THE EVOLUTION AND ENVIRONMENTS OF X-RAY EMITTING ACTIVE GALACTIC NUCLEI IN HIGH-REDSHIFT LARGE-SCALE STRUCTURES

The Cl1604 supercluster at z ≈ 0.9 is one of the largest structures studied at high redshift. It consists of at least 10 clusters and groups and spans 100 h⁻¹₇₀ Mpc along the line of sight and 13 h⁻¹₇₀ Mpc in the plane of the sky (Lubin et al. 2000; Gal & Lubin 2004; Gal et al. 2008; Lemaux et al. 2009). The massive member clusters Cl1604+4304 and Cl1604+4321 were first discovered in the optical survey of Gunn et al. (1986). The proximity of the clusters suggested that they were components of a larger structure. Further wide field imaging has revealed 10 distinct red galaxy overdensities, suggesting the existence of a supercluster (Lubin et al. 2000; Gal & Lubin 2004; Gal et al. 2008). Spectroscopic observations have confirmed four of the overdensities to be clusters with velocity dispersions in excess of 500 km s⁻¹, while four others were confirmed to be poor clusters or groups with dispersions in the range 300–500 km s⁻¹ (Postman et al. 1998, 2001; Gal et al. 2005, 2008).

The two most massive clusters in Cl1604 have associated diffuse X-ray emission. Cl1604+4304 and Cl1604+4314, hereafter Clusters A and B, have measured bolometric X-ray luminosities of 15.76 ± 1.48 and 11.64 ± 1.49 × 10⁴³ h⁻²₇₀ erg s⁻¹ and X-ray temperatures of 3.50^+1.82_{− 1.08} and 1.64^+0.65_{− 0.45} keV, respectively (Kocevski et al. 2009a). While these values place Cluster A on the σ–T curve for virialized clusters, Cluster B is well off from it, suggesting that Cluster A is relaxed while Cluster B is not. All other clusters have only an upper limit on their bolometric luminosity of 7.4 × 10⁴³ h⁻²₇₀ erg s⁻¹ (Kocevski et al. 2009a).

While many galaxies in Cl1604 have substantial [O ii] emission, near-infrared spectroscopy has shown that a significant portion of this emission is due to contributions from low-ionization nuclear emission-line regions (LINERs) and Seyferts (Lemaux et al. 2010). Also, Kocevski et al. (2011a) studied 24 μm selected galaxies in and around three clusters and three groups in Cl1604 using the Multiband Imaging Photometer for Spitzer (MIPS; Rieke et al. 2004) and found evidence for recent starburst activity and an infalling population. Analysis of the morphologies using the Advanced Camera for Surveys (ACS; Ford et al. 2003) on the Hubble Space Telescope (HST) revealed that many of these 24 μm bright galaxies were disturbed, indicating mergers and interactions were likely responsible for starburst activity.

We refer the reader to Kocevski et al. (2009a), Gal et al. (2008), and Lemaux et al. (2011) for more details on the data processing, supercluster properties, and observations.

The Cl0023+0423 structure at z ≈ 0.84, hereafter Cl0023, was also discovered as an overdensity in the optical survey of Gunn et al. (1986). The structure was later observed by Oke et al. (1998) using the Low-Resolution Imaging Spectrograph (LRIS; Oke et al. 1995) on the Keck 10 m telescope, where the overdensity was resolved into two structures. Further study has shown that the structure consists of four merging galaxy groups separated by approximately 3000 km s⁻¹ in radial velocity and ∼0.25 Mpc on the plane of the sky (Lubin et al. 1998, 2009). The constituent groups have measured velocity dispersions within 1.0 h⁻¹₇₀ Mpc of 480 ± 170, 430 ± 70, 290 ± 80, and 210 ± 30 km s⁻¹ (Lubin et al. 2009). Simulations suggest that the groups will merge to form a cluster of mass ∼5 × 10¹⁴ M_☉ within ∼1 Gyr (Lubin et al. 1998).

Lubin et al. (2009) found Cl0023 to have a large blue population, with 51% of the galaxies bluer than their red galaxy color–color cut, down to an i'-band magnitude of 24.5. Spectroscopic analysis found that ∼80% of the galaxies had measurable [O ii] emission, which, because of the large blue population, is most likely due to ongoing star formation.⁵

We refer the reader to Lubin et al. (2009) and Kocevski et al. (2009c) for more details on the supergroup properties and observations.

The Cl1324 supercluster is an LSS at z ≈ 0.76. The two most massive clusters in the structure, Cl1324+3011 at z = 0.76 and Cl1324+3059 at z = 0.69, were first discovered in the optical survey of Gunn et al. (1986). Because of the proximity of the clusters on the sky and in redshift space, this structure was chosen for the ORELSE survey to investigate the possible existence of structure in the field. Wide-field imaging has revealed a total of ten clusters and groups, detected through red galaxy overdensities, and, despite extensive spectroscopy, only four have been spectroscopically confirmed as constituent clusters or groups (see R. R. Gal et al. 2012, in preparation).

Cl1324+3011 was previously studied in Lubin et al. (2002, 2004), where a velocity dispersion of 1016⁺¹²⁶_{− 93} km s⁻¹ and a temperature of 2.88^+0.71_{− 0.49} keV, using XMM-Newton, were measured for the cluster. According to these measurements, the cluster does not fall close to the σ–T curve for virialized clusters, which would imply that it is not well relaxed. New Chandra results for Cl1324 are presented in N. Rumbaugh et al. (2012, in preparation), including new X-ray temperatures for Cl1324+3011 and Cl1324+3059. In addition, we present here updated velocity dispersions for these two clusters of 930 ± 120 and 870 ± 120 km s⁻¹, respectively. The new measurements place Cl1324+3011 closer to the σ–T curve for virialized clusters, only offset by ∼1σ. While the older measurements suggested the cluster was not relaxed, the new measurements are more consistent with virialization. Similar to Cl1324+3011, Cl1324+3059 is offset from the curve, but still by less than 1σ.

The photometric and spectroscopic observations of the Cl1324 supercluster will be covered in full in R. R. Gal et al. (2012, in preparation). In this paper, we present redshift histograms of the full structure and velocity dispersions for the four confirmed groups and clusters, as well as the Chandra observations.

The X-ray-selected cluster RXJ1821.6+6827, hereafter RXJ1821, at z = 0.82, was the highest redshift cluster discovered in the ROSAT North Ecliptic Pole (NEP) survey, where it is also referred to as NEP5281 (Gioia et al. 2003; Henry et al. 2006). Using XMM-Newton data, the cluster was found to have slightly elongated diffuse X-ray emission with a measured bolometric luminosity of 1.17^+0.13_{− 0.18} × h⁻²₇₀ erg s⁻¹ and a temperature of 4.7^+1.2_{− 0.7} keV (Gioia et al. 2004). The same study measured a velocity dispersion of 775⁺¹⁸²_{− 113} km s⁻¹ using 20 cluster members. Later analysis by Lubin et al. (2009) used 40 galaxies within 1 Mpc to measure a velocity dispersion of 926 ± 77 km s⁻¹. Redshift histograms of RXJ1821 are characteristic of a single, isolated structure, although a small kinematically associated group has been detected to the south (Lubin et al. 2009). While the temperature and dispersion measurements place the cluster near the σ–T relation for virialized clusters, the elongated X-ray emission could be indicative of still ongoing formation of the cluster.

