THE CLUSTER LENSING AND SUPERNOVA SURVEY WITH HUBBLE: AN OVERVIEW

Recent observations suggest that galaxy clusters formed earlier in our universe than in simulated ΛCDM universes. These observations include the detection of perhaps unexpectedly massive galaxy clusters at z > 1 (Stanford et al. 2006; Eisenhardt et al. 2008; Jee et al. 2009, 2011; Huang et al. 2009; Rosati et al. 2009; Papovich et al. 2010; Schwope et al. 2010; Gobat et al. 2011; Foley et al. 2011) and the finding that some clusters at intermediate redshift (z ∼ 0.3) have denser cores than clusters of similar mass produced in simulations (Broadhurst et al. 2008; Broadhurst & Barkana 2008; Oguri et al. 2009; Richard et al. 2010; Sereno et al. 2010; Zitrin et al. 2011a). While the evidence to date for early cluster growth is suggestive, possible explanations include departures from the Gaussian initial density fluctuation spectrum or higher levels of dark energy in the past, so-called Early Dark Energy (EDE; Fedeli & Bartelmann 2007; Sadeh & Rephaeli 2008; Francis et al. 2009; Grossi & Springel 2009). If a significant quantity of EDE (for example, Ω_DE ∼ 0.1 at z = 6) suppressed structure growth in the early universe, then clusters would have had to start forming sooner to yield the numbers we observe today. These scenarios remain allowable within current observational constraints as described in the above papers, although some non-Gaussian models can be ruled out by using the cosmic X-ray background measurements (Lemze et al. 2009). The CLASH data permit significant advances to be made toward supporting or rejecting observational evidence for early cluster growth by measuring core densities for a larger, less biased sample of clusters.

In cosmological simulations, cold-dark-matter-(CDM)-dominated halos of all masses consistently evolve to have a roughly "universal" density profile that steepens with radius. Functional forms that fit such a profile well include the "NFW" profile (Navarro et al. 1996, 1997) and the Einasto/Sérsic profile (Sérsic 1963; Einasto 1965; Navarro et al. 2004, 2010). Furthermore, each simulated halo's core density is related to the background density of the universe at the halo's formation time. Halos that form later, including the most massive galaxy clusters, are found to have the least dense cores in a relative sense. Determining the relationship between the shape and depth of a halo's gravitational potential and its total mass as a function of time thus provides fundamental constraints on structure formation.

In practice, the relative core densities, or "concentrations," are measured (both in simulated and observed halos) as c_vir = r_vir/r₋₂, a ratio between the virial radius and the inner radius at which the density slope of the fitted profile is isothermal (ρ∝r⁻²). Analyses of gravitational lensing spanning such a large range of radii allow one to map the (primarily dark) matter profiles of observed halos and measure their concentrations.

DM profiles are best mapped in cluster cores using strong-lensing analysis of multiband HST imaging. The lensing-based mass profile mapping is extended to the virial radii of clusters using weak-lensing analysis of wider field ground-based multiband imaging, such as from Subaru. Results from Umetsu et al. (2011a, 2011b) for four of the currently best-studied (non-CLASH) clusters are shown in Figure 1.

Zoom In Zoom Out Reset image size

Joint modeling of strong and weak lensing (SL+WL) yields significantly better constraints on concentrations than either probe alone. Quantitatively, Meneghetti et al. (2010) found their joint SL+WL analyses of simulated clusters yield concentrations to ∼11% accuracy, while WL-only and SL-only analyses yielded ∼33% and ∼59% scatters, respectively. Figure 2 demonstrates how SL and WL analyses combine to yield robust constraints on the mass and concentration of CL0024+17 (Umetsu et al. 2010).

Zoom In Zoom Out Reset image size

The best-studied (SL+WL) galaxy clusters to date have been found to have overly high concentrations (dense cores) compared to halos in N-body simulations with similar masses, as shown in Figure 3 (Broadhurst et al. 2008; Oguri et al. 2009; Sereno et al. 2010) and detailed in Table 1. Recent simulations (Prada et al. 2011) yield cluster concentrations that are over 50% higher than previous simulations (e.g., Duffy et al. 2008). This is a product of upturns as a function of both mass and redshift found in these newer simulations, which are not yet understood. Similarly, clusters have also been found to have somewhat larger than expected Einstein radii, a direct and particularly accurate measure of the projected mass in a halo's core (Broadhurst & Barkana 2008; Richard et al. 2010; Zitrin et al. 2011a, 2011b).

Zoom In Zoom Out Reset image size

Unfortunately, the best-studied clusters to date have also been among the strongest gravitational lenses known. Such lensing-selected clusters are highly biased toward halos with high concentrations, both intrinsically and as projected on the sky due to halo elongation along the line of sight (Hennawi et al. 2007; Oguri & Blandford 2009; Meneghetti et al. 2010, 2011). These biases are estimated to lead to systematically higher concentrations by as much as 50% or more. However, such a bias is insufficient to account for the discrepancy between the observations and the predictions, that is until a very recent analysis of new simulations was performed (see Figure 3 and discussion below).

