STAR FORMATION QUENCHING IN HIGH-REDSHIFT LARGE-SCALE STRUCTURE: POST-STARBURST GALAXIES IN THE Cl 1604 SUPERCLUSTER AT z ∼ 0.9

Our ground-based optical photometry consists of two pointings of the Large Format Camera (LFC; Simcoe et al. 2000) in Sloan r', i', and z' and two pointings of the Carnegie Observatories Spectroscopic Multislit and Imaging Camera (COSMIC; Kells et al. 1998) in Cousins R and Gunn i filters, both on the Palomar 5 m telescope. The layout of the observations can be found in Figure 4 of Gal et al. (2008). These data were reduced using the Image Reduction and Analysis Facility (IRAF; Tody 1986) with a set of publicly available routines. The detailed data reduction process is described in Gal et al. (2005). The 5σ limiting magnitudes are 25.2, 24.8, and 23.3 in r', i', and z', respectively. However, we have improved the source detection process from the previous work, as detailed in Section 3.1.

Near-infrared (NIR) J- and K_s-band images were obtained with the Wide-Field Camera (WFCAM; Casali et al. 2007) on the United Kingdom Infrared Telescope. This camera has four widely separated detectors, and four tiled pointings completely fill in a square with 075 sides. Observations were taken between 2007 and 2010, in ∼09 seeing. Images were taken with individual exposure times of 20 s in J and 10 s in K_s, with microstepping among exposures to recover some of the resolution lost to WFCAM's large pixels. Total integration times per pixel were 1500 s and 1800 s in J and K_s, respectively. Data were processed by the WFCAM reduction pipeline at the Cambridge Astronomy Survey Unit. The "deepleavestacks" for each pointing were retrieved from the WFCAM science archive and combined using the software task SWarp (Bertin et al. 2002).

Part of the Cl 1604 field is covered by a 17-pointing HST Advanced Camera for Surveys (ACS) mosaic in the F606W and F814W filters. Observations consist of two pointings from GO-9919 (PI: H. C. Ford) with effective exposure times of 4840 s and 15 pointings from GO-11003 (PI: L. M. Lubin) with effective exposure times of 1998 s. The integration time of 1998 s results in 5σ point-source limiting magnitudes of 27.2 and 26.8 mag in F606W and F814W, respectively. In terms of physical units, it corresponds to 94.0 L_☉ pc⁻² (Ascaso et al. 2014). Detailed descriptions of the observations, data reduction, and photometry are provided in Kocevski et al. (2011a). In our study, the HST data are used to obtain optical galaxy colors, examine blended sources in LFC images, and determine galaxy morphologies.

The entire Cl 1604 field was imaged by the Spitzer Infrared Array Camera (IRAC) at 3.6 and 4.5 μm and by the Multiband Imaging Spectrometer (MIPS) at 24 μm. For the observational program, we refer the reader to Kocevski et al. (2011a). Data were reduced using the standard Spitzer Science Center (SSC) reduction pipeline. For IRAC images, the reduced basic calibrated data (BCD) images were further processed using a modified version of the SWIRE survey pipeline (Surace et al. 2005a) to remove instrumental artifacts. The processed images were co-added using the SSC MOsaicker and Point source EXtractor (MOPEX) version 18.4.9. We created the Fiducial Image Frame using both IRAC channels, resulting in a unified astrometric reference for both channels. The permanently damaged pixels of the detector and bad pixels in each BCD were masked out when co-adding. The pixel scale was set to 06 pixel⁻¹.

The MIPS data reduction is described in Kocevski et al. (2011a). With a total exposure time of 1200 s pixel⁻¹, we achieved a 3σ flux limit of 40 μJy. Following the recipe and the synthetic spectra of Chary & Elbaz (2001) and Dale & Helou (2002), we calculated the total IR luminosity (L_TIR; 8–1000 μm) of each 24 μm source. The SFR can be directly obtained from L_TIR by the L_TIR–SFR relation from Kennicutt (1998). Our 40 μJy flux limit corresponds to an L_TIR of 3 × 10¹⁰ L_☉ at z = 0.9 and an SFR of 5.2 M_☉ yr⁻¹ (Kocevski et al. 2011a).

The spectroscopic data were obtained using the Low-Resolution Imaging Spectrometer (LRIS; Oke et al. 1995) and DEep Imaging Multi-Object Spectrograph (DEIMOS; Faber et al. 2003) on the Keck 10 m telescopes.

The early LRIS campaign targeted galaxies in the vicinity of two of the constituent clusters, Cl 1604+4304 and Cl 1604+4321 (see Oke et al. 1995; Gal & Lubin 2004, for further details), down to i' < 23. Data were taken using the 400 line mm⁻¹ grating blazed at 8500 Å, with a central wavelength of 7000 Å. This setup provides spectral coverage from 5500 to 9500 Å, with a resolution of ∼7.8 Å. The LRIS data were processed using standard IRAF tasks and scripts.

The bulk of the redshifts used in this study come from observation of 18 slitmasks with DEIMOS taken between 2003 May and 2010 June. The first 12 slitmasks targeted mainly galaxies with 20.5 ⩽ i' ⩽ 24, with higher priority given to red galaxies based on their r' − i' and i' − z' colors. The remaining six slitmasks were designed to obtain a magnitude-limited sample to a depth of F814W ∼ 23.5 across a 167 × 5' subsection of the field running roughly north to south covering clusters Cl 1604+4304 and Cl 1604+4321 (hereafter clusters B and D, adopting the naming convention of Gal et al. 2008).

All DEIMOS slitmasks were observed with the 1200 line mm⁻¹ grating with an FWHM resolution of ∼1.7 Å (68 km s⁻¹) and a typical wavelength coverage of 6385–9015 Å. The exposure frames for each DEIMOS slitmask were combined using a modified version of the DEEP2 spec2d package (Davis et al. 2003; Newman et al. 2013). This package combines the individual exposures for each mask and performs wavelength calibration, cosmic-ray removal, and sky subtraction on a slit-by-slit basis, generating a processed two-dimensional spectrum for each slit. The spec2d pipeline also generates a processed one-dimensional spectrum for each slit.

In this work, we use objects with high-quality redshifts (Q ⩾ 3), measured from at least one secure feature and one marginally detected feature. We adopted the redshift range of the Cl 1604 supercluster to be 0.84 < z < 0.96. Our spectroscopic catalog contains 2445 objects, of which 531 are supercluster members with Q ⩾ 3.

We first created a unified astrometric reference frame using SWarp, spanning the spatial extent of the ground-based and IRAC images. All of the ground-based images were then resampled onto this common astrometric system with a new pixel scale of 0174 pixel⁻¹, the smallest pixel scale of all the images. Regions with no data were filled with zeroes, and a flag map for each image was created to indicate regions of no data or bad pixels. We then co-added the resampled versions of the WFCAM J and K, LFC r', i', and z', and COSMIC R and i' pointings to produce an ultradeep optical–NIR detection image. Photometry was performed by running SExtractor in dual-image mode, using the ultradeep image as the detection map, from which objects were detected and photometric apertures were determined. We then applied the same photometric apertures to each single-band resampled image. Using the deep stacked images as the detection map has three merits. First, the combined images provide significantly improved detection ability for faint objects compared to each single-band image. Second, this procedure ensures that photometry in each band is performed using apertures of the same size. Finally, the catalog produced in each band has, by construction, the same sources, so they are trivial to combine.