Lubin et al. (2009) measured a blue fraction of only 24% for RXJ1821, down to a magnitude limit of i' = 24.5. They found a population dominated by massive, old (formation epoch of z_f ∼ 2–3) galaxies, along with fainter galaxies that were quenched more recently. They also found that 36% of the galaxies had detectable [O ii] emission. Near-IR spectroscopy of a subset of these [O ii] emitting galaxies suggests that some of the emission is due to LINER or AGN activity (Lemaux et al. 2010).

We refer the reader to Lubin et al. (2009) for more details on the data processing, cluster properties, and observations.

The z = 0.69 cluster RXJ1757.3+6631, hereafter RXJ1757, was discovered as part of the ROSAT NEP survey, where it is also identified as NEP200 (Gioia et al. 2003). Gioia et al. (2003) found the structure to have an X-ray luminosity of 8.6× 10⁴³ h⁻²₇₀ erg s⁻¹ in the 0.5–2.0 keV band. The structure is dominated by a single, large cluster. In this paper, we present a velocity dispersion, redshift histograms, and analysis of X-ray point sources for RXJ1757, none of which have been previously published.

Ground-based optical imaging data were obtained with the Large Format Camera (LFC; Simcoe et al. 2000) on the Palomar 5 m telescope. Observations were taken using the Sloan Digital Sky Survey r', i', and z' filters. The 5σ point source limiting magnitudes for the five fields ranged from 25.5–25.1, 25.0–24.5, and 23.6–23.3 in the r', i', and z' bands, respectively.

Cl1604 was also imaged using ACS. The HST imaging for Cl1604 consists of 17 ACS pointings designed to image 9 of the 10 galaxy density peaks in the field. Observations were taken using the F606W and F814W bands. These bands roughly correspond to broadband V and I, respectively.

Our photometric catalog is complemented by an unprecedented amount of spectroscopic data. For this part of the study, we used the Deep Imaging Multi-Object Spectrograph (DEIMOS; Faber et al. 2003) on the Keck II 10 m telescope. In addition, Cl1604 and RXJ1821 have some LRIS coverage (see Oke et al. 1998; Gal & Lubin 2004; Gioia et al. 2004). DEIMOS has a wide field of view (169 × 50), high efficiency, and is able to position over 120 targets per slit mask, which makes the instrument ideal for establishing an extensive spectroscopic catalog. We targeted objects down to an i'-band magnitude of 24.5. On DEIMOS, we used the 1200 line mm⁻¹ grating, blazed at 7500 Å, and 1'' slits. These specifications create a pixel scale of 0.33 Å pixel⁻¹ and an FWHM resolution of ∼1.7 Å, or 68 km s⁻¹. The central wavelength was varied from structure to structure and sometimes between different masks for the same field. Central wavelengths for the spectroscopic observations for the five fields and the approximate spectral coverages are displayed in Table 2. When more than one central wavelength was used per field, a range is given. Total exposure times for the observations are in the range of 1–4 hr per mask.

Spectroscopic targets were chosen based on color and magnitude. The number of spectroscopic targets in each field is shown in Table 3. Redshifts were determined or measured for all targets and given a quality flag value, Q, where Q = 1 indicates that we could not determine a secure redshift, Q = 2 means a redshift was obtained using features that were only marginally detected, Q = 3 means one secure and one marginal feature were used to calculate the redshift, and Q = 4 means at least two secure features were used. Those sources determined to be stars were given a flag of −1. See Gal et al. (2008) for more details on quality flags and the spectral targeting method. For our analysis, redshifts with Q = −1, 3, 4 were deemed satisfactory, and the number of such sources in each field is shown in Table 3.

Spectroscopic data have been previously presented for the Cl1604 supercluster, Cl0023, and RXJ1821 as part of the ORELSE survey. We present new data for each of these structures, as well as for the Cl1324 supercluster and RXJ1757. We include ten DEIMOS masks for Cl1324, with exposure times ranging from 6635 s to 10,800 s, and four DEIMOS masks for RXJ1757, with exposure times ranging from 7200 s to 14,730 s. We include a total of 18 spectroscopic masks for Cl1604, 6 more than what were included in Gal et al. (2008).⁶ We include nine total masks for Cl0023, four more than in Kocevski et al. (2009c). We include three masks for RXJ1821, one more than in Lubin et al. (2009). Many of the new targets were optical counterparts to X-ray sources. From our measured redshifts with Q = 3 or 4 from the new masks, we find 114 new galaxies in Cl1604 within 0.84 < z < 0.96, 104 new galaxies in Cl0023 within 0.82 < z < 0.87,⁷ and 5 new galaxies in RXJ1821 within 0.80 < z < 0.84.

All X-ray imaging of the clusters was conducted with the Advanced CCD Imaging Spectrometer (ACIS) of the Chandra X-ray Observatory, using the ACIS-I array (PI: L. M. Lubin).⁸ This array has a 169 × 169 field of view. Some of the five structures were imaged with one pointing and some with two, but every pointing had the same approximate total exposure time of 50 ks. Cl0023, RXJ1821, and RXJ1757 were each imaged with one pointing of the array. Cl1604 and Cl1324, with angular sizes in excess of 20', were observed with two pointings each. For Cl1604, the two pointings are meant to cover as much of the structure as possible, and there is a small overlap (∼30 arcmin²). For Cl1324, the two pointings are centered near the two largest and originally discovered clusters Cl1324+3011 and Cl1324+3059. There is an approximately 13' gap between the north and south pointings.

In this paper, we present new Chandra data for Cl1324, RXJ1821, and RXJ1757.

The reduction of the data was conducted using the Chandra Interactive Analysis of Observations 4.2 software (CIAO; Fruscione et al. 2006). Each observation was filtered by energy into three bands: 0.5–2 keV (soft), 2–8 keV (hard), and 0.5–8 keV (full). Data were checked for flares using dmextract and the Chandra Imaging and Plotting System (ChIPS) routine lc_clean. Exposure maps were created using the routine merge_all. For vignetting correction, exposure maps were normalized to their maximum value, then images were divided by this normalized exposure map.

To locate point sources, the routine wavdetect was run on each observation, without vignetting correction, using wavelet scales of 2^i/2 pixels, with i ranging from 0 to 8. A threshold significance of 10⁻⁶ was used, which would imply fewer than one spurious detection per ACIS chip, which has dimensions of 1024 × 1024 pixels (0492 per pixel). However, this assumes a uniform background, which is almost certainly not the case. To measure realistic detection significances, we instead used photometric results explained below. Point source detection was carried out on images in each of the soft, hard, and full bands separately. For Cl1604 and Cl1324, wavdetect was run on each of the two pointings separately. Output object positions from the three different bands were cross-correlated to create one final composite list for each field.

We carried out follow-up photometry on the point sources. Circular apertures containing 95% of the flux were created for each point source using the point-spread function (PSF) libraries in the Chandra calibration database. For Chandra, the PSF depends on both energy and off-axis angle. For the soft and hard bands, respectively, we used the PSF libraries for energies of 1.497 and 4.510 keV. Photometry was carried out on the vignetting-corrected images in the soft and hard bands. Background counts for each source were calculated in annuli with inner and outer radii of 1.2 × R₉₅ and 2.4 × R₉₅, where R₉₅ is the radius of the circular aperture containing 95% of the flux. Since only 95% of the flux is enclosed, net counts calculated from the apertures were multiplied by 1/0.95 to recover all the counts. Full band counts were calculated by summing those from the soft and hard bands. The results of the photometry were used to calculate detection significances for the sources in each of the three bands using

$$\begin{equation} \sigma = C/(1.0+\sqrt{0.75 + B}), \end{equation} \tag{ 1 }$$

where C is the net photon counts from the source and B is the background counts (Gehrels 1986). X-ray sources with significances <2σ were rejected as spurious. With ∼75% of accepted sources having detection significances >3σ, and the remaining ∼25% with detection significances between 2σ and 3σ, we expect a spurious detection rate of ≲ 1.5%, based on a normal distribution.

