind. Both radio (Smail et al. 1999) and mid-IR (Dressler et al. 2009) observations have found that some k+a galaxies may host a residual ongoing SF activity, highly obscured by dust and therefore not visible in the O ii line. These works show that k+a galaxies with ongoing SF are a minority and that the residual SFR is significantly lower than the SFR during the previous burst (Dressler et al. 2009). Moreover, the separation between k+a's and e(a)'s is commonly set to a low EW of emission lines (usually 5 Å  in O ii) that is determined by the standard detection limit in distant spectroscopic surveys. This limit is necessarily arbitrary: lower or upper detection limits can result in a different mixing between the two classes. The lines used in the classification and the adopted limits for their strength vary significantly in the literature and this obviously hinders the comparison between different works and with the existing modeling. Any adopted choice of line combination leads to some level of bias, incompleteness and possible misclassification between k+a's and e(a)'s. For example, the use of the Hβ line in emission has been strongly advocated as a substitute of O ii by Yan et al. (2006) to avoid incompleteness in the post-starburst sample due to AGN-dropouts. In terms of the Hβ-based classification, it is still unclear how post-starbursts can be separated from ongoing dusty starbursts. Furthermore, e(a) spectra are sometimes referred to as "post-starbursts" in the literature (e.g., Tremonti et al. 2007; see also Le Borgne et al. 2006), because the association between strong Balmer lines and a post-starburst history is common knowledge in the astronomical community, unlike the association between Balmer lines and ongoing starbursts. Of course it remains possible that some e(a)'s are truly post-starburst systems whose lines in emission are due to a small residual SF activity (Poggianti et al. 1999). Finally, spectrophotometric modeling with selective extinction shows that regular SF histories can account for e(a) spectra at z > 0.4, depending on the timescale of the SF decline, without the need to invoke a starburst (J. Fritz & B. M. Poggianti 2009, in preparation).

Regardless of the intrinsic caveats in the spectral interpretation on a galaxy-by-galaxy basis, the simultaneous study of emission and Balmer features is the primary source of information to unveil the recent SF history of galaxies observed by spectroscopic surveys: identifying k+a and e(a) spectra allows to recognize strong variations of the SF activity (a truncation or a burst), and to study how they change with z and environment. In this paper, we investigate the occurrence and properties of post-starburst galaxies and dusty starburst candidates at z = 0.4–0.8. We make use of the ESO Distant Cluster Survey data set (Section 2), adopt the spectral classification described in Section 3, and consider different environments: clusters, groups, and the field, as defined in Section 4. We show how the proportion of post-starburst galaxies and dusty starburst candidates depends on environment (Section 5) and how it varies in clusters with the cluster properties (Section 5.2). We next consider composite spectra as an alternative tool to study the environmental dependence of SF histories (Section 6). We present the detailed properties of galaxies of different spectral types: their morphologies (Section 7.1), masses, luminosities, and mass-to-light ratios (Section 7.2), local density, and radial distributions (Section 7.3). Finally, in Section 8 we summarize our results and discuss the implications for the origin and the evolution of post-starburst galaxies. In the Appendix, we search for spectroscopic evidence of galactic winds in our Balmer-strong population.

In the following, we adopt the convention that the EW(O ii3727) is positive in emission while the EW(Hδ4101) is positive in absorption. All equivalent widths and cluster velocity dispersions are given in the rest frame. We assume a Λ cold dark matter (ΛCDM) cosmology with (H₀, Ω_m, Ω_λ) = (70, 0.3, 0.7).

The ESO Distant Cluster Survey (hereafter, EDisCS) is a multiwavelength survey of galaxies in 20 fields containing galaxy clusters at z = 0.4–1.

Candidate clusters were selected as surface brightness peaks in smoothed images taken with a very wide optical filter (∼ 4500–7500 Å) as part of the Las Campanas Distant Cluster Survey (LCDCS; Gonzalez et al. 2001). The 20 EDisCS fields were chosen among the 30 highest surface brightness candidates in the LCDCS, after confirmation of the presence of an apparent cluster and of a possible red sequence with VLT 20 minutes exposures in two filters (White et al. 2005).

For all 20 fields, EDisCS has obtained deep optical photometry with FORS2/VLT (White et al. 2005), near-IR photometry with SOFI/NTT (A. Aragón-Salamanca et al. 2009, in preparation), multislit spectroscopy with FORS2/VLT (Halliday et al. 2004; Milvang-Jensen et al. 2008), and MPG/ESO 2.2/WFI wide field imaging in VRI. ACS/HST mosaic imaging in F814W of 10 of the highest redshift clusters has also been acquired (Desai et al. 2007). Other follow-up programmes include XMM-Newton X-Ray observations (Johnson et al. 2006), Spitzer IRAC and MIPS imaging (programs PIs: Desai, Finn, Rudnick), Hα narrowband imaging (Finn et al. 2005), additional FORS2 imaging and spectroscopy in 10 EDisCS fields (Douglas et al. 2007), 2dF/AAT, VIMOS/VLT and IMACS Magellan wide-field spectroscopy.

Spectroscopic targets were selected from I-band catalogs (Halliday et al. 2004). Conservative rejection criteria based on photometric redshifts (Pelló et al. 2008) were used in the selection of spectroscopic targets to reject a significant fraction of nonmembers, while retaining a spectroscopic sample of cluster galaxies equivalent to a purely I-band selected one. A posteriori, we verified that these criteria have excluded at most 1%–3% of cluster galaxies (Halliday et al. 2004; Milvang-Jensen et al. 2008). The spectroscopic selection, observations, and catalogs were presented in Halliday et al. (2004) and Milvang-Jensen et al. (2008). Out of the 20 EDisCS fields, we exclude here the two fields that lack several masks of deep spectroscopy (cl1122.9-1136 and cl1238.5-1144; Halliday et al. 2004). For our analysis, we use all the structures in the remaining fields within 0.1 in z from the redshift targeted for spectroscopy in each field. The structures used in this paper are listed in Table 1.