The LFC r', i', z' and COSMIC R and i' data were photometrically calibrated using SDSS. We refer the reader to Gal et al. (2008) for further details of the calibration process. For regions covered by overlapping pointings in the same filter, we use the measurement with lowest photometric error for our combined catalog. The WFCAM J and K data were calibrated using the same procedure but with photometry defined by the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006), and all magnitudes were converted to the AB system using J_AB = J_2MASS + 0.89384, K_AB = K_2MASS + 1.84024. The final catalog was corrected for Galactic reddening using the dust map from Schlegel et al. (1998).

Because the point-spread function of IRAC is significantly different from the ground-based images, we made a separate deep IRAC detection image by stacking the resampled IRAC ch1 and ch2 images and co-adding them. The photometry in the two IRAC bands was performed in a similar manner as for the ground-based images. The IRAC deep stacked image was used for object detection. We used a fixed aperture of 19 for measurement, which is recommended by the SWIRE survey, and applied aperture corrections to account for the finite aperture size (Surace et al. 2005b). MOPEX outputs IRAC images with flux densities in MJy sr⁻¹, which we convert to fluxes in μJy using

$$\begin{equation} 1\ {\rm MJy\ sr^{-1}} = 8.46152\ \mu \;{\rm Jy} \end{equation} \tag{ 1 }$$

and from μJy to AB magnitude with

$$\begin{equation} M_{AB} = -2.5 \log (\mu \;{\rm Jy}) + 23.9. \end{equation} \tag{ 2 }$$

Finally, we produced our master catalog by cross-matching the ground-based and IRAC catalog using simple nearest-neighbor matching. The offset limit was set to 15 between the two catalogs.

We derived galaxy stellar masses for the full Cl 1604 spectroscopic sample using the Le Phare code (Arnouts et al. 1999; Ilbert et al. 2006), based on χ² fitting analysis of the spectral energy distributions (SEDs). The SED templates were generated with the stellar population synthesis package developed by Bruzual & Charlot (2003). We adopted an exponentially decreasing SFH, SFH ∝e^−t/τ, with nine different values of decaying timescale τ, ranging from τ = 0.1 Gyr to τ = 30 Gyr. Three different metallicities, 0.2 Z_☉, 0.4 Z_☉, and Z_☉, were used. We used a Calzetti et al. (2000) extinction law with E(B − V) = 0, 0.1, 0.2, 0.3, 0.4, and 0.5.

We included photometry from r', i', z', J, K_s, and Spitzer 3.6 μm and 4.5 μm, giving seven total bands for our SED fitting. We required an object to be detected in at least four bands for SED fitting. The HST ACS has significantly better angular resolution than ground-based telescopes and Spitzer, and the photometric apertures in F606W and F814W are much smaller than those in other bands. Therefore, these two bands were not used for our SED fitting. A total of 525 objects were included in our analysis.

The SED fitting was performed in two steps. First, all available ground-based and Spitzer channel 1 and 2 photometric measurements for each object were used for fitting. After the initial fit, we rejected a photometric data point if it contributed more than 25% of the total χ² for that particular fit and deviated more than 10σ from the model, and we repeated the fit a second time with the remaining bands. This method assures that our SED fits are not affected by catastrophic mismeasured photometry with incorrectly measured small errors.

The uncertainty in the stellar mass of each galaxy is estimated by running 1000 Monte Carlo trials wherein the magnitude in each band is offset by a Gaussian random error, with the width of the Gaussian based on the measured photometric error of that object in that band. The standard deviation of stellar masses of 1000 trials is taken to be the stellar mass uncertainty. Figure 1 plots the mass uncertainty versus stellar mass. Filled and open circles are objects detected in all seven bands (r', i', z', J, K_s, 3.6 μm, 4.5 μm) and less than seven bands, respectively. First, we find that, for a given mass, whether or not an object is detected in all seven bands does not strongly affect the derived mass uncertainty. On the other hand, the mass uncertainty is mass dependent. For objects with stellar mass <10¹⁰ M_☉, the error increases significantly. The median and average Δlog (M/M_☉) are 0.23 and 0.33, while for massive objects (M > 10¹⁰ M_☉) they are 0.14 and 0.13, respectively. This is due to the less massive galaxies also being less luminous in the K_s, 3.6 μm, and 4.5 μm bands and thus having correspondingly larger photometric errors.

Zoom In Zoom Out Reset image size

Due to the typically poorer seeing in our imaging data compared to the spectroscopy, some of the spectroscopically identified galaxies are blended in ground-based images. Normally, this blending prevents us from obtaining correct photometry for these galaxies, and thus we cannot measure their stellar masses. Such objects were excluded from our analysis unless we could ensure that the blending did not significantly affect the photometry. First, if the object is separable in our HST imaging, and our source of interest is brighter than the contaminating blended source by more than 2 mag in both F606W and F814W, we include it in our catalog. Second, if an object in the redshift range of interest blends with a background high-redshift (z > 4) spectroscopically identified object, we considered the effect of blending to be insignificant (see Lemaux et al. 2009, for an estimate of limiting magnitudes of high-redshift galaxies in the Cl 1604 field). After rejecting blended objects, we obtain stellar masses for 489 galaxies with 0.84 < z < 0.96.

Regardless of the physical trigger(s) that result(s) in a post-starburst galaxy, there are two basic components: (1) a significant star formation period in the recent past, followed by (2) no ongoing star formation. Practically, the [O ii] λ3727 doublet emission line is often used as an indicator of ongoing star formation, and the Hδ absorption feature as a proxy for the presence of A-type stars (e.g., Poggianti et al. 1999; Dressler et al. 1999; Oemler et al. 2009; Vergani et al. 2010). If one uses these spectroscopic diagnostics, a galaxy in its post-starburst phase should have no or little [O ii] emission due to the lack of H ii regions that are forming O- and B-type stars, but have strong Hδ absorption from the dominant A-type star population born in the recent past.

While [O ii] and Hδ are usually used to select post-starburst samples, other diagnostics also have been adopted. For example, Blake et al. (2004) suggest using a combination of Hδ, Hγ, and Hβ equivalent widths (EWs). A sample selected in this way can reduce contamination from dust-obscured star-forming galaxies, in which the [O ii] emission is highly attenuated. Instead of using Balmer line EWs, Quintero et al. (2004), Yan et al. (2006), and Yan et al. (2009) decomposed galaxy spectra into young and old components, representing A- and K-type stars, respectively, and then used their ratio to indicate the relative stellar populations.