Due to the low number of photons observed for many sources, we opted to normalize a power-law spectral model to the net count rate of individual point sources to determine fluxes. We assumed a photon index of γ = 1.4, which is the approximate slope of the X-ray background in the hard band (Tozzi et al. 2001; Kushino et al. 2002). Count rates were calculated by dividing net counts by the nominal exposure time at the aim point of the appropriate observation. The galactic neutral hydrogen column density was calculated at the aim point of each observation using the Colden tool from the Chandra proposal toolkit, which uses the data set of Dickey & Lockman (1990). Hydrogen column densities and derived net count rate to unabsorbed flux conversion factors for each field are summarized in Table 4. Conversion factors were determined separately for the different pointings of Cl1604 and Cl1324, but did not differ to the three significant figures listed in the table.

Table 4. Count Rate to Flux Conversion Factors and H i Column Densities

Structure	$N_{{\rm H\,\mathsc{i}}}$a	Count Rate to	Count Rate to
	(1020cm−2)	Flux Conversion,	Flux Conversion,
		Soft Bandb	Hard Bandb
Cl1604	1.2	6.12	22.2
Cl0023	2.7	6.37	22.3
Cl1324	1.1	6.16	22.2
RXJ1821	5.6	8.23	22.3
RXJ1757	4.0	6.60	22.2

Notes. ^aGalactic neutral hydrogen column density, using the data set of Dickey & Lockman (1990). ^bX-ray net count rate to unabsorbed flux conversion factor, in units of 10⁻¹² erg cm⁻².

Download table as:  ASCIITypeset image

To search for AGNs within the individual clusters, we matched X-ray and optical sources. In order to increase our completeness, the input to the matching included all point sources detected by wavdetect, regardless of significance in any of the three bands. We used the maximum likelihood ratio technique described in Kocevski et al. (2009a), which was developed by Sutherland & Saunders (1992) and also used by Taylor et al. (2005) and Gilmour et al. (2007). Our technique is similar to Kocevski et al. (2009a), but with a few key differences. The main statistic calculated in each case is the likelihood ratio (LR), which estimates the probability that a given optical source is the genuine match to a given point source relative to the arrangement of the two sources arising by chance. The LR is given by the equation

$$\begin{equation} {\rm LR}_{i,j} = \frac{w_i \exp \big({-}r_{i,j}^2/2\sigma _j^2\big)}{\sigma _j^2}. \end{equation} \tag{ 2 }$$

Here, r_{i, j} is the separation between objects i and j, σ_j is the positional error of object j,⁹ and w_i = 1/n(< m_i) is the inverse of the number density of optical sources with magnitude brighter than m_i. The inclusion of the latter quantity is designed to weight against matching to fainter optical objects. However, in our analysis, we found that this particular weighting, used by Kocevski et al. (2009a), of 1/n(< m_i) gave too much favor to bright objects, even when they were much farther from an X-ray object than a faint source. We adjusted the weighting to w_i = n(< m_i)^−1/2. We found that this change did not have a large overall effect but changed some borderline cases where a bright object with a large separation from the point had been chosen over a dimmer, much closer object. This includes one case, which prompted the adjustment, where spectroscopy showed that an M-type star was matched instead of a probable AGN, even though the M star was three times farther from the X-ray source.

For each field, except Cl1604, n(< m_i) was measured using i' magnitudes from our LFC catalogs. For Cl1604, ACS data were also available, but these observations did not cover the entire field. All objects were matched to the LFC catalogs. When possible, objects were also matched using the F814W magnitude from the ACS catalogs and matches to ACS objects took precedence over matches to LFC objects.

From the LR, we use Monte Carlo simulations to derive the probability that a given match is genuine. We ran 10,000 trials for each X-ray object. In each trial, the object's position was randomized and the LR was calculated based on nearby optical sources. The LR for a given X-ray object to optical counterpart pairing, LR_{i, j}, was compared against the distribution of LRs from the 10,000 Monte Carlo trials. We calculated the reliability as

$$\begin{equation} R_{i,j} = 1 - \frac{N({{\rm LR}}_j > {{\rm LR}}_{i,j})}{{\rm 10{,}000}}, \end{equation} \tag{ 3 }$$

where N(LR_j > LR_{i, j}) is the number of matches, to any optical source, across all 10,000 trials for that X-ray object with LR greater than LR_{i, j}. R_{i, j} can be interpreted as the probability that optical source i is the true match of X-ray source j, in the case of only one optical candidate. When there are multiple candidates, we used the method of Rutledge et al. (2000) to calculate the probability that optical source i is the true match of X-ray source

$$\begin{eqnarray} &&P_{i,j} = \frac{R_{i,j}\prod _{k \ne i}^N (1-R_{k,j})}{S}. \end{eqnarray} \tag{ 4 }$$

The probability that no optical source is the true match is

$$\begin{eqnarray} &&P_{{{\rm none},j}} = \frac{\prod _{k = 1}^N (1-R_{k,j})}{S}, \end{eqnarray} \tag{ 5 }$$

where N is the total number of optical candidates and S is a normalization factor defined so that ∑^N_{i = 1}(P_{i, j}) + P_{none, j} = 1.

For an X-ray source with a single optical counterpart, a match was considered genuine if P_{none, j} < 0.15. For X-ray sources with multiple optical counterparts, a genuine match was chosen if ∑_iP_{i, j} > 0.85 (which is equivalent to P_{none, j} < 0.15) and P_{i, j} > 4∑_{k ≠ i}P_{k, j} for any one object i. If the first condition was true, but the second was not, all objects with P_{i, j} > 0.2 were considered as matches. In subsequent sections, only one optical counterpart was considered for each X-ray source. In almost all cases, the highest probability match was used. However, in several cases, spectra of the primary and secondary matches indicated that the secondary match was an AGN, and thus more likely to be a genuine match. Note that our threshold is a deviation from Kocevski et al. (2009a). They used P_{none, j} < 0.2 instead of 0.15. We decided to use the more stringent threshold of 0.15, which has been used by others (Mann et al. 2002; Taylor et al. 2005), to better limit the number of false matches. The more restrictive cutoff omitted ≲ 10 sources per field. We determined this threshold through visual inspection of potential matches. This calibration entailed determining at what approximate level of P_{none, j} most matches visually seemed spurious. However, optical candidates above our threshold were also visually scrutinized (≲ 3% of the total), and some were accepted after this inspection where we felt the matching algorithm had failed. Note that setting a threshold for genuine matches is not entirely objective, and a precedent has been set for accomplishing this with visual inspection (e.g., Mann et al. 1997). In Table 3, we list, for each field, the number of X-ray sources detected at >3σ (>2σ) in one of the three bands, as well as the number of those sources matched to optical counterparts.