Typically, FORS2/VLT spectra of greater than 100 galaxies per cluster field were obtained, with exposure times of 4 hr for the high-z sample and 2 hr for the mid-z sample. The analysis presented in this paper is based on these data.

Given the long exposure times, the success rate of our spectroscopy (number of redshift/number of spectra taken) is 97% above the magnitude limit used in this study (see Poggianti et al. 2006 for details). The completeness of our spectroscopic catalogs, which depends on galaxy magnitude and distance from the cluster center, was computed for each cluster in Poggianti et al. (2006). This is used in this paper to weight the contribution of each galaxy and therefore correct the sample for spectroscopic incompleteness. Typically, our spectroscopy samples a region out to a clustercentric radius equal to R₂₀₀. This is defined to be the radius delimiting a sphere with interior mean density 200 times the critical density and is commonly used as an approximation for the cluster virial radius. The R₂₀₀ values for our structures are computed from the velocity dispersions using Equation (1) in Poggianti et al. (2006).

The EDisCS spectra have a dispersion of 1.32 Å  pixel⁻¹ or 1.66 Å  pixel⁻¹, depending on the observing run. They have a FWHM resolution of ∼6 Å, corresponding to rest-frame 3.3 Å  at z = 0.8 and 4.3 Å  at z = 0.4. The equivalent widths of and Hδ were measured from the spectra using a Gaussian line-fitting technique, as outlined in Poggianti et al. (2006), using the same technique of Dressler et al. (1999). Our measurements are based on the visual inspection of each one-dimensional and two-dimensional spectrum. Each emission line detected in a given one-dimensional spectrum was confirmed by visual inspection of the corresponding two-dimensional spectrum. This method is especially useful for assessing the reality of weak lines, to avoid false positive and false negative detections of O ii emitters. Error bars on EWs were estimated allowing continuum variations according to error spectra computed as in P. Sanchez-Blazquez et al. (2009, in preparation).²²

We do not attempt to separate a possible AGN contribution to the O ii line, or to exclude galaxies hosting an AGN. We are unable to identify AGNs in our data, since the traditional optical diagnostics are based on emission lines that are not included in the spectral range covered by most of our spectra (see Poggianti et al. 2009; P. Sanchez-Blazquez et al. 2009, in preparation, for a discussion of AGN occurrence in our red sample).

Only galaxies with an absolute V magnitude brighter than M^lim_V are considered in the following. M^lim_V varies with redshift between −20.5 at z = 0.8 and −20.1 at z = 0.4 to account for passive evolution. In the following, we will refer to this as the magnitude-limited sample.

In the following, we adopt a spectral classification similar to that proposed by the MORPHS collaboration (Dressler et al. 1999). This method of classification is based on the equivalent widths of the O ii line in emission and the Hδ line in absorption, which are the best indicators of ongoing and recent SF, respectively, in optical spectra of galaxies at z = 0.4–1.

The classification allows us to broadly divide galaxies into groups according to their SF history. In particular, it aims to identify galaxies with an ongoing (starburst) and a recent (post-starburst) episode of SF, to distinguish them from passively evolving galaxies and quiescently star-forming galaxies. This scheme is based upon and is a development of earlier spectral interpretation of strong Balmer absorption in the spectra of galaxies in intermediate redshift clusters (Dressler & Gunn 1983; Couch & Sharples 1987).

The classes that we will use in the following are:

• 1.  
  Passively evolving galaxies—k class showing no spectral sign of ongoing or recent SF during the past 1–1.5 Gyr. We define them as those with no securely detected O ii line in emission and EW(Hδ) < 3 Å  and refer to them as "k" galaxies (with a spectrum similar to a K-type star). In the following, we will refer occasionally to "k(e)" galaxies, which correspond to those with an otherwise k-type spectrum and an additional very weak O ii line in emission (less than 5 Å), whose reality is confirmed by the two-dimensional spectrum.
• 2.  
  Post-starforming/post-starburst galaxies—k+a class. Spectrophotometric modeling has shown that the lack of significant O ii emission (EW<5 Å) and the strength of Hδ (EW>3 Å) imply that the SF activity was terminated sometime between 5 × 10⁷ and 1.5 × 10⁹ years prior to observation. We refer to these as "k+a" (a mix of K-type and A-type stars) galaxies.
• 3.  
  Quiescent star-forming galaxies—e(c) class, which have spectra with a moderate O ii line in emission (EW = 5–25 Å) and an EW(Hδ) < 4 Å. These spectral characteristics are similar to those of normal spiral galaxies of types Sa to Sd in the local universe. They are consistent with a continuous SF history during which a galaxy experienced no sudden variation (either starburst or SF-truncation) in its SF activity.
• 4.  
  Dusty starburst candidates—e(a) and e(a)+ classes. E(a)'s are defined to have an EW(O ii)>5 Å  and a strong Hδ line in absorption (EW(Hδ)>4 Å). As discussed in Section 1, at low redshift these line strengths are observed to be rare among normal spirals, while they are common among dusty starbursts and luminous infrared galaxies. These observations and spectrophotometric modeling of e(a) galaxies suggest that this spectral combination arises in dusty, ongoing starbursts. As discussed in Section 1, in principle the e(a) signature can arise from truly post-starburst systems with a residual SF activity, although examples of such a case have not yet been securely identified observationally in galaxy clusters.Hδ-strong galaxies such as e(a)'s typically also have strong higher-order Balmer lines, i.e., strong H (epsilon, 3970 Å), Hζ (zeta, 3889 Å), Hη (eta, 3835 Å), and Hθ (theta, 3798 Å), as shown in the bottom right panel of Figure 1.²³ The line that correlates most strongly with Hδ and is the best indicator of recent SF among the high-order Balmer lines is Hζ (zeta), that is a useful substitute for Hδ whenever the latter is unavailable.In EDisCS spectra, we noted a large number of galaxies that have EW(Hδ) < 4 Å  and therefore strictly speaking are not classifiable as e(a)'s, but have very strong higher-order Balmer lines similar to those of e(a)'s. An operational definition to categorize this type of spectra is to select those with Hζ>4 Å. In the following, we will refer to these as "e(a)+" spectra. Their similarity to e(a)'s is visible comparing the two bottom panels in Figure 1.E(a)+ spectra only differ from e(a)'s because the Hδ line in absorption is slightly more filled in by emission. In the following, we consider them separately from both e(a)'s and e(c)'s, to allow for the possibility that at least some e(a)+ galaxies may be candidate dusty starbursts too.
• 5.  
  Strong emission line starbursts—e(b) class, defined to have an EW(O ii)>25 Å. Both in the local universe and in clusters at z ∼ 0.5, galaxies with a strong O ii line are generally low-mass, very late-type galaxies (Poggianti et al. 1999). In Section 7, we show that this is the case also in the EDisCS sample. These spectra belong to "starbursts" in the sense that their current SFR is much higher than their past average, but they can be either starbursts or simply galaxies with a regular SFR of a late-type and low extinction levels.Following Dressler et al. (1999), we present in Figure 2 a summary view of the equivalent width limits adopted to define the spectral types described above.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.1. Robustness of the Spectral Type Interpretation