We adopt the conventional criteria using limits on [O ii] emission and Hδ absorption EWs to be consistent with the majority of the literature in selection methodology. In the literature, various EW cuts have been adopted, depending on both data quality and the desired contamination level. For example, Dressler et al. (1999) and Poggianti et al. (1999) considered EW([O ii]) > − 5 Å as "absent," Vergani et al. (2010) adopted EW([O ii]) > − 3 Å for the zCOSMOS survey, and Goto (2005) applied a more stringent cut of EW([O ii]) > − 2.5 Å for SDSS. This inconsistency among studies leads to difficulty in comparing results. In this paper we define two samples, both with EW(Hδ) > 3 Å, but with different EW([O ii]) limits, to examine the effects of different selection criteria. The O5 sample consists of galaxies satisfying the EW(Hδ) limit and with EW([O ii]) > − 5 Å. The more stringent O3 sample contains galaxies meeting the EW(Hδ) limit and with EW([O ii]) > − 3 Å. For discussion purposes, we further define an intermediate O35 sample as galaxies belonging to the O5 sample but not the O3 sample, i.e., galaxies with EW(Hδ) > 3 Å and −5 Å < EW([O ii]) < − 3 Å. Errors in the EW measurement of our K+A samples are typically 1–3 Å. We will discuss this effect on sample selection in Appendix B. Out of 489 galaxies with stellar masses, 48 galaxies belong to the O5 sample and 31 galaxies belong to the O3 sample. In this paper, if not further specified, the term "K+A galaxy" refers to the larger O5 sample.

We estimate the SFR corresponding to our adopted EW([O ii]) limit using the conversion between [O ii] luminosity, L([O ii]), and SFR of Kewley et al. (2004). To convert EW([O ii]) to L([O ii]), we use the i'-band magnitude to estimate the continuum as it covers the wavelength of [O ii] emission at z ∼ 0.9. For a galaxy at z = 0.9 (the median redshift of the Cl 1604 supercluster), i' = 22.2 (the median i' magnitude of the O5 sample), and assuming E(B − V) = 0.3 (Lemaux et al. 2012), EW([O ii]) = −5 Å and −3 Å correspond to SFR of ∼1.3 M_☉ yr⁻¹ and ∼0.8 M_☉ yr⁻¹, respectively.

To examine the large-scale environmental effects on K+A galaxies, we attempt to assign each spectroscopically confirmed member to one of the eight member clusters and groups in the Cl 1604 supercluster, using both the galaxy's position and velocity offset relative to the central position and redshift of each cluster/group. We associated a galaxy with a specific cluster or group if (1) its velocity offset from the systemic velocity of the cluster or group is δ_v < ±3σ_v, where σ_v is the velocity dispersion of the system, and (2) its projected distance to the cluster or group center r_proj ⩽ 2 R_vir, where R_vir is the virial radius.

We determined R_vir of each cluster or group using the relation R_vir = R₂₀₀/1.14 (Biviano et al. 2006; Poggianti et al. 2009), where R₂₀₀ is the radius at which the mean density of a cluster or group is equal to 200 times the critical density of the universe at the corresponding redshift. R₂₀₀ is calculated by

$$\begin{equation} R_{200} = \frac{ \sqrt{3} \sigma _v }{ 10 H(z)}, \end{equation} \tag{ 3 }$$

where H(z) is the Hubble parameter at redshift z.

The spatial distribution of galaxies in the Cl 1604 supercluster is shown in Figure 2. Each black dot represents a galaxy associated with a specific group or cluster, while gray dots are superfield members. Dashed circles indicate a projected distance of 2 R_vir from the center of each cluster/group. Squares, stars, and crosses represent cluster, group, and superfield K+A galaxies, respectively. Table 1 summarizes the properties of each cluster and group, including the R.A., decl., redshift, velocity dispersion, virial radius, total number of members, and number of members in the O5 and O3 K+A samples (Gal et al. 2008; Lemaux et al. 2012; Rumbaugh et al. 2013). In total, we have 489 galaxies with reliable stellar masses and [O ii] and Hδ EWs. We associated 297 galaxies in the field with a specific cluster or group in the Cl 1604 complex. The remaining 192 galaxies fall within the redshift range of the supercluster but are found in filamentary structures near clusters or groups (Gal et al. 2008; Lemaux et al. 2012). We refer to these galaxies as the superfield sample hereafter.

Zoom In Zoom Out Reset image size

Table 1. Properties of the Cl 1604 Galaxy Clusters and Groups

Name	αJ2000	δJ2000	$\bar{z}$	σv	Rvir	N	NO5	NO3
				(km s−1)	($h_{70}^{-1}$ Mpc)
Cluster A	241.09311	43.0821	0.8984	722.4 ± 134.5	0.98 ± 0.18	36	10	8
Cluster B	241.10796	43.2397	0.8648	818.4 ± 74.2	1.13 ± 0.01	82	11	8
Group C	241.03142	43.2679	0.9344	453.5 ± 39.6	0.60 ± 0.05	24	4	3
Cluster D	241.14094	43.3539	0.9227	688.2 ± 88.1	0.92 ± 0.12	91	2	2
Group F	241.20104	43.3684	0.9331	541.9 ± 110.0	0.72 ± 0.15	25	2	1
Group G	240.92745	43.4030	0.9019	539.3 ± 124.0	0.73 ± 0.17	23	4	2
Group H	240.89890	43.3669	0.8528	287.0 ± 68.3	0.40 ± 0.10	10	4	2
Group I	240.79746	43.3915	0.9024	333.0 ± 129.4	0.45 ± 0.13	6	1	0
Superfield						192	10	5
Total						489	48	31

Download table as:  ASCIITypeset image

3.5.1. Equivalent Widths and D_n(4000)

EWs of the [O ii] and Hδ features of each galaxy were measured using the bandpass method following Lemaux et al. (2010). The bandpass measurements were performed by defining two "continuum" bandpasses, slightly blueward and redward of the spectral feature, which are used to estimate the stellar continuum across the emission feature. An additional "feature" bandpass is defined to encompass the spectral line. Any pixels with large variance values were removed from the continuum bandpasses. We do not remove similar pixels in the feature bandpass. The EW is defined as

$$\begin{equation} {{\rm EW}(\mathring{\rm{A}})} = \sum \limits _{i=0}^n \frac{F_i - C_i}{C_i} \Delta \lambda _{r,i}, \end{equation} \tag{ 4 }$$

where F_i is the flux in the ith pixel in the feature bandpass, C_i is the continuum flux in the ith pixel over the same bandpass, and Δλ_{r, i} is the rest-frame pixel scale of the spectrum (in Å pixel⁻¹). Uncertainties in the EW were derived using a combination of Poisson errors on the spectral feature and the covariance matrix of the linear continuum fit and are given by (Bohlin et al. 1983)

$$\begin{equation} \sigma _{{\rm EW}}({\rm \mathring{\rm{A}}}) = \sqrt{\left(\sum \limits _{i=0}^n \frac{\sigma _{F,i}\Delta \lambda _i}{C_i} \right)^2 + \left(\sigma _C \sum \limits _{i=0}^n \frac{F_i\Delta \lambda _i}{C_i^2} \right)^2 }. \end{equation} \tag{ 5 }$$

We adopt the bandpasses for [O ii] and Hδ in Fisher et al. (1998), with modification by eye for each galaxy spectrum to avoid poorly subtracted airglow lines and to avoid contaminating features near the spectral lines of interest. A handful of galaxies were observed twice by DEIMOS in the survey campaign, with EW([O ii]) measurable for six such galaxies while EW(Hδ) can be measured in three galaxies. For each galaxy, we measured the EWs and the uncertainties of each spectral line for both observations. We compare the difference of EW of the same spectral line measured from the two observations, ΔEW, to the EW uncertainty, σ_EW. From the nine total measurements, we find that the median value of ΔEW/σ_EW is ∼1, suggesting that the EW uncertainty in Equation (5) is a fairly good description of the actual EW uncertainty.