The Cl1604 supercluster, Cl0023, and RXJ1821 have all been studied previously as part of the ORELSE survey (see Section 2 for individual references), although we have gathered new data on each, as described in Section 3.1. While individual clusters in Cl1324 have been studied, the properties of the supercluster as a whole have not. In this section we do a preliminary exploration of this structure, which will be covered more thoroughly in an upcoming paper (R. R. Gal et al. 2012, in preparation). The cluster RXJ1757 was studied only as a part of the ROSAT NEP survey (Gioia et al. 2003), in little detail. Here, we present new redshift histograms of these last two structures derived from our ORELSE data.

4.1.1. Cl1324

Figure 1(a) shows all confirmed redshifts in the spatial vicinity of Cl1324. We can see two peaks in the histogram, at z ≈ 0.695 and z ≈ 0.755. These peaks coincide with the two largest clusters in the structure, Cl1324+3011 and Cl1324+3059. From the distribution of red galaxies, we find ten overdensities in the supercluster, some of which can be observed in the redshift histogram. So far, we have confirmed four clusters and groups to be constituents, shown in Table 5, with coordinates, redshifts, and measured velocity dispersions given. Additional multiobject spectroscopy to confirm the nature of the other red galaxy overdensities is planned.

Zoom In Zoom Out Reset image size

Table 5. Cl1324: Properties of Confirmed Groups and Clusters

Name	Alt.	R.A.	Decl.	σa
	Name	(J2000)	(J2000)
Cl1324+3011	A	13 24 48.7	+30 11 48	930 ± 120
Cl1324+3059	I	13 24 50.5	+30 58 19	870 ± 120
Cl1324+3013	B	13 24 21.5	+30 13 10	820 ± 240
Cl1324+3025	C	13 24 01.8	+30 25 05	220 ± 100b

Notes. ^aVelocity dispersion measured in km s⁻¹, within 1 Mpc. ^bDispersion calculated using only eight galaxies.

Download table as:  ASCIITypeset image

4.1.2. RXJ1757

Figure 2 shows CMDs for all five fields. All spectroscopically confirmed supercluster/cluster members are shown. Squares indicate the confirmed X-ray AGNs within each structure, which are analyzed in Section 6. The red sequence for each field is delineated by dotted lines. Red-sequence fits for each field were calculated using a linear fitting and σ-clipping technique. First, a fit to a linear model, of the form

$$\begin{equation} C = C_0 + m \times B \end{equation} \tag{ 6 }$$

where C is either r' − i' or F606W–F814W and B is either i' or F814W, was carried out on member galaxies within a chosen magnitude and color range using a χ² minimization (Gladders et al. 1998; Stott et al. 2009). The fit was initialized with a color range chosen "by eye" to conform to the apparent width of the red sequence of the structure. The magnitude bounds were defined as the range where the photometric errors were small ($\sigma _{i^{\prime },{\rm F814W}} \lesssim 0.05$). After the initial fit, colors were normalized to remove the slope. The color distribution was then fit to a single Gaussian using iterative 3σ clipping. At the conclusion of the algorithm, the boundaries of the red sequence were defined by a 3σ offset from the center, except for Cl1604 and Cl1324. For the two superclusters, the color dispersion was inflated due to the large redshift extent of these structures, and 2σ offsets were used to achieve reasonable boundaries. For every field except Cl1604, the LFC r' − i' CMDs are shown. For Cl1604, ACS data were available and were used in place of LFC data because of their superior precision. Note, however, that two of the AGNs in Cl1604 are outside our ACS pointings, so that our analysis using ACS data only includes eight AGNs in the Cl1604 supercluster.

Zoom In Zoom Out Reset image size

While all the structures show a substantial number of galaxies on the red sequence, there are large differences in the blue populations. Qualitatively, we observe a lower blue fraction in the CMDs of RXJ1821 and RXJ1757 than in those of Cl0023 and Cl1604. This is quantified in Table 6, where the blue fraction¹⁰ for all confirmed cluster/supercluster members with i' or F814W magnitudes brighter than 23.5 are displayed in the first column. We can see that Cl0023 and Cl1604 have the bluest galaxy populations, while RXJ1757 has the highest fraction of galaxies on the red sequence of all the fields. Since this could be due to the low completeness of spectroscopic coverage in this field, we attempted to make corrections with two approaches: (1) using our efficiency of spectroscopically confirming structure members to estimate the total number of member galaxies and (2) correcting using measurements of the background galaxy density. All efforts to statistically estimate the true blue fraction yielded large errors making the measurements highly uncertain. Instead, we created a blue fraction measurement metric that could be compared between fields (see also Lubin et al. 2009). The different fields have a varying number of spectral masks, with several different spectroscopic priorities. However, the first several masks, excluding those designed to specifically target X-ray matched sources, have similar priorities for choosing targets.¹¹ Therefore, we chose to compare blue fractions only among sources in the first two masks for each field, since these sources should represent similar populations. We chose two masks because this is the number of masks for RXJ1757 excluding those where objects matched to X-ray point sources were preferentially targeted. We confirmed that this sampling is representative of the entire galaxy population by recreating composite spectra using only the first two spectral masks. Since there are no large differences between the average spectral features, the method should be accurate. The results of this comparison are displayed in the second column of Table 6. However, we did not calculate a blue fraction for Cl1604 in this way, for two reasons. First, there are a total of 24 spectroscopic masks for Cl1604, from several different telescopes. Choosing which ones to include is difficult and it may be impossible to select a population congruous with any of the other fields in this manner. Second, our spectroscopy for Cl1604 is relatively complete, so we are confident in the blue fraction measured down to F814W = 23.5.

Examining the results, we can see the same color hierarchy in the structures for both methods of measuring the blue fraction, with RXJ1757 having the smallest fraction and Cl1604 and Cl0023 the largest. These large blue fractions, in particular those for Cl1604 and Cl0023, are consistent with the Butcher–Oemler effect (Butcher & Oemler 1984). We also note that the isolated X-ray-selected clusters, RXJ1757 and RXJ1821, are the reddest structures, suggestive of a more advanced, dynamically relaxed evolutionary state. The color hierarchy suggests a similar ranking of galactic star formation in the five structures, which we can explore with our spectroscopic data.

Using our spectroscopic data, we examine the typical star formation history of the galaxies in each structure. We formed composite spectra by co-adding the individual spectra of galaxies within each structure, according to the method of Lemaux et al. (2009, 2011). We analyze these spectra in terms of two important features relevant to star formation: the [O ii] and Hδ lines. The Hδ absorption is indicative of a population of A and B stars, which disappears ∼1 Gyr after the cessation of star formation within a galaxy, due to the lifetime of A stars (Poggianti & Barbaro 1997). If star formation is ongoing, a population of O stars, which have weaker hydrogen features, can dominate the continuum and wash out this absorption line. Infilling can also occur from Balmer emission from H ii regions. The [O ii] emission line has been used as an indicator of star formation, especially as a proxy for the Hα emission line at higher redshifts when Hα has shifted out of the optical range (Poggianti et al. 1999). However, recent analysis using near-IR spectroscopy of sources from the Cl1604 supercluster and RXJ1821 has compared [O ii] and Hα emission, finding that a significant portion of [O ii] emission can come from LINER- or Seyfert-related processes (Lemaux et al. 2010; Kocevski et al. 2011a). These results are supported by those of Yan et al. (2006), albeit using a lower redshift sample. In light of this, caution must be exercised when interpreting [O ii] measurements. For an additional diagnostic, we measure the D_n(4000) strength which is an indicator of mean stellar age (Kauffmann et al. 2003).