The EDisCS data set allows us to study galaxies in a wide range of environments using homogeneous data. In this paper we consider the following.

• 1.  
  Clusters are defined as structures with a velocity dispersion σ>400 km s⁻¹. Our clusters, listed in Table 1, span the entire range of cluster masses with velocity dispersion from 400 to 1100 km s⁻¹.
• 2.  
  Groups are structures with 160 < σ < 400 km s⁻¹ with at least 8 spectroscopic members. Groups are further split into groups with high- and low-O ii content, depending on their fraction of O ii-emitters. High/low-O ii groups have an O ii fraction higher than 80% and lower than 50%, respectively.The high-O ii groups and the clusters define a broad anticorrelation in a diagram of O ii fraction versus cluster velocity dispersion σ (f_O ii = −0.74(σ/1000) + 1.115). Low-O ii groups are "outliers" in this relation (Figure 4 in Poggianti et al. 2006). In the following, we will refer collectively to "O ii-outliers" as those groups and clusters that have a low O ii content for their σ compared to the rest of the structures. Specifically, the outliers have an O ii fraction lower by at least 25% than the fraction derived from the best-fit relation at its velocity dispersion, thus f_O ii < −0.25 + (−0.74(σ/1000) + 1.115) (see Figure 10 and Equation (2) in Poggianti et al. 2006).
• 3.  
  Poor groups are galaxy associations between 3 and 6 galaxies, identified as described. No velocity dispersion measurement is attempted for them.
• 4.  
  The "field" includes those galaxies that do not belong to any cluster, group, or poor group.

Only groups, poor groups, and field galaxies within ±0.1 in redshift from the cluster targeted in each field were considered. With this choice, the spectroscopic catalog of each environment is equivalent to a purely I-band selected sample with no selection bias, although it should be kept in mind that there may be a preference for our field galaxies to lie in the filaments close to our clusters and groups.

We begin our analysis by examining the relative fractions of each spectral type in the different environments. The raw fractions, measured for the whole magnitude-limited spectroscopic sample with no radial constraints, are given in Table 3. Table 4 lists the fractions for the magnitude-limited sample weighted for spectroscopic incompleteness and considering only galaxies within R₂₀₀. The results of Table 4 are visually illustrated in Figure 3.

Zoom In Zoom Out Reset image size

Tables 3 and 4 refer to a magnitude-limited sample of galaxies. We also define a mass-limited sample including, in all environments, only galaxies with stellar masses >5.0 × 10¹⁰ M_☉. This is the stellar mass of a galaxy with luminosity M_V = −20.3 (our magnitude limit at the average redshift) that has the highest mass-to-light ratio we observe in our sample, M/L_V = 4 M_☉/L_☉ (Section 7.2). Once corrected for spectroscopic incompleteness, our sample is complete above this mass limit for all spectral types and all redshifts. Galaxy stellar masses and absolute luminosities were computed as described in Poggianti et al. (2008). The fractions of each spectral type for the mass-limited sample are presented in Table 5.

As far as the post-starburst galaxies are concerned, the tables and Figure 3 show that the proportion of k+a galaxies is comparably high in clusters and in low-O ii groups (about 10%), while the poor groups and the field contain fewer post-starburst galaxies, and no k+a galaxy is detected in the groups with high O ii.

This has two main implications. First of all, these results highlight, once more, the differences in the galaxy populations of groups of similar velocity dispersion, but different O ii content. We have noted before that the properties of galaxies in low-O ii groups resemble those in the cores of far more massive clusters, in terms of their fractions of passively evolving and early-type galaxies, their O ii EW distribution (Poggianti et al. 2006), and their steep relation between average star-forming fraction and local galaxy density (Poggianti et al. 2008). We now find that these systems are similar to more massive clusters also for the conspicuousness of their post-starburst population. In contrast, high-O ii groups of similar masses appear to be the least-favorable environment for the production of post-starburst galaxies.

Second, post-starburst galaxies are observed for the first time to occur in significant proportions also in a subset of the groups, while so far, at z > 0.4, they were believed to be a phenomenon related principally to massive clusters. As discussed in Section 1, previous work found a low k+a fraction in the general field and a high k+a fraction in clusters at z = 0.4–0.5. Intermediate and low-mass clusters and groups at these redshifts had not been studied in detail to date. Hints of a sizeable post-starburst population in low-z groups were found by Zabludoff & Mulchaey (1998). Our result shows that a mechanism switching off SF on a short timescale must operate efficiently also in some (but not all) groups.

We note that the k+a fraction computed over all groups (low- and high-O ii groups together) is low, of the order of that computed in the field (see line "All groups" in the tables). This might explain the recent result of Yan et al. (2008), reporting low k+a fractions in the DEEP2 field and groups at z = 0.8 and a 2.5σ higher fraction in the field than in groups. Given the different environment and post-starburst definitions, a straight comparison cannot be performed, but their group sample is likely dominated by a combination of our poor groups, high-O ii groups and low-O ii groups: our poor group and overall group k+a fractions (e.g., 0.03 and 0.05, respectively, see Table 4) and field fractions (0.06) are probably qualitatively consistent with their result.