We also measure EW(Hδ) and the D_n(4000) spectral indices of the composite spectra of our K+A samples to examine their average properties (see Section 3.5.2). The D_n(4000) is measured using the ratio of the blue (3850–3950 Å) and red (4000-4100) continua as defined by Balogh et al. (1999). Mean flux density values are calculated from the σ-clipped spectrum of each region, with the D_n(4000) index defined as D_n(4000) = 〈F_{λ, r}〉/〈F_{λ, b}〉. Errors on the D_n(4000) index are calculated from the variance spectrum in each region, again using σ-clipping to avoid regions of poor night-sky subtraction or regions that fell within in the 10 Å CCD chip gap.

3.5.2. Composite Spectra

We generated composite spectra of K+A samples to examine their average properties. We generated composite spectra for 24 μm detected and undetected K+A galaxies separately because they may have distinct properties, as suggested by the SFRs implied by their IR luminosities.

We made the composite spectra following the method of Lemaux et al. (2009). Each spectrum was first normalized by the galaxy's total spectral flux prior to stacking and then co-added using variance weighting (see Lemaux et al. 2009). We measured EW(Hδ), D_n(4000), and EW([O ii]) in each composite spectrum, and these are provided in Table 2.

Table 2. Spectral Indices of Composite Spectra of K+A Galaxies

Sample	EW(Hδ)	Dn(4000)a	EW([O ii])	〈LTIR〉	〈E(B − V)〉
	(Å)		(Å)	(1010 L☉)
O5 non-IR	4.12 ± 0.17	1.594 ± 0.007	−0.61 ± 0.20	<3.0	<0.62
O5 IR	5.27 ± 0.19	1.406 ± 0.006	−3.41 ± 0.20	21.5	0.79
O3 non-IR	3.86 ± 0.18	1.610 ± 0.008	−0.41 ± 0.21	<3.0	<0.50
O3 IR	6.22 ± 0.33	1.454 ± 0.009	−1.48 ± 0.35	12.4	0.57

Note. ^aNot corrected for extinction. See Section 3.5.2 for details.

Download table as:  ASCIITypeset image

Because D_n(4000) is not impervious to dust, the D_n(4000) of 24 μm detected galaxies must be corrected for dust attenuation. For example, the D_n(4000) of a galaxy with a 0.5 Gyr simple stellar population and E(B − V) = 0.5 would be ∼0.1 larger than a galaxy with the same stellar population but no dust (MacArthur 2005). To account for this effect, we estimate the average E(B − V) by comparing the SFR derived from [O ii] and L_TIR (Lemaux et al. 2013) by

$$\begin{equation} \langle E(B-V) \rangle = \frac{2.5\log (\frac{{\rm SFR}(\langle L_{{\rm TIR}} \rangle) }{ {\rm SFR}(\langle L_{{[\rm O\,\scriptsize{II}]}} \rangle) }) }{ k^{\prime }_{[ O\,\scriptsize{II} ]} }, \end{equation} \tag{ 6 }$$

where $k^{\prime }_{{\rm [O\,\scriptsize{II}]}}$ is the Calzetti et al. (2000) reddening curve evaluated at the wavelength of [O ii].

We list the estimated 〈E(B − V)〉 in Table 2. Given the E(B − V) we derived, the dereddened D_n(4000) of 24 μm detected galaxies would be up to ∼0.15 and ∼0.10 smaller than the measured values for the O5 and O3 samples, respectively. For 24 μm undetected galaxies, the 〈E(B − V)〉 cannot be derived in this way. We calculate an upper limit, using 〈L_TIR〉 < 3 × 10¹⁰ L_☉. On the other hand, if we take the median E(B − V) from the result of SED fitting, E(B − V) = 0.1, the effect is ∼0.02 (MacArthur 2005).

To determine the true fraction of K+A galaxies, we estimated the spectroscopic incompleteness as a function of color and magnitude in each system in the Cl 1604, for galaxies with F814W ⩽ 24, or i' ⩽ 24 in regions not covered by ACS. For each system, we first plotted the color–magnitude diagrams (CMDs) of ACS photometric sources and all spectroscopic samples with ACS imaging in projection on the sky. The CMD was then separated into bins of 0.5 mag in color (F606W − F814W) and 1 mag in magnitude (F814W) over the color range 0 ⩽ F606W − F814W ⩽ 2.5 and the magnitude range 20 ⩽ F814W ⩽ 24. The completeness was calculated for each color–magnitude bin for a given system by dividing the number of objects with high-quality redshifts by the number of photometric sources. Figure 3 shows the spectroscopic completeness dependence on magnitude, color, and host system. Also in Figure 3 we plot the radial dependence of the spectroscopic completeness in each cluster and the group composite. The completeness depends more strongly on a galaxy's location in color–magnitude space than on its distance to the host system center. Therefore, we compute the color–magnitude-dependent completeness in each system as a whole but do not further divide each system into smaller regions, resulting in more robust statistics for the dependence on color and magnitude.

Zoom In Zoom Out Reset image size

Assuming that galaxies in the same color–magnitude bin have similar physical properties, that is, our spectroscopic samples are representative in each color–magnitude bin, the number of members in each color–magnitude bin was then corrected for incompleteness, and summing over all bins gave the estimate total number of members with ACS coverage in each system. However, the ACS imaging does not cover the full extent of the spectroscopically mapped areas of the supercluster. Therefore, to obtain a full estimate of the cluster membership, we must further correct for this incomplete ACS coverage. We apply a multiplicative factor to the ACS-based galaxy numbers, which is computed as the ratio of the total number of spectroscopic members to those with ACS coverage (N_spec/N_{spec + ACS}) within each structure. This correction factor is simply the inverse of the ACS sampling fraction in the region of each system. The K+A population was corrected in a similar manner. We first calculated the fraction of K+A galaxies in each color–magnitude bin. The number of K+A galaxies missed by our incomplete spectroscopy is then the number of missed system members multiplied by the K+A fraction in each bin.

Table 3 lists the number of galaxies in each sample, as well as the completeness-corrected and uncorrected K+A fractions. The completeness-corrected K+A fraction in each system is plotted in Figure 4. We also show the number and fraction of only 24 μm undetected K+A galaxies in Table 3 and Figure 4. In this paper, all discussions of the K+A fractions will use the completeness-corrected values.

Zoom In Zoom Out Reset image size

Table 3. Prevalence of K+A Galaxies

System	N	O5	O3	O5-IR	O3-IR	$f^{{\rm uncorr}}_{O5}$	$f^{{\rm uncorr}}_{O3}$	$f^{{\rm uncorr}}_{O5-IR}$	$f^{{\rm uncorr}}_{O5-IR}$	$f^{{\rm corr}}_{O5}$	$f^{{\rm corr}}_{O3}$	$f^{{\rm corr}}_{O5-IR}$	$f^{{\rm corr}}_{O5-IR}$
Cluster A	34	10	8	8	6	0.29	0.24	0.24	0.18	0.19	0.13	0.11	0.09
Cluster B	73	11	8	7	6	0.15	0.11	0.10	0.08	0.12	0.08	0.06	0.05
Cluster D	80	2	2	1	1	0.03	0.03	0.01	0.01	0.05	0.03	0.02	0.02
Groups	82	15	8	10	7	0.18	0.10	0.12	0.09	0.11	0.06	0.06	0.03
Superfield	164	10	7	7	5	0.06	0.04	0.04	0.03	0.04	0.02	0.02	0.01
All Galaxies	437	48	33	33	25	0.11	0.08	0.08	0.06	0.07	0.04	0.04	0.02

Notes. Only galaxies with F814W < 24, or i' < 24 for those not covered by ACS, are included. The K+A fractions after completeness correction are derived as described in the text. Columns with "−IR" indicate K+A galaxies not detected at 24 μm.