Figure 3(a) shows [O ii] versus Hδ equivalent widths for the composite spectra of members of the five structures. The dotted line represents the average spectral properties for a cluster population composed of various fractions of "normal" star-forming and quiescent galaxies (Dressler et al. 2004), based on data from the Two-Degree Field (2dF) Galaxy Redshift Survey (Colless et al. 2001). Asterisks on this line represent a cluster population composed of (from left to right) 20%, 40%, 60%, 80%, and 100% star-forming galaxies. For a cluster whose average galaxy lies above this line, the Hδ line is too strong to be produced by normal star formation, requiring some contribution from starbursting or post-starbursting galaxies. The dashed lines enclose 95% of the normal star-forming galaxies (Oemler et al. 2009). The shaded regions, which are based on the spectral types of Dressler et al. (1999) and Poggianti et al. (1999), denote the region of this phase space inhabited by (starting from the upper right and moving counterclockwise) starburst (dark blue), post-starburst (green), quiescent (red), and normal star-forming galaxies (light blue). Examining the positions of the structures in our sample on this plot, we can see that RXJ1757 and RXJ1821 have mostly quiescent populations, with each cluster having ≲ 20% normal star-forming galaxies. Because RXJ1757 is offset from the 2dF line, there may be some contribution from post-starburst galaxies, but the fractional contribution is low. The stronger [O ii] emission in the Cl0023 and Cl1604 composite spectra suggests that these structures have a higher fraction (≳ 40%) of continuously star-forming galaxies. These structures have larger blue fractions than the others, so it is unlikely that the increased EW([O ii]) in their average spectra is due to LINER processes, which are primarily associated with red-sequence galaxies. The EW(Hδ) measured from the Cl0023 and Cl1604 composite spectra are significantly in excess of the 2dF line, suggesting a substantial contribution from starbursting or post-starburst galaxies. Cl1324 is in an intermediate range, with ∼30% normal star-forming galaxies and an observed EW(Hδ) for its galaxy population smaller than that of Cl0023 and Cl1604.

Zoom In Zoom Out Reset image size

Our conclusions based on the [O ii] and Hδ lines are supported by the corresponding D_n(4000) measurements (see Table 7). These results are illustrated in Figure 3(b), where average measurements of EW(Hδ) and D_n(4000) are plotted for the five structures. For comparison, we also indicate ranges of EW(Hδ)–D_n(4000) phase space spanned by four different Bruzual & Charlot (Bruzual 2007) models for various times after the starburst. The four models include a single burst (τ = 0.01) and a secondary burst of 20%, 10%, and 5% by mass which occurs 2 Gyr after the initial burst. All models have solar metallicity and are corrected for extinction using E(B − V) =0.25 and a Calzetti et al. (2000) extinction law. Because Bruzual & Charlot models only incorporate stellar light, emission infill corrections were made for all EW(Hδ) measurements using relationships between Hα and [O ii] taken from Yan et al. (2006) and Hα and Hδ from Schiavon et al. (2006). Although these corrections are not perfect (e.g., see the measurement for Region 1), these values should be accurate to within ±0.5 Å. We can see that RXJ1757 and RXJ1821 have the largest continuum break strengths, indicating that they possess the oldest average stellar populations of the five structures. The average galaxy in these structures has had 1–3 Gyr since its last starburst, according to the Bruzual (2007) model. The other structures have smaller average D_n(4000) measurements, consistent with the results of Figure 3(a) showing larger fractions of star-forming galaxies and younger galaxy populations. According to the Bruzual (2007) model, the average galaxy in Cl1324, Cl0023, and Cl1604 has had a progressively shorter time since the last starburst. Altogether, these spectral results parallel what was found with the blue fractions, in that the reddest structures are also the ones with the most evolved stellar populations and the lowest fraction of star-forming galaxies.

Despite our extensive spectroscopic sample, there are too few X-ray AGNs in any individual cluster (and even supercluster) to draw statistical conclusions. Hence, we divide our structure sample into two categories. The first contains Cl0023 and Cl1604 which have the highest level of ongoing star formation and starburst activity, as shown in the preceding sections. The second category, consisting of Cl1324, RXJ1757, and RXJ1821, contains structures whose member galaxies are typically quiescent or forming stars at a lower rate. We refer to these two categories as "unevolved" and "evolved," respectively, as a means of describing their typical galaxy populations. While we acknowledge that the terms "more evolved" and "less evolved" would be more appropriate, we adopt the less accurate denominations for brevity. In addition, these terms do not necessarily imply differing levels of cluster dynamical evolution or even a clear temporal sequence from one category to the other. We note that, while some clusters or groups may not fit well with the global characteristics of their parent supercluster (i.e., Cluster A of Cl1604), we cannot examine all of the AGNs on a cluster-by-cluster basis. As we will discuss in Section 6.1, many of the AGNs in the superclusters and the supergroup are not associated with any one cluster or group. So, when analyzing these AGNs, we take the parent supercluster or supergroup as a whole.

The particular segregation of our structure sample is motivated by the clear distinction between the structures shown in Figure 3. The abundant star formation in the unevolved structures suggests the presence of a large gas reservoir in many of their member galaxies. We might expect that this same gas is available for AGN fueling. Conversely, the typical galaxy in the evolved sample has likely consumed most of the available gas in prior star formation episodes, leaving less fuel for the AGNs. In the following sections, we examine whether the properties of the AGNs in the two categories reflect this distinction, and what we can learn about the relationship of star formation to AGN activity in LSSs.

Examining the spatial distribution of AGNs located within each cluster can give insight into what processes triggered their nuclear activity. In Figures 7, 8, and 9, we show the spatial distributions on the sky and redshift distributions of the five structures studied here. The AGNs are marked in red, and their positions and characteristics are given in Table 8.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Of particular note is the lack of AGNs in dense cluster centers. Indeed, we find 19% of AGNs in cluster cores, defined as being within a projected distance of 0.5 Mpc to the nearest cluster or group.¹³ An additional 27% lie on the outskirts of clusters (projected distances between 0.5 and 1.5 Mpc), and a majority (>50%) of AGN host galaxies lie more than 1.5 Mpc in projected distance from the nearest cluster or group.

These results are consistent with previous work studying the spatial distribution of AGNs, in that the AGNs tend to be located outside of clusters or in their outskirts. First, the distribution of AGNs in Cl1604, previously studied by Kocevski et al. (2009a), is consistent with the other four structures here. Our results are also consistent with those of Gilmour et al. (2007), who found that AGNs in the A901/902 supercluster at z ≈ 0.17 tend to avoid the densest areas. Several other studies have also found that X-ray AGNs tend to reside in regions of moderate density similar to group environments, up to z ∼ 1 (Hickox et al. 2009; Silverman et al. 2009a). It is thought that regions of intermediate density, such as the outskirts of clusters, are the most conducive to galaxy–galaxy interactions because of their elevated densities, compared to the field, but relatively low velocities (Cavaliere et al. 1992). Since we find more of the AGNs in these areas, this lends support to the theory that mergers or tidal interactions are one of the main instigators of AGN activity. Since only ∼20% inhabit dense cluster cores, processes which preferentially occur in these regions, such as ram pressure stripping, are probably not responsible for triggering AGNs in cluster galaxies. However, the association between AGNs and these regions could also be related to higher gas availability in galaxies farther from cluster cores.