It is also of note that the k+a fraction we observe in clusters is significantly lower than that measured by the MORPHS collaboration, the only previous large spectroscopic survey of several distant clusters at redshifts similar to ours (Dressler et al. 1999). The MORPHS work has the same EW measurement method and spectral classification, and similar spectral quality, galaxy magnitude, and radial limits of the present work, therefore a direct comparison is appropriate. In 10 clusters at z = 0.4–0.5, the MORPHS reported a total k+a fraction of ∼20% (Poggianti et al. 1999), which is approximately twice the value we observe in EDisCS. In the following, the analysis of the detailed cluster properties will shed light on the origin of the different k+a incidence in the two samples (Section 5.2).

The conclusions discussed above are valid both for the magnitude-limited and mass-limited samples. Figure 3 and Tables 3–5 also show that the environments with the highest k+a fraction (i.e., clusters and low-O ii groups) are those with the highest fraction of passively evolving (k) galaxies, and lowest fractions of actively star-forming (emission-line) galaxies. At z = 0.4–0.8, those environments that host few currently star-forming galaxies are also efficiently truncating SF in them.

5.1. Quenching Efficiency and Starburst Frequency

Each galaxy in our sample is observed in one of three main evolutionary phases: while it is passively evolving (k/k(e)), recently star-forming (k+a) or currently star-forming (any of the emission-line classes). It is useful to examine the incidence of post-starburst galaxies within the "active" population, defined to include k+a galaxies and galaxies of all emission-line types. The active population represents the population of galaxies that were all star-forming ∼2 Gyr before the epoch of observation.

Hence, the k+a/active fraction can be considered a sort of "quenching efficiency": the efficiency of a given environment in truncating SF in star-forming galaxies, which have presumably been recently accreted by the cluster.²⁴

The environmental variations of the SF histories are even more striking when considering the spectral fractions relative to the active population instead of the fractions in the entire cluster population. The k+a/active fractions are listed in Table 6 and shown in Figure 4 for the various EDisCS environments. Post-starburst galaxies represent 20%–30% of the active galaxy populations in EDisCS clusters and low-O ii groups, while they compose only less than 10% of the active galaxies in the other environments.

Zoom In Zoom Out Reset image size

Table 6. Fractions in the Mass-selected EDisCS Sample Within R₂₀₀ Weighted for Spectroscopic Incompleteness

Galaxy Type	k+a/Active	e(a)/Active	e(a)/Emission
Clusters	0.23 ± 0.04	0.23 ± 0.04	0.33 ± 0.05
All groups	0.09 ± 0.05	0.39 ± 0.10	0.46 ± 0.12
Groups low-O ii	0.30 ± 0.13	0.31 ± 0.13	0.44 ± 0.18
Groups high-O ii	0.00 ± 0.02	0.43 ± 0.14	0.46 ± 0.15
Poor groups	0.06 ± 0.05	0.28 ± 0.09	0.32 ± 0.11
Field	0.09 ± 0.06	0.11 ± 0.09	0.18 ± 0.12
MORPHS clusters	0.40 ± 0.02	0.21 ± 0.02	0.37 ± 0.06
MORPHS "field"	0.09 ± 0.03	0.15 ± 0.04	0.20 ± 0.07

Notes. The "active" class includes: k+a, e(a), e(a)+, e(c), e(b), e. "Emission" includes e(a), e(a)+, e(c) and e(b). The results for the magnitude-limited sample are similar, always within the error bar. The term "All groups" includes both low- and high-O ii groups, but not the poor groups. Also shown are results from the MORPHS collaboration, for their approximately magnitude-limited sample.

Download table as:  ASCIITypeset image

The same table shows also a comparison with the MORPHS cluster and field results at z = 0.4–0.5. The k+a/active incidence is even higher in MORPHS than in EDisCS clusters (40% versus 23%), and it is equal in the EDisCS and MORPHS field (9%).

Considering now the dusty starburst candidates, Table 6 lists both the e(a)/active fraction and the fraction of e(a)'s relative to the number of emission-line galaxies with an assigned spectral type (the sum of e(a)'s, e(a)+'s, e(c)'s, and e(b)'s). The e(a)/emission fraction, graphically displayed in Figure 4, is related to the occurrence of dusty starburst candidates among star-forming galaxies, whose SF has not yet terminated.

Previous works (Poggianti et al. 1999; Dressler et al. 1999, 2004, 2008) observed a high frequency of e(a) galaxies both in clusters and in the field at z = 0.4–0.5, suggesting that the presence of a dusty starburst is not caused by the cluster environment (see also Ma et al. 2008). Table 6 shows that the e(a)/active and e(a)/emission fractions observed in the EDisCS and MORPHS clusters and field are very similar.

Besides confirming the ubiquity of e(a) spectra in all environments at z = 0.4–0.8, we find that the highest e(a) frequency is observed in groups (∼45% of the star-forming galaxies), while lower fractions (20–30% of the star-forming galaxies) are observed in clusters, poor groups, and the field. Interestingly, the e(a)/emission fraction is high in all types of groups, being similarly high in low- and high-O ii groups. Groups with σ = 150–400 km s⁻¹ are therefore the most favorable environment for triggering an e(a) spectrum. The frequency of dusty starburst candidates in groups could be related to the high incidence of mergers expected in these environments.

The results presented so far highlight the strong environmental dependence of the frequency of k+a spectra, and a milder but noticeable variation in the proportion of dusty starburst candidates with environment. It is worth noting that even in a mass-limited sample the different mix of SF histories in the various environments could arise from a different distribution of galaxy masses with environment (see Poggianti et al. 2008). In order to test whether our results are driven by this effect, we constructed a mass-matched sample, drawing, for each cluster galaxy, a galaxy of similar mass in each of the other environments. All conclusions presented in this section remain unchanged: strong environmental variations in the spectral fractions persist in the mass-matched samples. Hence, variations in the galaxy mass distribution are not responsible for the observed variations in spectral fractions with environment. Galaxies of similar masses must present significantly different SF histories depending on environment.