Download table as:  ASCIITypeset image

We differentiated red and blue galaxies in the Cl 1604 supercluster by using the HST ACS CMDs. The F606W and F814W passbands are effectively rest-frame U and B bands at z ∼ 0.9. The red sequence in each constituent system is determined by minimizing the χ² of colors and magnitudes of member galaxies within a certain range to a linear model of the form

$$\begin{equation} F606W - F814W = y0 + m \times F814W. \end{equation} \tag{ 7 }$$

The detailed process is described in Lemaux et al. (2012). Because of the small number of galaxies in each group, we combined all five groups into a composite. Table 4 summarizes the parameters of the red sequence fits for each system. For clusters, the red sequence width is defined by ±3σ, while for the group composite and field samples, ±2σ is adopted because heterogeneous subsamples are included across a variety of redshifts, resulting in a larger color dispersion.

For galaxies covered by our ACS imaging, their morphologies were visually classified by one of the authors (L.M.L.). First, each galaxy is classified as either early-type (E or S0) or late-type (Sa through Sd and irregular). Second, galaxies showing tidal or disturbed features or ongoing mergers are further labeled as interacting. To estimate the reliability of our visual classification, a random subset of 150 supercluster members were presented to two of the authors (L.M.L. and R.R.G.) for classification. The result was then compared to the original classification. This process tests the consistency of visual classification from single observer and variation among different classifiers. We estimate that roughly 5%–10% of our sample may be morphological misclassified (Lemaux et al. 2012).

We present thumbnail images of the O3 and O35 samples in Figures 5 and 6, respectively. Table 5 summarizes the statistics of the visual morphological classifications. Numbers in parentheses indicate how many galaxies in each sample are also detected at 24 μm. For both the more stringent O3 sample and the less restrictive O5 sample, K+A galaxies have heterogeneous morphologies. They are predominantly early-type galaxies and not interacting or merging with another galaxy, but late-type or interacting systems also constitute a non-negligible portion. We will discuss the morphological heterogeneity in Section 5.2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 7 shows the CMD and color–stellar mass diagram of each cluster, the group composite, and the superfield in the Cl 1604 supercluster. K+A galaxies are labeled with open symbols. In the CMDs, the red sequence is shown by the dashed lines. In Figure 8 we plot the completeness-corrected color and stellar mass distributions of red, blue, and K+A galaxies, respectively. Each galaxy is weighted by the inverse of the completeness calculated in Section 3.6. For the stellar mass distribution, we only consider galaxies with log (M/M_☉) ⩾ 10 for two reasons. First, the uncertainty in stellar mass from SED fitting increases significantly below log (M/M_☉) = 10 (Figure 1). Second, for a passive galaxy with a near-instantaneous burst (τ = 0.1 Gyr) SFH, solar metallicity, and formation redshift z_f = 3, F814W = 24 (the magnitude limit down to which we correct for the completeness) roughly corresponds to log (M/M_☉) = 10. This result means that red galaxies with log (M/M_☉) < 10 are excluded by the magnitude cut. Therefore, when examining stellar mass distributions, we only discuss galaxies more massive than log (M/M_☉) = 10.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We see that the majority of K+A galaxies lie in the red sequence, with a minority in the blue cloud. Those K+A galaxies with bluest colors are in the O5 sample, possibly due to higher residual SFRs. The median F606W − F814W color of all K+A galaxies is 1.74, similar to the red sequence (1.79) and distinct from that of blue cloud galaxies (1.27) for galaxies with log (M/M_☉) ⩾ 10. K+A galaxies occupy the massive end of the mass spectrum, comparable to red sequence galaxies. The massive nature of these K+A galaxies is consistent with those in the field at moderate to high redshifts (Vergani et al. 2010). While Vergani et al. (2010) argued that K+A galaxies with 0.48 < z < 1.2 in the COSMOS field are on average more massive than red quiescent galaxies, we cannot distinguish the masses of K+A galaxies from red sequence galaxies in Cl 1604. The median stellar masses of red sequence galaxies and K+A galaxies are log (M/M_☉) = 10.54 (10.49 for the O3 sample) and log (M/M_☉) = 10.52, respectively. Given the uncertainty of our stellar mass estimate, Δlog (M/M_☉) = 0.14, the difference is insignificant. The similar colors and masses of K+A and red sequence galaxies suggest that, at least in high-density regimes, K+A galaxies are likely the progenitors of some red sequence galaxies, or vice versa.

On the other hand, blue-cloud galaxies are less massive than K+A galaxies, with a median stellar mass of log (M/M_☉) = 10.35 and the number counts increasing to lower mass. The difference between the median stellar masses of K+A galaxies and blue-cloud galaxies is ∼10¹⁰ M_☉, about 50% of the stellar mass of an "average" blue-cloud galaxy. In our sample, a K+A galaxy has an SFR less than 1 M_☉ yr⁻¹. Assuming that the SFR of an "average" blue-cloud galaxy drops to that of a K+A galaxy immediately and retains SFR ∼1 M_☉yr⁻¹ for 1 Gyr, the stellar mass only increases by ∼10⁹ M_☉ and thus cannot reach the mass of an "average" K+A galaxy passively. This result suggests that, for an average blue-cloud galaxy turning into a K+A galaxy, there should be an epoch of rapid mass assembly before the star formation shuts down. Further discussion will be presented in Section 5.2

The Cl 1604 supercluster contains galaxies in a variety of clusters and groups, with velocity dispersions ranging from ∼300 to 800 km s⁻¹, as well as those not belonging to any bound structure, providing a variety of environments all at similar redshifts. The prevalence of K+A galaxies ranges from 4% in the superfield to 19% in Cluster A. Even the three clusters, which have similar velocity dispersions and thus possibly similar mass, do not have similar K+A fractions. This variation among individual clusters has been shown in several previous studies (Dressler et al. 1999; Tran et al. 2003; Poggianti et al. 2009). One source of this variation may be the cluster mass, as Poggianti et al. (2009) found from 20 intermediate-redshift clusters in the ESO Distant Cluster Survey (EDisCS). They found that the K+A fraction correlates with the velocity dispersion of the host cluster, after binning their sample into three velocity dispersion bins. However, in each bin, the variation in the K+A fraction is still large. This variation is also seen in Cl 1604, where the three clusters have quite similar velocity dispersions, but in Cluster A, K+A galaxies are almost four times more prevalent than in Cluster D, whose K+A fraction is comparable to the superfield. Although the dependence on cluster mass may be present, it cannot be the only factor. The dynamical state of each cluster may also be important.