Although we find that most of the X-ray AGNs do not reside in the cluster cores, a number of studies have measured the fraction of cluster galaxies that host AGNs (e.g., Martini et al. 2007). Therefore, we attempt to measure the AGN fraction for the individual clusters within the five structures in our sample. Because we are limited by the spectral completeness of our sample, we make a composite measurement of the most massive, well-sampled clusters: Cl1604A, Cl1604B, Cl1604C, Cl1604D, Cl1324A, Cl1324I, and RXJ1821. We compare to the results of Martini et al. (2007) for low-redshift clusters (0.06 < z < 0.31), who used galaxies with M_R < −20, within approximately 2000 km s⁻¹ of the mean redshift of cluster members for each system, and within the field of view of their Chandra observations, which ranged from 1.2 to 4.5 Mpc in width. To approximate these criteria, we adopt a magnitude cutoff that roughly corresponds to $M_{i^{\prime }/{\rm F814W}}=-19$ and use galaxies within 1 Mpc and Δz = 0.01 of the cluster centers and redshifts. We measure the combined AGN fraction for the seven clusters listed above to be 0.012  ±  0.007. This is consistent with the results of Martini et al. (2007), who measured f_A(M_R < −20; L_x > 10⁴² h⁻²₇₀ erg s⁻¹) ≈ 1%. We note, however, that a number of arbitrary definitions went into our measurement. In addition, our spectroscopy of optical counterparts to X-ray sources is significantly incomplete, in contrast to Martini et al. (2007). Correcting for this incompleteness contributes the largest source of error to our measurement. Because of the large uncertainties, we refrain from drawing any conclusions from our measurement.

Figure 2 presents CMDs, which are described in Section 4, for all five fields. Confirmed AGN members of the structures are shown with blue squares. While the AGNs in Cl1324, RXJ1821, and RXJ1757 preferentially reside within the bounds of, or very close to, the red sequence, the AGN hosts in Cl0023 and Cl1604 are more spread out. Previous work, including Kocevski et al. (2009a) and a number of wide-field surveys (Sánchez et al. 2004; Nandra et al. 2007; Georgakakis et al. 2008; Silverman et al. 2008), has found an association between AGN activity in galaxies and the transition onto the red sequence (possibly for a second time) in the green valley. While these galaxies could be evolving from the blue cloud onto the red sequence, it is also possible that they could have moved down off the red sequence after a tidal interaction or merger and are evolving back (Menanteau et al. 2001; Kocevski et al. 2009a). We note, however, that some studies using mass-selected samples have that AGN hosts have found a color distribution more similar to that of normal galaxies (Silverman et al. 2009b; Xue et al. 2010).

In addition, Cardamone et al. (2010), using a sample of galaxies with redshifts 0.8 < z < 1.2, have found that many green-valley AGN hosts are dust-reddened blue cloud members, so that AGN host colors acquire the bimodality apparent in the general galaxy population. However, Rosario et al. (2011) have also examined the impact of extinction on the colors of AGN hosts and did not find a significant impact for galaxies in the redshift range 0.8 < z < 1.2, although bimodality may be introduced at higher redshifts. To address this issue for our study, we are planning to implement spectral energy distribution (SED) fitting to evaluate the impact of extinction on the broadband colors of our sample. Preliminary results from SED fitting of the Cl1604 hosts suggest that extinction levels in our sample are not as drastic as those presented by Cardamone et al. (2010). It is also possible that AGN host colors are contaminated by the AGNs themselves. However, this is unlikely because (1) almost all AGN hosts in our sample have rest-frame X-ray luminosities below the quasi-stellar object (QSO) level of 10⁴⁴ erg s⁻¹ (see Section 6.4) and (2) Kocevski et al. (2009b) found that, in Cl1604, AGN hosts with blue cores did not have a rising blue continuum indicative of QSO activity. Therefore, we proceed to investigate the AGN association with the transition zone and to explore differences in the evolutionary states of their host galaxies in each field by examining color offsets of the AGN hosts from the red sequence.

Histograms of offsets from the center of the red sequence are shown in Figure 10. The first two panels present offsets in terms of color. In the top panel, only the structures where LFC data were used are shown, which is every field except Cl1604. The middle panel shows only Cl1604, for which we used ACS colors. In the bottom panel, all five structures are shown, with normalized offsets. In order to compare the ACS and LFC data, we scale by the red-sequence width, W_RS. We define W_RS as the distance from the center of the red-sequence fit to its boundary (see Section 4.2 and Figure 2). On this plot, AGN hosts on the red sequence will then be located between −1.0 and 1.0.

Zoom In Zoom Out Reset image size

With red-sequence offsets, we can quantitatively examine the green valley. In Figure 10(b), we plot a histogram of RS offsets, measured from the ACS data, for the AGNs in the Cl1604 supercluster. For comparison, we overplot a scaled distribution of RS offsets for all spectroscopically confirmed supercluster members with ACS photometry. In the scaled histogram, we can clearly see an area of reduced number density between the red sequence and blue cloud. For Cl1604, the green valley can be approximated as the region −3W_RS < ΔC < −W_RS, where ΔC is the offset of an AGN host from the center of the red sequence. Only ∼17% of all confirmed supercluster members with ACS data fall within this region. However, five out of eight of the Cl1604 AGN hosts with ACS data reside within it. While it is unclear how well this definition of the green valley extends to LFC data, because of larger photometric errors, we can see in Figure 10(c) that 36% of all host galaxies have −2W_RS < ΔC < −W_RS and 60% are in the range −3W_RS < ΔC < −W_RS.

While many galaxies in both the evolved and unevolved structures lie in the "green valley" region, the percentage of AGNs on the red sequence is somewhat higher in the evolved structures compared to the unevolved structures, with 30% and 20%, respectively. Examining Figure 10, we can see the distribution of AGN hosts in evolved structures is clustered closer to the red sequence, while in the unevolved structures, this distribution has a large tail extending into the blue cloud. Indeed, none of the AGN host galaxies in the evolved structures have ΔC < −3W_RS, whereas four of the X-ray AGN hosts in the unevolved structures have red-sequence offsets below this limit. Although these results are suggestive (and unaffected by our inclusion of the <3σ X-ray sources), the two distributions are not statistically different based on the K-S test.

Morphological analysis by Kocevski et al. (2009a) has shown that ∼67% of the X-ray AGNs in Cl1604 have had recent mergers or tidal interactions, which could fuel star formation through starburst events. More recent mergers or interactions are one possible explanation for some of the color differences that we see between the AGN host galaxies in the evolved and unevolved structures. In particular, we find that nine of the AGN host galaxies are members of a kinematic close pair with a relative line-of-sight velocity of Δv ⩽ 350 km s⁻¹ and projected physical separation (on the plane of the sky) of Δr_p ⩽ 70 h⁻¹₇₀ kpc (e.g., Lin et al. 2007). Two are in Cl0023, five in Cl1604, one in Cl1324, and one in RXJ1821 (see Table 8). Those AGN hosts in pairs include three of the four galaxies with the largest red-sequence offsets (i.e., the bluest), all of which are members of the unevolved structures. Based on their z' magnitudes or measured stellar masses (in the case of the Cl1604 members; see Lemaux et al. 2011), seven out of the nine kinematic pairs have flux or mass ratios of ≳ 50%, implying a major merger scenario.

The differences in color and, perhaps, merger activity are likely related to the increased level of star formation and starburst activity in the unevolved compared to evolved structures (see Section 4.3). To explore the connection between the AGN and star formation history, we can use our high-resolution spectroscopy to examine the average spectral properties of their host galaxies.