5.2. Dependence on Cluster Properties

In Sections 5.0 and 5.1, we have shown that the fractions of post-starburst and, to a smaller extent, of dusty starburst candidates depend on the global environment, when environment is subdivided into clusters, different types of groups, poor groups, and the field. To investigate whether the SF histories depend on the cluster and group properties in more detail, we now examine how the spectral fractions depend on the velocity dispersion of the system. The latter can be considered to be a proxy for the system mass, being M ∝ σ³ at a fixed redshift.

The fractions of k+a's among all galaxies and among active galaxies are listed for individual systems in Table 1 and shown in the left and right panels of Figure 5, respectively.²⁵ In these plots, we indicate with empty symbols groups and clusters with a low-O ii content for their σ, i.e., the outliers in the O ii–σ diagram. For comparison, we also show the values for the MORPHS clusters, whose spectral fractions are taken from Poggianti et al. (1999).

Zoom In Zoom Out Reset image size

Both the k+a/all and—more importantly—the k+a/active fractions of nonoutliers appear to increase with cluster velocity dispersion. The O ii outliers, instead, depart from this correlation and possibly follow a parallel one, with higher fractions at a given σ. In fact, low-O ii groups have a k+a incidence similar to more massive clusters. For EDisCS clusters the correlation probabilities are significant (99.1% and 99.98%, respectively) only when O ii outliers are excluded. Including both EDisCS and MORPHS points, the Spearman's test for all points (O ii outliers and nonoutliers) yields a 99.1% and 99.7% probability of a correlation, respectively.²⁶

A summary view of the trends with σ is shown in Figure 6, where we have grouped EDisCS structures into three velocity dispersion bins (σ>750 km s⁻¹, 400 < σ < 750 km s⁻¹, and σ < 400 km s⁻¹), keeping low- and high-O ii groups separate (empty and solid low-σ points). This figure shows how, on average, the k+a fractions in nonoutliers correlate with σ. It is useful to keep in mind that this is an average trend, while the scatter for individual clusters can be appreciated in Figure 5, where for example a few intermediate-mass clusters with a null k+a fraction are visible. Figure 6 also visualizes the fact that, while the average fractions in high-O ii groups are compatible with the field and poor group values, the other environments have average k+a fractions higher than the field/poor groups.

Zoom In Zoom Out Reset image size

In contrast to k+a's, neither the e(a)/all, nor the e(a)/active, nor the e(a)/emission fractions, even just for nonoutliers, depend on velocity dispersion (see Figure 7). Hence, while the "quenching efficiency" depends on the cluster properties, the incidence of dusty starburst candidates among all and among active galaxies does not.

Zoom In Zoom Out Reset image size

The trend of the post-starburst fractions with velocity dispersion is a remarkable result that could not be investigated in previous studies due to the limited cluster mass range explored. It is worth noting that the correlation of the post-starburst fraction with cluster mass is likely to be the cause of the differences in the global spectral fractions for EDisCS and MORPHS clusters that were discussed in Section 5 (see Figure 5 and Table 6). The EDisCS sample includes clusters as massive as those in the MORPHS sample, but also a number of lower mass systems. Therefore, the "average" EDisCS cluster has a lower mass than the "average" MORPHS cluster, and this is reflected in the lower global k+a fraction in EDisCS clusters.

The observed trends of the k+a fractions with σ raise two main questions:

• 1.  
  Why are more massive systems progressively more efficient in quenching SF (the k+a/active fraction)?
• 2.  
  What physical conditions and/or structure history set the outliers apart from the other systems, and cause them to quench SF with a higher efficiency than other systems of their mass?

These questions will be addressed in Section 8, after having gathered additional observational clues.

We proceed comparing the k+a fractions with the star-forming fraction in Figure 8, where the latter includes galaxies with EW(O ii)>3 Å  as in Poggianti et al. (2006). Both the k+a/all and k+a/active fractions anticorrelate with the fraction of O ii emitters (96.4% and 99.1% in EDisCS, 99.98% and 99.91% if both EDisCS and MORPHS are included). The peculiarity of most outliers disappears in this diagram, because they align with the rest of the systems.

Zoom In Zoom Out Reset image size

We note that, at a fixed O ii fraction f_O ii, both the k+a/all and the k+a/active fractions can be at most equal to (1−f_O ii). As a consequence, in Figure 8 there are "zones of avoidance," shown as shaded regions, that cannot be populated by data points due simply to the definition of the quantities on the X and Y axes. However, this upper limit cannot induce forced anticorrelations as those we observe.

To further investigate whether the anticorrelation could arise as a result of the definition of the different galaxy classes, we perform two types of simulations, simulating 200 clusters with a distribution of number of member galaxies equal to that observed. For the first simulation, we show in Figure 8 the position of the simulated systems starting from a uniform distribution of f(O ii)'s between 0 and 1 and allowing the k+a fraction to vary randomly and uniformly between 0 and (1−f(O ii)) (black small circles). The simulated systems are scattered all over in the k+a/all versus f(O ii) diagram, except for the avoidance zone. In the k+a/active versus f(O ii) diagram, most points of the random distribution are concentrated in a stripe parallel and close to the avoidance limit. In both cases, the observed anticorrelations are incompatible with the distribution expected from the simulation, as confirmed by a two-dimensional Kolmogorov–Smirnov (KS) test.

In the second test, in all of the simulated clusters we draw random realizations assuming binomial distributions with probabilities of the galaxies being star-forming equal to 50%, and adopting a 30% probability that any one of the non-star-forming galaxies is a k+a. This test is shown as the open red circles in Figure 8. The cloud of simulated clusters illustrate something akin to a point spread function in these two diagrams. As expected, the simulated points cluster around f_O ii = 0.5 and f_k+a = 0.15. They display a very large scatter in k+a/all fraction, failing to account for the observed trend. A two-dimensional KS test rules out the hypothesis that the observed and the simulated distributions are drawn from the same parent distribution.