In addition to the prevalence of K+A galaxies, we also examined the spatial distributions of K+A galaxies in the clusters and groups. Figure 9 shows the cumulative projected radial distributions of K+A galaxies, along with those of red and blue galaxies, in clusters and groups as a function of virial radius. We combine clusters and groups by normalizing the projected distance of each galaxy by the virial radius of its parent system. In clusters, K+A galaxies are almost all located at <1R_vir from the center but not in the infalling region of 1 < R_vir < 2, while K+A galaxies in the lower-mass groups can be found at all radii. This distribution is unlikely an artifact of our spectroscopic sampling, as the average completeness in clusters and groups does not strongly depend on the projected distance to system centers (Figure 3). Thus, the distribution of K+A galaxies may be a clue to the preferred locations where quenching occurs in clusters and groups. Both the variation in K+A prevalence and radial distribution will be discussed in Section 5.

Zoom In Zoom Out Reset image size

The three clusters in the Cl 1604 supercluster are in different dynamical states, resulting in a range of physical properties (Gal et al. 2008; Kocevski et al. 2009; Lemaux et al. 2012). Among the three clusters, Cluster A is the most X-ray luminous and most massive and shows clear velocity segregation between red and blue galaxies (Gal et al. 2008; Kocevski et al. 2009). It likely formed at an earlier epoch so that its primordial galaxy population had more time to virialize and establish a different dispersion than the infalling galaxy population. Cluster A also falls on the X-ray–optical scaling relation, the σ_v–T relation (Kocevski et al. 2009; Rumbaugh et al. 2013). This relation arises if the ICM shares the same dynamics as the cluster galaxies, so that the ICM temperature scales with the cluster velocity dispersion as σ_v∝T^1/2, further supporting Cluster A as a virialized structure. Cluster B has the second-highest X-ray luminosity but no velocity segregation. It also has been shown that the velocity dispersion of its member galaxies is higher than inferred from the temperature of the ICM (Kocevski et al. 2009; Rumbaugh et al. 2013). This excess in velocity dispersion is often interpreted as an unvirialized system, with high-velocity infalling galaxies inflating the velocity dispersion. Gal et al. (2008) also noted that the redshift distribution of Cluster B shows evidence of a substructure or a triaxial system. These dynamics of its member galaxies indicate that Cluster B is not yet fully virialized and may still be undergoing collapse. Cluster D has the most spectroscopically confirmed members in the Cl 1604 supercluster but is not detected in X-rays, suggesting that it has not yet developed a dense ICM. In addition, its members have a filamentary distribution (Gal et al. 2008), also indicative of an unvirialized system.

The X-ray and optical properties of these three clusters suggest that they are at different dynamical states of evolution, with Cluster A the most evolved and Cluster D the most dynamically young. This difference among their dynamical states also leaves imprints on their galaxies' star formation properties. Kocevski et al. (2011a) found increased star formation activity in Cluster D compared to Clusters A and B and to the field, where the fraction of cluster members detected at 24 μm is higher in Cluster D than in the other systems. From composite spectra of the galaxies in each system, Cluster D members on average have the highest EW([O ii]) and lowest D_n(4000), suggesting a higher SFR and younger stellar age (Lemaux et al. 2012). The elevated rate of 24 μm sources is interpreted as enhanced star formation in clusters associated with actively infalling galaxies, with the likely trigger being interactions between galaxies (Kocevski et al. 2011a). This scenario is also supported by the radial distribution of 24 μm detected galaxies in clusters, which is less centrally concentrated than that of 24 μm undetected galaxies (Figure 8 of Kocevski et al. 2011a).

Contrary to 24 μm sources, the K+A fraction varies among clusters but with an inverse trend, with the most evolved Cluster A having a high K+A fraction and the dynamically youngest Cluster D having few K+A galaxies. Figure 4 plots the K+A fraction in each system for our various sample selection criteria. This variation is the same in all sample selections and could reflect a real difference among clusters.

As argued in Kocevski et al. (2011a), starbursts induced by interactions and mergers between infalling galaxies are occurring in Cluster D, which results in the elevated 24 μm source fraction seen in Cluster D. Meanwhile, we do not see the aftermath of these starbursts, the post-starburst galaxies, in Cluster D. The K+A fraction is comparable to that of the superfield, showing that in this early stage of cluster formation, where a cluster is actively accreting galaxies, starburst activity is triggered, but star formation quenching is not yet enhanced. The enhancement occurs at a later stage of cluster assembly, when the starbursts fade and move into their post-starburst phase. Cluster B is likely an example of such an environment, where the K+A fraction increases to 11%, about twice that of Cluster D and the superfield.

The high K+A fraction in Cluster A is intriguing. In the Cl 1604 groups, 11% of members are K+A galaxies, comparable to the fraction in Cluster B but much lower than the 19% seen in Cluster A. If mergers and interactions are the only routes producing K+A galaxies, the higher K+A fraction in Cluster A implies a higher frequency of merger-induced starbursts in Cluster A than in the Cl 1604 groups. The enhanced merger rate in cluster cores had been proposed by Struck (2006) and observed by Oemler et al. (2009) in a cluster at z ∼ 0.4. For this enhancement to happen, galaxies fall into the cluster as bound pairs or small groups rather than individually. The gravitational perturbation from the cluster to the orbit of the bound groups may cause the orbit to shrink, leading to a higher merging probability. If a substantial fraction of member galaxies of Cluster A were recently accreted via this route, it would explain the high K+A fraction. However, the relaxed dynamical state of Cluster A suggests against this scenario. Cluster A does not exhibit substructure that indicates a recent infall of a group of galaxies (Gal et al. 2008).

Alternatively, Dressler et al. (2013) suggested that not every merger turns into a K+A galaxy, but passive galaxies merging specifically with a smaller star-forming galaxy can create K+A phases. The excess of K+A galaxies in Cluster A may arise from a distinct underlying galaxy population in Cluster A compared to the other systems. In Cluster A, red sequence galaxies compose ∼60% of members (after correcting for incompleteness), much higher than the ∼30% in Cluster B, Cluster D, and the groups and ∼20% in the field. The high fraction of red sequence galaxies appears to be in line with the scenario of Dressler et al. (2013), so that mergers in Cluster A are more likely to involve a passive galaxy as a major component, thus producing more K+A galaxies. For this assertion to hold, there should be more early-type K+A galaxies in Cluster A than in other systems in order to account for the excess of K+A galaxies. After examining morphologies of K+A galaxies in Cluster A, we found five early-type galaxies out of nine K+A galaxies with ACS imaging. The early-type fraction of K+A galaxies in Cluster A is not higher than that in the whole Cl 1604 supercluster (see Table 5). Thus, the presence of more passive galaxies in Cluster A does not lead to a higher K+A fraction, but the numbers are too small to draw a definitive conclusion.