We measure the average spectral properties of the AGN host galaxies in the five ORELSE structures using three composite spectra: one comprised of six AGNs from Cl0023,¹⁴ one comprised of all ten AGNs from Cl1604, and one comprised of the ten AGNs from the combined fields of Cl1324, RXJ1821, and RXJ1757. The last spectrum combines all the evolved structures, necessitated by the low number of AGNs in RXJ1821 and RXJ1757 compared with the other structures. These three composite spectra are shown in Figure 11, and measurements of spectral features are listed in Table 9.

Zoom In Zoom Out Reset image size

Table 9. Spectral Measurements of AGN Hosts

Bin	Number of	EW([O ii])a	EW(Hδ)a,b	Dn(4000)c	Ca ii H+H/Ca ii Kd
	Galaxies	(Å)	(Å)
Cl0023	6	−11.20 ± 0.21	0.36 ± 0.39	1.05 ± 0.003	0.42 ± 0.07
Cl1604	10	−7.09 ± 0.17	2.14 ± 0.19	1.22 ± 0.005	1.58 ± 0.13
Evolved Sys.e	10	−12.92 ± 0.21	4.51 ± 0.15	1.44 ± 0.006	1.02 ± 0.02

Notes. ^aMeasured using bandpasses from Fisher et al. (1998). ^bCorrected for infill from emission, based on EW([O ii]). We use the method described in Kocevski et al. (2011a), although we made a slightly different choice for the relationship between EW(Hα) and EW([O ii]) appropriate for Seyfert galaxies. ^cMeasured using bandpasses from Balogh et al. (1999). ^dThe equivalent widths are measured by fitting a line to the background and a Gaussian to the absorption feature. ^eIncludes Cl1324, RXJ1821, and RXJ1757.

Download table as:  ASCIITypeset image

First, we can see that the AGN hosts in all three groupings have substantial [O ii] emission. In Cl1604 and the evolved structures, most of this emission is probably from the AGNs, rather than star formation. Six of the ten AGN host galaxies in our Cl1604 sample were analyzed by Kocevski et al. (2011a) using the Keck II Near-infrared Echelle Spectrograph (NIRSPEC; McLean et al. 1998). For five out of the six targets, Kocevski et al. (2011a) found that the [N ii]/Hα flux ratio was too large for a normal star-forming galaxy, which implies that AGNs are the dominant contributor to [O ii] emission (Kauffmann et al. 2003). The Cl1324, RXJ1821, and RXJ1757 structures have AGNs mostly near or on the red sequence. In addition, we found in Section 4 that star formation is low in all three structures (≲ 30% of all galaxies are star forming). Because of this low star formation rate, and because Lemaux et al. (2010) found that ∼90% of the red [O ii] emitters are dominated by LINER/Seyfert emission, it is likely that most of the [O ii] emission from the AGN host galaxies in the evolved structures comes from the AGNs as well.

Deciphering the origin of the [O ii] emission in the AGN host galaxies in Cl0023 is not as straightforward. While some AGN host galaxies in Cl0023 are on the red sequence, the structure also has the bluest host galaxies in our sample. Also, as discussed in Section 4, there is significant star formation in the general population of Cl0023. While this is also true of Cl1604, the NIRSPEC results of Kocevski et al. (2011a) showed most of the [O ii] emission in the AGN host galaxies in that structure comes from the AGNs. However, we do not have any near-IR spectroscopic data for the Cl0023 AGN, so we must use other means to determine the emission source. The average [Ne iii]/[O ii] ratio of the Cl0023 X-ray AGN hosts is 0.429, typical of type-2 AGN emission, emission from metal-poor star formation, or a superposition of the two processes (Nagao et al. 2006; Trouille et al. 2011). Note that Cl1604 and the evolved structures have values of the [Ne iii]/[O ii] ratio of 0.549 and 0.229, respectively, also typical of type-2 AGN emission. Combining this result with the blue colors of the Cl0023 hosts and the high fraction of star-forming galaxies, it is likely that the observed [O ii] emission of the AGN hosts in Cl0023 is due to a combination of normal star formation and type-2 AGN activity.

Examination of the Balmer features reveals further insight into the star formation histories of the AGN host galaxies. Specifically, based on a single stellar population model, the EW(Hδ) rises quickly from zero after a starburst in a galaxy, peaks after about 300–500 Myr, and then declines back to approximately zero at ∼1 Gyr after the burst (Poggianti & Barbaro 1997). Infill can complicate the interpretation of this feature, so care must be taken when measuring the line strength. For the composite spectrum of the Cl0023 AGN hosts, the equivalent width of Hδ, attempting to correct for infill, is consistent with zero. However, we observe strong emission from other Balmer features, EW(Hβ) = −7.81 Ź⁵ and EW(Hγ) = −2.55 Å, suggesting that emission infill has a significant effect on the measured EW(Hδ). This infill could be due to emission from H ii regions, emission from AGNs or some other LINER processes, or from continuum emission produced by O stars. Since we observe Hβ and Hγ in emission, it is unlikely that O stars are solely responsible for the observed EW(Hδ). The Balmer emission lines observed in the Cl0023 AGN hosts are not broad¹⁶ and are quite strong. In the average type-2 AGNs, a large fraction of the Balmer emission originates from star formation (Kauffmann et al. 2003), which suggests ongoing star formation in the Cl0023 hosts. Compared to Cl0023, Hγ emission from the Cl1604 hosts is low, with EW(Hγ) = −0.69 ± 0.17, which is consistent with a lower level of star formation in the supercluster.¹⁷ Balmer lines for the evolved structures are in absorption, EW(Hβ) = 0.48 ± 0.12 and EW(Hγ) = 1.41 ± 0.16, consistent with an even lower level of star formation. These results combined with the earlier result analyzing the average [Ne iii]/[O ii] ratios of the AGN hosts strongly indicates that star formation is occurring in the Cl0023 galaxies coevally with AGN activity, while less star formation activity is occurring in the other structures.

The Hδ equivalent widths, combined with measurements of the 4000 Å break, suggest that starbursts have occurred more recently in the average Cl1604 AGN host compared to those in the evolved structures. Larger values of D_n(4000) indicate a more passive galaxy, with an older average stellar population, which could mean that more time has passed since the cessation of star formation (Balogh et al. 1999; Kauffmann et al. 2003). The EW(Hδ) measured from the Cl1604 composite is 2.14 ± 0.19 Å, roughly half of that from the evolved structures composite. Similarly, the Cl1604 composite has a lower value for D_n(4000) than those in the evolved structures, indicating that the average Cl1604 host is more actively star forming or has a younger stellar population on average. This result is supported by the bluer colors of the Cl1604 AGN hosts compared to those in the evolved structures. All of these results imply the average Cl1604 host has had a starburst more recently than the average AGN host in the evolved structures.

The AGN host composite spectrum for Cl0023 has a particularly low D_n(4000) measurement (see Table 9). Since we found that this composite spectrum (which excluded one broad-line source) was consistent with type-2 AGNs, the AGNs themselves should not contribute most of the blue continuum. This points to a stellar source, particularly O and B stars. We would then expect significant star formation in the Cl0023 hosts within the last 10–100 Myr, as indicated by the other spectral features as well.