These simple tests show that the mere definitions of the plotted quantities is not sufficient to explain the observed distribution of points, suggesting a physical origin for the link between the frequency of the various spectral classes.

To investigate such an origin, we ran a third simulation, starting from the observed distribution of f_O ii as measured by Poggianti et al. (2006). In this work, the f_O ii fraction was found to follow a general broad anticorrelation with cluster velocity dispersion, although with a number of outliers. For the galaxies determined to be nonemission-line galaxies in each cluster we then randomly determine whether they are k+a galaxies, adopting a 30% probability that any one of the nonemission-line galaxies is a k+a. Results are shown as blue open triangles in Figure 8 and are compatible with the trends observed in both panels. For this third case, KS tests allow the observed and the simulated distributions to be drawn from the same parent distribution (the KS probabilities for the distributions not to be drawn from the same parent distribution are 0.109 and 0.703).

Assuming a constant k+a probability among nonemission-line galaxies (0.3 in the third simulation) physically means a higher k+a production rate per emission-line galaxy in systems with lower star-forming fractions. In fact, it is important to keep in mind that the population of emission-line galaxies is the "reservoir" to produce k+a's, thus a fixed k+a probability among nonemission-line galaxies corresponds to adopting a higher conversion efficiency from emission-line to k+a in systems with lower star-forming fractions.

Therefore, the anticorrelation in Figure 8 might point to a connection between the "quenching efficiency" and the fraction of cluster galaxies that still retain an ongoing SF activity or, equivalently, the fraction of cluster galaxies that are devoid of SF (1−f(O ii). Naively, this connection seems to imply that the observed proportion of passively evolving galaxies in a system depends directly on the capacity of the cluster/group to truncate the SF activity in currently infalling galaxies. Assuming that the SF activity, once quenched, does not re-start at later times, this result is perhaps unsurprising if environments that are good at quenching at the moment we observe them have been good at quenching also at previous times, even if the quenching mechanism may not be the same over time. Attempting an interpretation of the observed anticorrelation, however, requires an articulated model for the accretion and quenching history of clusters, that is beyond the scope of this paper.

Composite cluster spectra can be obtained combining the light of individual galaxies. This method is complementary to the analysis of single galaxy spectral fractions, and provides a measurement of the "global" SF history integrated over all galaxies in a cluster (Dressler et al. 2004; hereafter D04). The high S/N of composite spectra allows a robust comparison of the global stellar history as a function of environment and redshift.

A composite spectrum for each of our clusters was obtained by summing the spectra of cluster members, after normalizing each spectrum by its mode and assigning a weight equal to its observed I-band luminosity. Each spectrum was further weighted for spectroscopic incompleteness as a function of galaxy magnitude and position in the cluster (see Section 2).²⁷ The EW(O ii) and EW(Hδ) for our composite spectra are listed in Table 1.

As suggested by D04, the most useful diagnostic diagram is a plot of EW(Hδ) versus EW(O ii) (Figure 9). In this diagram, clusters with a Balmer-strong (e.g., a prominent post-starburst) population will lie at higher EW(Hδ) for a given EW(O ii) than clusters without a Balmer enhancement. The local reference location in this plot is taken from D04: low-z clusters are shown as empty squares and have low-O ii and low Hδ. The expected location of the "normal" field population is shown by the dashed and dotted lines, also taken from D04. These represent mixes of different proportions of passive and quiescently star-forming galaxies, which are supposed to represent the field at different z's and to follow the variation in the star-forming fraction with redshift. MORPHS clusters at z = 0.4–0.5 (filled triangles) show an Hδ excess and occupy a region of the diagram separated from the location of both low-z clusters and the evolving field. We identify the region occupied by MORPHS clusters as a box in the upper left corner of Figure 9.

Zoom In Zoom Out Reset image size

EDisCS clusters and groups are shown as solid circles. Only a few of them lie in the upper left box where MORPHS clusters lie. These are namely Cl 1103a, Cl 1119, Cl 1054-12a, and, marginally, Cl 1216, Cl 1420, and Cl 1354. The great majority of the low-O ii groups are therefore found inside the Hδ-excess box. Other Hδ-excess systems might be Cl 1138a and Cl 1232.²⁸ The O ii and Hδ EWs of the composite spectrum of the Hδ-excess systems (named "Excess clusters and groups") are given in Table 1, and coincide precisely with those in the MORPHS cluster composite, taken from D04 and shown as large empty triangle.

In contrast, the majority of our other clusters and groups display no clear Hδ excess, being located around the region of the diagram corresponding to combinations of "normal" star-forming and passively evolving galaxy spectra. Composite high-O ii groups, poor groups, field and all our cluster and groups together (large empty circle, asterisk, cross, large filled circle, respectively) all lie close to the "normal galaxy" sequence.

The analysis of cluster composite spectra thus reinforces the conclusions based on individual galaxy spectra and presented in the previous section: only a subset of the distant clusters and groups have an outstanding Balmer-strong population. The latter appears to be conspicuous in low-O ii groups and some clusters, that tend to be mainly the most massive ones. The rest of our clusters and groups, as well as the poor groups and the field, do not show a notable Hδ-excess.

In the following, we analyze other galaxy properties and the local environment of galaxies of different spectral types that may shed some light on the origin of the various SF histories and the evolutionary link between them.

7.1. Morphologies

Visual morphologies based on Hubble Space Telescope Advanced Camera for Surveys (HST/ACS) images are available for 10 of the 20 EDisCS fields (Desai et al. 2007). The morphological distribution of galaxies of different spectral types is shown in Figure 10. The majority of passively evolving (k) and k(e) galaxies are ellipticals and S0's, with a small tail of Sa's and later types. Post-starburst (k+a) galaxies are mostly early disk galaxies of types S0's and Sa's. Actively star-forming galaxies, as expected, are mostly spirals, with e(c)'s being a mix of Sb's, Sc's, and some earlier types, e(a)'s and e(a)+'s being typically Sb and Sc galaxies, and e(b)'s tending to have on average later type morphologies, principally Sc's to Irregulars. Most irregular galaxies have an e(b) spectrum, as in the local universe.