These results lead us to consider another possibility: interaction between galaxies and the ICM. Violent processes such as ram pressure stripping could remove cold gas from a cluster galaxy in ∼10⁷ yr (Abadi et al. 1999), producing a K+A spectrum. To remove all the gas of a Milky Way like galaxy, the ram pressure should meet the requirement (Gunn & Gott 1972; Treu et al. 2003)

$$\begin{eqnarray*} \rho _{{\rm gas}} v_i^2 &{>} & 2.1 \times 10^{-12} \;{ {\rm N}\; {\rm m}^{-2}} \left(\frac{v_{{\rm rot}}}{220 \;{ {\rm km}\; {\rm s}^{-1}}} \right)^2 \nonumber\\ && \times \left(\frac{r_h}{10 \;{ {\rm kpc}}}\right)^{-1} \left(\frac{\Sigma _{{\rm H\,{\scriptsize {I}}} } }{8 \times 10^{20} m_H \;{{\rm cm}}^{-2}} \right), \end{eqnarray*}$$

where ρ_gas is the gas density, v_i is the velocity of the galaxy, and the rotational velocity v_rot, scale length r_h, and H_I surface density of the galaxy in question are expressed in units for a Milky Way like galaxy. Making use of Chandra X-ray data, the gas density profile of Clusters A and B can be derived from the best-fit isothermal β-models from Rumbaugh et al. (2013). For a Milky Way like galaxy with v_i = 800 km s⁻¹, the stripping radii for both Clusters A and B are ∼0.3 R_vir. That means that ram pressure stripping is effective only for galaxies passing very close to the cores of Clusters A and B. For Cluster B, still undergoing collapse, some of the members would be just accreting onto the cluster and not yet have passed close to the cluster core. Although the ICM in the core of Cluster B is dense enough to strip a galaxy's gas, not enough galaxies have traveled deep enough into the cluster core and experienced this effect. On the contrary, galaxies in the dynamically evolved Cluster A have had a higher chance to interact with the dense ICM because they had been in the cluster for a longer period of time. In this case, besides galaxy mergers, an extra mechanism kicks in, boosting the K+A fraction in Cluster A.

The three clusters in Cl 1604 may be demonstrating one of the evolutionary tracks for galaxies assembling into clusters. At the early stage, a starburst may be triggered by a galaxy merger while a galaxy is infalling into the cluster. On average, the cluster has an actively star-forming population, and quenching is not yet enhanced. At a later stage, those galaxies have past their starburst phase and turn into post-starburst galaxies, while new galaxies are still infalling into the cluster, with starbursts being triggered. As the cluster evolves, fewer galaxies assemble into the cluster and it becomes virialized, with a dense ICM in the center. Earlier starbursts fade out, and the ram pressure becomes strong in the center, producing more post-starburst galaxies. However, this broad picture is based on only three clusters in the Cl 1604 supercluster. Future analysis of the full ORELSE survey, which has similar-quality DEIMOS spectral coverage in 20 large-scale structures at z ∼ 1, will be able to provide a more complete picture on the coevolution of galaxies and clusters.

We have seen in Table 5 that about 30% (13/43) of K+A galaxies exhibit interaction features. This simple morphological observation suggests that at least some K+A galaxies are related to galaxy mergers or interactions. Numerical simulations have demonstrated that galaxy mergers can result in SFHs that produce K+A spectral features (Bekki et al. 2005; Snyder et al. 2011). A galaxy merger could trigger a short period of elevated star formation activity, followed by a rapid fall-off. A few hundred million years after the peak of the starburst, the merger appears as a K+A galaxy, even before star formation activity fully stops. The whole K+A phase would last for another few hundred million years to as long as 1 Gyr, depending on the physical properties of the merging galaxies (Snyder et al. 2011). On the other hand, disturbed morphological features fade away at a timescale comparable to that of the SFR decline (Lotz et al. 2008, 2010a, 2010b). The merger-induced tidal arms can be prominent for ≲ 500 Myr, but the low surface brightness tidal tails around the spheroid remnant are difficult to observe locally and nearly impossible at z ≳ 0.5 (Mihos 1995; Yang et al. 2004). It can be expected that at higher redshifts, z ∼ 0.9, the observability of tidal arms decreases. Therefore, some merger-induced K+A galaxies may not be found to have interaction features, which means that the true population of merger-induced K+A galaxies could be much higher than the observed value of 30% of the whole K+A population. Also with HST images, Vergani et al. (2010) reported a high incidence of asymmetry in K+A galaxies at 0.48 < z < 1.2, supporting the hypothesis that K+A galaxies have experienced mergers or interactions in the recent past.

The spatial distribution of K+A galaxies in clusters and groups in the Cl 1604 supercluster is also consistent with a merger origin. In clusters, galaxy mergers are expected to happen more often at the outskirts than in the core, while in groups, the group centers are the preferred place for mergers. In the Cl 1604 supercluster, the starburst activity distribution gives a consistent picture, where starbursts in clusters are found in infall regions and within R_vir in groups (Kocevski et al. 2011a). So, for a starburst happening at the infall region of a cluster, the remnant likely ends up close to the cluster core after a few hundred million years while infalling into the cluster. For a starburst galaxy near the core of a group, it can end up anywhere during the K+A phase, depending on whether it is moving toward or away from the core.

Some of our K+A galaxies are bright at 24 μm, implying that they still host substantial star formation. These galaxies do not meet the strict definition of post-starburst galaxies (no or little ongoing star formation). Sometimes they are suggested to be dusty-starburst galaxies and are therefore excluded from post-starburst galaxy samples (Oemler et al. 2009). Figure 10 presents the cumulative projected radial distributions of 24 μm undetected K+A galaxies, 24 μm detected K+A galaxies, and 24 μm detected non-K+A galaxies in clusters and groups. We found that the spatial distributions of K+A galaxies, whether they are detected at 24 μm or not, are similar in clusters. We performed K-S tests on the radial distributions of these different samples. In clusters, there is a strong suggestion (K-S significance = 0.981) that the 24 μm detected K+A galaxies and their 24 μm undetected K+A counterparts are drawn from the same distribution. On the contrary, the 24 μm detected K+A galaxies and 24 μm detected non-K+A galaxies do not have similar radial distributions, with a K-S likelihood of only 0.068 of being the same. The radial distributions in clusters suggest that 24 μm detected K+A galaxies are possibly associated with 24 μm undetected K+A galaxies but not with other 24 μm detected galaxies.

Zoom In Zoom Out Reset image size

Among all 24 μm detected K+A galaxies, the average IR luminosity for those with and without interaction features is 3.1 × 10¹¹ L_☉ and 1.3 × 10¹¹ L_☉, respectively. For 24 μm detected non-K+A galaxies, the average IR luminosity is 2.2 × 10¹¹ L_☉. The average IR luminosity of 24 μm detected, interacting K+A galaxies suggests elevated star formation activity in the interacting K+A galaxies, and the SFR drops as the interaction feature fades away. These 24 μm detected K+A galaxies may be younger merger-induced K+A galaxies, whose SFR has not yet dropped to an undetectable level. This phase should last at most a few hundred million years. In Clusters A and B, it takes ∼1 Gyr for a galaxy with a velocity of 800 km s⁻¹ to traverse ∼1 R_vir. Taking projection effects into account, the short-lived 24 μm detected K+A phase is hard to spatially distinguish from the parent K+A galaxy sample. The D_n(4000) indices and EW(Hδ) of composite 24 μm detected and 24 μm undetected K+A galaxies (Table 2) are also suggestive that both populations have truncated star formation (Le Borgne et al. 2006; Lemaux et al. 2012), where the 24 μm detected population is younger, indicated by its stronger EW(Hδ) and smaller D_n(4000).