Related to the D_n(4000) measurement, the Ca ii H+H and the Ca ii K lines also provide information on star formation. For F, G, and K stars, the ratio of these lines is constant, while the Ca ii H+H/Ca ii K ratio increases for A and B stars as the overall Ca ii strength decreases and the H strength increases (Rose 1985). The Ca ii H+H/Ca ii K ratio is 1.58 ± 0.13 for Cl1604 and 1.02 ± 0.02 for the evolved structures (see Table 9), consistent with the evolved structures having (on average) older stellar populations. We do see a decrease in the overall strengths of both Ca ii lines in the average spectrum of the Cl0023 host galaxies relative to the other structures; however, we actually measure a dramatic decrease in the ratio for Cl0023 (0.42  ±  0.07), the opposite of what is expected from a population of A and B stars. The most likely explanation is significant H emission, which would be in concert with the other observed Balmer emission. The H emission could be coming from some combination of AGNs and H ii regions, which would support previous conclusions about the level of activity in the Cl0023 hosts.

Altogether, the composite spectra of the AGN hosts in all three bins of Figure 11 show that the average host galaxy has ongoing star formation or has had star formation within the last ∼1 Gyr. However, the hosts in Cl1604 and the evolved structures each have, on average, less ongoing star formation than Cl0023, as evidenced by larger values of D_n(4000) and the absence of the Balmer emission that is observed in Cl0023. These differences suggest a progression in the temporal proximity of the last starburst event, with the hosts in Cl0023 having significant ongoing star formation characteristic of a current starburst, to those in Cl1604 and the evolved structures each having successively more time since the last significant starburst event. We have confirmed that none of our results based on the composite spectra are changed by removing the four lowest significance (<3σ) X-ray sources from our sample, with most spectral measurements remaining the same within the errors.

With our analysis of the composite spectra of AGN host galaxies in the different structures, we can compare the average properties of these hosts with the average properties of all spectroscopically confirmed galaxies within the same structures. In Section 4, we found that the galaxy populations in the evolved structures were largely quiescent, with little or no contribution from starburst or post-starburst galaxies. In contrast, the populations in the unevolved structures were comprised of large fractions of star-forming galaxies, with a more significant contribution from starburst or post-starburst galaxies. When comparing these results to the average spectral properties of the AGN hosts, we find that, in all cases, the average AGN host galaxy has a younger stellar population than the average galaxy in the parent structure, irregardless of the evolved or unevolved classification. This result holds even when comparing to member galaxies outside the dense cluster cores (>0.5 Mpc), where the vast majority of X-ray AGNs reside.

The most prominent difference comes from the evolved structures where their average AGN host galaxy has significantly larger EW(Hδ) and smaller D_n(4000) than the average structure member, indicative of a post-starburst galaxy with a substantial star formation event within the last ∼1 Gyr. Such galaxies make a small contribution to the overall population in the evolved structures, which is largely quiescent. While we do observe clear differences between the spectra of the AGN hosts, with those in Cl0023 having significant ongoing star formation to those in the evolved structures having the most time since the last significant starburst event, these differences are not nearly as pronounced as the differences between the average galaxy in each structure. This suggests that AGN activity has a common origin associated with current or recent star formation.

In this section, we explore the differences between the structures and AGN host properties based on the X-ray luminosities of the confirmed AGNs. We calculate rest-frame luminosities for X-ray point sources with known redshifts. K-corrections were carried out using the power-law spectral models for sources, with a photon index of γ = 1.4, described in Section 3. Luminosities are measured in the X-ray soft, hard, and full bands. A histogram of full-band rest-frame luminosity,¹⁸ binned by evolved and unevolved structures, is shown in Figure 12. The luminosity distributions in the soft and hard bands are similar to the one shown.

Zoom In Zoom Out Reset image size

In the left panel of Figure 12, we can see that the AGNs in the unevolved structures have higher X-ray luminosities than those in the more evolved structures. K-S tests show that the distributions of the two bins are different at the 99% level in each of the three bands. This statistically significant result is independent of our inclusion of the four low-significance (<3σ) X-ray sources.

While 10 out of 17 AGNs in the unevolved structures have full band luminosities above L_x = 10^43.3 h⁻²₇₀ erg s⁻¹, there are no AGNs above this limit in Cl1324, RXJ1757, or RXJ1821. We follow a Bayesian approach, using Poisson statistics to calculate the likelihoods, to estimate the probability of finding no high L_x AGNs in the evolved structures (N^E_det = 0) given the detection rate in the unevolved structures. Specifically, we calculate P(N^E_det = 0|N_det^UE, N^UE_targ, N_targ^E). Here, N^UE_det is the number of high L_x AGNs that are spectroscopically confirmed members in the unevolved structures, and N^UE_targ and N^E_targ are the total number of high L_x AGNs that were targeted for spectroscopy in the unevolved and evolved structures, respectively. Based on this calculation, the probability of finding no high L_x AGNs in the evolved structures is only 0.25%. This result is likely related to the smaller fractions of blue cloud galaxies (and overall more quiescent populations) in Cl1324, RXJ1757, or RXJ1821 and, thus, the unavailability of large gas reservoirs.

At the faint end, there are four sources in the evolved structures below 10^42.5 h⁻²₇₀ erg s⁻¹, all of which are members of Cl1324. However, the X-ray source counts, in all five fields, are significantly incomplete at these luminosity levels with only ∼5% of optically matched X-ray sources below this limit, most of which are <3σ detections. Therefore, we cannot say anything definitive about the lack of faint sources in the unevolved structures; however, if we remove the four low-luminosity sources, the difference between the luminosity distributions in the evolved and unevolved samples is still significant at a 95% level according to the K-S test. The reasonably high significance is clearly due to the lack of high-luminosity sources in the evolved structures.

6.4.1. Relation to Host Galaxy Color

6.4.2. Relation to Average Spectral Properties

To explore the origin of the variations observed in Figure 12, we examine the average spectral properties of the host galaxies within the four distinct regions. Specifically, the sample is split by X-ray luminosity at L_x = 10^43.3 and red-sequence offset, scaled by RS width, to delineate regions containing high L_x blue cloud (Region 1), high L_x (Region 2) and low L_x (Region 3) green valley, and low L_x red-sequence (Region 4) host galaxies. In Figure 13, we plot the measured EW(Hδ) versus D_n(4000) from the spectral composites in the four regions. For comparison, we also plot post-starburst temporal regimes derived from four Bruzual (2007) models, described in Section 4.3.

Zoom In Zoom Out Reset image size

The small D_n(4000) and Hδ in emission indicate that the high L_x blue hosts are coeval with the starburst or ongoing star formation. As we examine galaxies in Regions 2 to 4 going from the high to low L_x hosts in the green valley to the low L_x hosts in the red sequence, the time since the burst gets progressively larger. While there is some degeneracy between time since burst and burst strength, it is clear that the low L_x green-valley hosts are either (1) further along since the burst than their high L_x counterparts or (2) had a weaker initial burst which could explain their lower X-ray luminosities as less gas would likely be funneled to the center. Our results are robust to removing the four lowest significance (<3σ) X-ray sources, as well as the four X-ray sources below L_x = 10^42.5 erg s⁻¹, where in both cases we are highly incomplete.

The most striking results from this spectral analysis are, first, that the average AGN host in every region is either in the process of having a starburst or has had one within last ≲ 1 Gyr. This global finding clearly demonstrates the close connection between starburst and AGN activity as normal star formation does not typically produce the Hδ values seen in these host galaxies. Second, we do not detect high X-ray luminosity, young (as indicated by time since starburst) galaxies in the evolved structures. This result implies that the entire galaxy population in these structures (certainly in the isolated, X-ray-selected ones) are more advanced, suggesting that the peak of gas consumption, seen in both star formation and AGN activity, occurred at an earlier time.