Zoom In Zoom Out Reset image size

In general, the morphological mix follows what is expected from the SF properties, with early-type, passively evolving galaxies and late-type star-forming galaxies, and k+a galaxies that are transition objects with early disk morphologies. The k+a morphological distribution we observe is shifted toward slightly earlier morphologies than that observed by the MORPHS (Dressler et al. 1999). Our k+a's are typically S0's and Sa's, while several MORPHS k+a's display later spiral morphologies.

7.2. Masses, Luminosities, and Mass-to-light Ratios

Stellar mass, absolute V magnitude, and stellar mass-to-light ratio distributions of galaxies of different spectral types are shown in Figure 11.

Zoom In Zoom Out Reset image size

As it is natural to expect on the basis of what has been shown above and in previous works, the spectral classification is also a sequence of decreasing average galaxy mass and decreasing mass-to-light ratio going from k/k(e), k+a, e(c), e(a)/e(a)+ to e(b) types. The absolute magnitude distribution, instead, is more uniform across spectral types, except for e(b) galaxies that are the least massive, but also the least luminous, class.

This strengthens some of the conclusions found by previous works, and provides some new insights:

• 1.  
  Strong emission-line (e(b)) galaxies are not the primary progenitors of k+a galaxies, being significantly less massive than k+a's (Poggianti et al. 1999). The obvious candidate progenitors of the bright/massive k+a's included in our sample are star-forming galaxies of the e(a), e(a)+, and e(c) types.
• 2.  
  Since the stellar mass distributions of k and k+a galaxies are rather similar, the fate of the observed k+a's (assuming they never resume a SF activity) is to increment the class of massive, passively evolving k galaxies. Given also their morphologies (Section 7.1), k+a's are massive galaxies in the transition phase from actively starbursting or star-forming disk galaxies to passively evolving early-type galaxies. This scenario has been proposed and discussed before in several works (e.g., Poggianti et al. 1999), although galaxy masses were not available then.
• 3.  
  If this scenario can explain the evolution with redshift of galaxy properties (SF activity and morphologies) in clusters and, as we now find, in the low-O ii groups, the paucity or even lack of k+a galaxies in other environments leads to the question of whether/how galaxies turn from blue actively star-forming to red passively evolving in high-O ii groups, poor groups, and the field. Currently it is unknown how to what extent the observed strong evolution of the number ratio of red to blue galaxies on a cosmic scale (e.g., Bell et al. 2007) can be accounted for by the presence of k+a galaxies in the different environments and, conversely, how many galaxies in each environment turn from blue to red without going through a k+a phase. It is hard to assess quantitatively the impact of k+a galaxies on the cosmic SF, not knowing the relative proportion of low-O ii and high-O ii groups, and what fraction of all galaxies could be processed through the low-O ii group environment.

7.3. Local Densities and Radial Distributions

The location of galaxies of different spectral types within the cluster may provide some clues about the physical cause for the spectral characteristics and, therefore, the SF pattern.

Distributions of projected local density for galaxies in our clusters and groups are shown in Figure 12 for each spectral class. The local galaxy density was computed within the circular area enclosing the closest 10 neighbors, as described in Poggianti et al. (2008). E(a), e(a)+ and e(b) galaxies are rare in regions of densities higher than 100 galaxies/Mpc². The same is true for the youngest k+a galaxies (solid histogram), identified as those with the strongest Balmer lines and blue colors (r.f. B − V ⩽ 0.7, EW(Hδ)>4Å and EW(Hζ)>5Å), that tend to inhabit, on average, lower density regions than the general k+a population.

Zoom In Zoom Out Reset image size

Even more significant is to look at the incidence of post-starbursts and dusty starburst candidates among the active population as a function of local density (Figure 13). The overall k+a/active and k+a/all fractions are rather flat with density, except for a possible peak at the lowest densities we sample. Isolating the youngest k+a's (shaded histogram in the left panel of Figure 13), their incidence tends to increase toward lower densities. A similar trend with local overdensity is reported by Yan et al. (2008) for the DEEP2 Galaxy Redshift Survey at z ∼ 0.8. An opposite trend with local density has been instead found by Ma et al. (2008) in a cluster at z = 0.5. Our histogram shows that a preference for underdense regions does not necessarily imply that k+a's occur preferentially outside clusters: all our young k+a's shown in Figure 13 reside in massive EDisCS clusters. This trend with local density is compatible for example with k+a's occurring at the edge of infalling structures, where they first encounter the turbulent ICM. This scenario is also consistent with the clustercentric distance distribution of k+a's shown below.

Zoom In Zoom Out Reset image size

Interestingly, the k+a/active trend with density is paralleled by the e(a)/active fraction that is consistent with a mild decrease with density (right panel in Figure 13). In contrast, the e(c)/active fraction has the opposite trend, and sharply increases toward the highest densities.

Finally, we consider the distribution of clustercentric radial projected distance of galaxies of different spectral types. Figure 14 shows the cumulative radial distribution in units of virial radii. The virial radius was computed as R₂₀₀/1.14 (Biviano et al. 2006; A. Biviano 2007, private communication). The distribution of all emission-line galaxies is shown in the left panel and contrasted with the distributions of k and k+a galaxies. The single emission-line classes are shown in the right panel.

Zoom In Zoom Out Reset image size

The SF sequence corresponds to a progressive shift in radial distributions: the oldest (k) galaxies are the most centrally concentrated, the recently star-forming (k+a) are less concentrated than k's and the currently star-forming (emission line types) galaxies are the least concentrated, as found before, e.g., by Dressler et al. (1999) and Moran et al. (2005). The distribution of emission-line galaxies differs from that of k and k+a galaxies at the 99.9% and 99.2% level, respectively. e(c)'s seem to be the most centrally concentrated and e(b)'s the least concentrated among emission-line galaxies, although numbers are too small for the differences among the various emission types to be statistically significant.

The radial distributions of galaxies within the cluster are consistent with a scenario in which the longer a galaxy has been inside a cluster, the longer ago was its SF terminated and the longer is the time it has had to dynamically relax within the cluster potential. The projected local density distributions show a tendency for young k+a galaxies to be located in low density cluster regions, as some of their most likely progenitors: e(a) galaxies.