Previous observations have also tested the idea that K+A galaxies are merger descendants. For example, Wild et al. (2009) compared K+A galaxies drawn from the VIMOS VLT DEEP Survey (Le Févre et al. 2005) with numerical simulations and showed that K+A galaxies are consistent with being merger remnants that underwent a strong starburst within the last few hundred million years. Studies on the kinematics and spatial distributions of stellar populations in early-type K+A galaxies with integral field unit spectroscopy also favor K+A galaxies as already coalesced gas-rich mergers (Pracy et al. 2009, 2012). These studies suggested that K+A galaxies are from gas-rich mergers, which induces likely strong starbursts. However, Dressler et al. (2013) claimed that passive galaxies are the principal progenitors of post-starburst galaxies, as they found that the ratio between passive and post-starburst galaxies stays nearly constant across a wide range of environments from isolated galaxies to cluster cores at z ∼ 0.5, which indicates a tight correlation between these two populations. They suggested that the dominant route of producing post-starburst galaxies is rejuvenated passive galaxies experiencing minor mergers or accretion.

With our data, we are not able to directly measure the strength of the past starburst in the K+A galaxies, but their morphologies can give us a clue about the progenitors. In the Cl 1604 supercluster, 13 of 43 K+A galaxies are late type, about one-third of the population. Among the late-type K+A galaxies, nine have interaction features. This simple observation shows that late-type galaxies constitute a non-negligible portion of the whole K+A population, but we cannot rule out the route from rejuvenated passive galaxies for at least some of the population for the remaining sample.

The stellar mass of each population can give some more insight. In Section 4.1 we found that, on average, the stellar mass of a typical K+A galaxy is similar to that of a red sequence galaxy and ∼50%(∼ 10¹⁰ M_☉) more than that of a blue-cloud galaxy. This result means that if an average blue-cloud galaxy turns into a K+A galaxy, its stellar mass should increase significantly in a short period of time. This increase can be achieved by a major merger but not by a minor merger. A major merger can produce a large amount of new stars rapidly, and the stellar mass of the merged galaxy is naturally nearly doubled by combining two progenitors with similar sizes. This scenario is only the case for blue-cloud K+A galaxies. Because the stellar masses of red sequence galaxies are similar to those of K+A galaxies, passive galaxies can experience a minor merger to create a K+A and do not have to merge with another massive galaxy.

The [O ii] λλ3727,3729 line is a popular star formation indicator because it is accessible by optical spectroscopy from the local universe out to z ∼ 1.5, allowing direct comparisons between samples over a wide range of cosmic time. However, [O ii] emission has been shown to be a flawed star formation indicator. [O ii] emission from LINERs/Seyferts can be as strong as that in star-forming galaxies. Therefore, setting an upper limit on [O ii] line strength tends to misidentify a LINER/Seyfert-harboring passive galaxy as a star-forming galaxy (Yan et al. 2006; Lemaux et al. 2010; Kocevski et al. 2011b). On the other hand, [O ii] emission in a dusty star-forming galaxy would be highly attenuated, with the resulting EW([O ii]) as low as that of a truly quiescent galaxy (Poggianti & Wu 2000). These potential problems are often not addressed because there is a shortage of usable information to determine the sample incompleteness and contamination. For the Cl 1604 supercluster, we are able to address these issues thanks to the extensive ancillary data.

5.3.1. LINER/Seyfert Activity

Lacking LINER/Seyfert diagnostics for the whole spectroscopic sample, i.e., the ability to make a BPT or pseudo-BPT diagram (Baldwin et al. 1981; Stasińska et al. 2006; Juneau et al. 2011), we estimated the number of LINER/Seyfert-harboring K+A galaxies based on a study of 17 [O ii]-emitting absorption-line-dominated galaxies in Cl 1604 presented in Lemaux et al. (2010). Their absorption-line-dominated spectra indicate that there is little ongoing star formation, so the [O ii] emission may come from other ionizing sources. From near-IR spectroscopy, Lemaux et al. (2010) found that about half of these [O ii]-emitting absorption-line-dominated galaxies exhibit EW([O ii])/EW(Hα) ratios higher than the typical observed value for star-forming galaxies. Furthermore, a majority of these galaxies have [N ii] λ6584 to Hα emission-line ratios consistent with a LINER/Seyfert origin, so the emission in many of these galaxies may be contributed by a LINER/Seyfert. Thus, these galaxies would have been categorized as K+A galaxies if not for the presence of a LINER/Seyfert. A strong color dependence is also observed: red [O ii]-emitting, absorption-line-dominated galaxies more likely harbor LINERs/Seyferts than blue ones.

We estimate the number of K+A galaxies that would be excluded due to [O ii] emission from a LINER/Seyfert in the following manner. From our spectroscopic sample, we first selected galaxies with (1) [O ii] emission; (2) absorption features such as Ca ii H, Ca ii K, and Balmer absorption lines; (3) no other emission features, such as Balmer lines; and (4) EW(Hδ) ⩾3 Å. These galaxies are candidate genuine post-starburst galaxies but were not classified as such because of [O ii] emission produced by a LINER/Seyfert. From the 17 galaxies in Lemaux et al. (2010), we calculated a color-dependent probability of an [O ii]-emitting, absorption-line-dominated galaxy being LINER/Seyfert-harbored for each 0.5 mag color bin. Applying this color-dependent probability, we derived the estimated number of LINER/Seyfert-harboring galaxies in our candidate sample. These galaxies with strong Hδ absorption and LINER/Seyfert activity are possibly missing from our K+A sample because LINERs/Seyferts contribute part of the [O ii] emission.

Figure 11 plots the estimated numbers of these Hδ-strong, LINER/Seyfert-harboring galaxies along with the O5 and O3 samples in each system. Because we cannot quantify the amount of [O ii] emission contributed from the LINER/Seyfert, this value should be taken as an upper limit on the number of missed K+A galaxies. In most of the systems, the number of LINER/Seyfert-harboring galaxies is comparable to that of the O3 sample and roughly half to two-thirds of the O5 sample. That is to say, K+A samples selected by means of EW([O ii]) may suffer from incompleteness up to ∼50%. Although it is derived from a limited sample size, this result at z ∼ 0.9 is similar to what has been found for lower-redshift SDSS galaxies (Yan et al. 2006).

Zoom In Zoom Out Reset image size

The exception is Cluster A, in which the number of LINER/Seyfert-harboring galaxies is only half of the O3 sample and one-third of the O5 sample. Because the estimate was made from a limited number of galaxies with LINER/Seyfert diagnostics, we cannot draw a firm conclusion whether or not this deviation is real. However, if Cluster A truly behaves differently from other systems, it suggests that the incompleteness of [O ii]-selected K+A samples could depend on the properties of their parent systems. Thus, any comparison of the K+A prevalence among different samples should consider this possibility. For example, if we assume that these LINER/Seyfert-harboring galaxies are all post-starburst galaxies, the total post-starburst fraction in Cluster A will still be higher than in Cluster B, but the difference becomes smaller. Future multi-object IR spectroscopy will be needed for post-starburst galaxy studies at high redshift in order to verify the effect of LINER/Seyferts and their possible environmental dependence.

5.3.2. Dust Obscuration