In this paper, we have found that the incidence of k+a (post-starburst) galaxies at intermediate redshifts depends strongly on environment. K+a galaxies at z = 0.4–0.8 preferentially reside in clusters and, unexpectedly, in a subset of the groups, those with a low-O ii fraction. In these environments, the SF activity is truncated on a short timescale (≪1 Gyr) in a large fraction (20%–30%) of the recently star-forming population.

In contrast, there are proportionally fewer or even no k+a galaxies in other environments, namely the field, the poor groups, and the groups with a high O ii fraction.

These results are based on the observed variation of the fractions of different spectral types with environment and are confirmed by the study of cluster and group integrated composite spectra.

The quenching efficiency—measured as the ratio between the number of k+a galaxies and the number of galaxies with an ongoing or recent SF activity—correlates with velocity dispersion. More massive systems have thus higher proportions of k+a galaxies, and higher quenching efficiencies.

In addition, low-O ii systems, characterized by a low-O ii fraction for their mass, display an excess of k+a galaxies compared to other systems of similar mass. Moreover, all systems follow a tight anticorrelation between the k+a/active galaxy fraction and the fraction of O ii emitters. This anticorrelation could imply that the number of passively evolving galaxies in a cluster is closely linked with the number of recently quenched galaxies.

Dusty starburst candidates present a very different environmental dependence from post-starburst galaxies. They are numerous in all environments at z = 0.4–0.8, representing at least 20% of the star-forming population, but they are especially numerous among star-forming galaxies in both low-O ii and high-O ii groups, where they represent 45% of the emission-line spectra (Table 6). This favors the hypothesis that a dusty starburst can be triggered by mergers that are expected to be common in groups.

Since dusty starburst candidates are present in similar proportions of the star-forming population in high- and low-O ii groups, while k+a galaxies are present only in the latter type of groups, it is reasonable to conclude that the SF enhancement and the truncation of the SF are not necessarily associated phenomena and are caused by different processes: the e(a) phase is not necessarily followed by a k+a phase.

The spectral classification scheme, from passively evolving (k) to post-starburst (k+a), to quiescently star-forming (e(c)) to starbursts and starburst candidates (e(a), e(a)+ and e(b)) is also a sequence of progressively later morphological types, lower galaxy stellar masses and lower mass-to-light ratios. We confirm that the properties of k+a galaxies are consistent with them being observed in a transition phase, at the moment they are rather massive S0s and Sa galaxies, presumably evolving from star-forming later types into passively evolving early-type galaxies. We also confirm that the clustercentric radial distribution is consistent with this evolutionary connection between the different types.

Interestingly, the observed frequency of k+a's shows no preference for the highest density regions. In fact, the incidence of young k+a's suggests there is a higher k+a production rate toward the low density regions within the cluster and group virial radius, although the number of galaxies is too low to draw definite conclusions.

8.1. Discussion

We would like to thank John Moustakas for helpful discussions and for sharing unpublished results, Fabio Barazza for providing his bar classification catalog, and the anonymous referee whose report triggered several improvements in the paper. B.M.P. thanks the Alexander von Humboldt Foundation and the Max Planck Instituut für Extraterrestrische Physik in Garching for a very pleasant and productive stay during which the work presented in this paper was carried out. B.M.P. acknowledges financial support from the National Institute for Astrophysics (INAF) through its PRIN-INAF2006 scheme. The Dark Cosmology Centre is funded by the Danish National Research Foundation.

Recently, Tremonti et al. (2007) have reported the discovery of massive gas outflows in galaxies with strong Balmer lines in absorption in SDSS spectra at z ∼ 0.6. Most of these spectra also display an O ii line in emission, judging from Figure 1 in Tremonti et al. (2007). Therefore, according to our classification scheme, they would belong to our e(a) class.

Evidence for galactic winds in these galaxies comes from the Mg ii 2796,2803 lines that are blueshifted by as much as ∼2000 km s⁻¹ with respect to the stars. A large mass of outflowing gas can be an efficient method for expelling the galactic gas and terminate the SF activity. The end of the SF in this case would be related to a phenomenon intrinsic to the galaxy itself such as an AGN.

184 of our spectra reach sufficiently out in the blue to cover the Mg ii 2796,2803 lines. Of these, 31 are cluster members, and of these 13 have spectral characteristics similar to those of Tremonti et al., i.e., have strong Balmer lines in absorption and relatively weak O ii lines. The great majority of these galaxies belong to our e(a) spectral class, with the exception of 1 k+a. The resemblance of these spectra with those of the galaxies in Tremonti et al. sample can be appreciated comparing our Figure 15 with their Figure 1.

Zoom In Zoom Out Reset image size

Measuring the position and strength of the Mg ii lines in the spectra of these 13 galaxies, we are able to identify the Mg II lines in most cases and we do not detect significant wavelength shifts in any case. Figure 15 shows the composite spectrum of our six best S/N cases, with the strong Balmer absorption lines (EW(Hδ)= 4 Å), the relatively weak O ii (EW=12 Å) and the Mg ii lines in place at approximately their rest-frame wavelength (right panel in Figure 15).

Moreover, Tremonti et al. (2007) have shown that there is an observed correlation between the outflow velocity v and the galaxy absolute B luminosity in both low-z and high-z samples. Given the moderate typical B luminosity of our small sample (〈M_B〉 = −20.4), the lack of a blueshift in the Mg ii line is consistent with the v–M_B correlation.

Thus, we find that galaxies with strong Balmer absorption and weak to moderate emission in our sample do not display signs of galactic winds. Our result shows that this combination of optical spectral features does not need to be generally associated with AGN-powered outflows. The Tremonti et al. galaxies were targeted for SDSS spectroscopy as quasar candidates, though they were subsequently classified as galaxies. Selection criteria other than the strength of Hδ and O ii must play a role for identifying galaxies with galactic winds.

Facilities: VLT (FORS2), HST (ACS), NTT (SOFI).