More robust conclusions require analysis of a larger, unbiased cluster sample. Progress toward this goal has been made by LoCuSS, the Local Cluster Substructure Survey (Smith et al. 2005). A large sample of 165 clusters between 0.15 < z < 0.30 was selected based on X-ray brightness. Strong-lensing analyses of 20 of these based on HST imaging (mostly single-band "snapshots") were presented by Richard et al. (2010). Okabe et al. (2010a) published weak-lensing analyses of 30 LoCuSS clusters, 22 of these being more "secure" based on multiband Subaru imaging, and 9 of these 22 overlapping with the Richard et al. (2010) subset. Subaru images are especially desirable for WL studies as they enable excellent galaxy shape measurements to be performed over a wide area. Multiband imaging is also critical to properly select background galaxies and avoid significantly diluting the weak-lensing signal (and thus the virial mass and concentration) with unlensed foreground galaxies (Medezinski et al. 2007, 2010). Stacked WL-only analyses have been performed on large cluster samples (Johnston et al. 2007; Mandelbaum et al. 2008), but we re-emphasize the need for combining strong + weak lensing analyses in addition to cross-comparisons with mass profile estimates from other techniques. For example, mass concentrations can be measured from X-ray profiles (Buote et al. 2007; Ettori et al. 2010), although these are subject to uncertainties due to assumptions about hydrostatic equilibrium. Importantly, cluster elongation along the line of sight (a potential bias in concentration measurements) can be measured by the combination of lensing and X-ray analysis (Morandi et al. 2011; Newman et al. 2011). The caustic technique (Diaferio & Geller 1997; Rines & Diaferio 2006) is, like lensing-based methods, independent of the dynamical state of the cluster and also provides an important cross check on mass estimates from lensing and gas kinematics. Further discussion of some of the previous results from various methods is given in Comerford & Natarajan (2007), Coe (2010), Rines et al. (2010), and King & Mead (2011).

Reconciliation of high observed concentrations with results from simulations may ultimately come from significantly reducing the observational sample bias along with finding higher concentrations in simulated clusters. Baryons, for example, are currently absent from those cosmological simulations large enough to produce massive clusters. Baryons can result in significant "adiabatic contraction" on galaxy scales, however they constitute a much smaller fraction of the mass on cluster scales. Simulations including baryons show that cluster halo concentrations are likely only varied by ∼10% or so relative to DM-only halos, and that the direction of variation (increase or decrease) is not even clear, depending on the gas physics assumed (Duffy et al. 2010; Mead et al. 2010; King & Mead 2011). Nonetheless, measuring the velocity dispersion of the brightest cluster galaxy (BCG) as an additional constraint on the inner (≲ 80 kpc) mass profile, where its stellar mass is a non-negligible component of the matter distribution, provides for a more thorough mapping of the total mass profile. Some clusters have been shown to have inner mass profiles that are shallower than NFW (Sand et al. 2004, 2008; Newman et al. 2009, 2011). Deviations of cluster mass profiles from NFW or Einasto forms at small and/or large radii may slightly bias concentration measurements (Oguri & Hamana 2011). The use of multiple probes of the matter distribution, as CLASH is designed to do, will enable such biases to be measured and the significance of deviations determined.

More recently, an analysis by Prada et al. (2011) found that clusters in the Bolshoi and MultiDark simulations have concentrations ∼50% higher than clusters in previous simulations. While the robustness of this new result is still being assessed, it raises the possibility that the combination of new observations of an unbiased sample of clusters and new simulations may be able to bridge the concentration gap. We stress here that estimates of the observational bias from previous cluster studies have large uncertainties and likely vary for each cluster. CLASH will determine mass profiles and concentrations for a new cluster sample free of lensing selection bias. As we demonstrate in Section 3.2, CLASH is designed to detect (or rule out) with 99% statistical confidence average deviations of 15% or more from predicted concentrations. However, given the outstanding uncertainties in the expected concentrations, we prefer to recast the problem as follows: CLASH will deliver robust observational concentration measurements for a sample of clusters that simulations will be tasked to reproduce. Ultimately, this will lead to a better calibration of many mass estimation techniques and, consequently, to a better understanding of structure formation on cluster scales and perhaps of our cosmological model as well.

The biggest cosmological surprise in decades came from observations of high-redshift SNe Ia, providing the first evidence that the expansion of the universe now appears to be accelerating (Riess et al. 1998; Perlmutter et al. 1999), and indicating that the universe is dominated by "dark energy." The presence of dark energy has galvanized cosmologists as they seek to understand it. Observations of high-redshift SNe Ia have continued to lead the way in measuring the properties of dark energy (e.g., Riess et al. 2011). The goal for cosmologists now is to measure the equation of state of dark energy, w = P/(ρc²), and its time variation in the hope of discriminating between viable explanations. A departure of the present equation of state, w₀, from −1 or a detection of its variation, ∂w/∂z, would invalidate an innate vacuum energy (i.e., the cosmological constant) as the source of dark energy and would point toward a present epoch of "weak inflation." A difference between the expansion history and the growth history of structure expected for w(z) would point toward a breakdown in general relativity as the cosmic scale factor approaches unity.

HST paired with ACS is a unique tool in this investigation, providing the only means to collect SNe Ia at 1 < z < 1.5, which, in turn, provide the only constraints we have to date on the time variation of w. From the 23 SNe Ia at z > 1 with HST data (Riess et al. 2004, 2007) we have learned: (1) that cosmic expansion was once decelerating before it recently began accelerating, (2) that dark energy, i.e., an energy density with w < 0, was already present during this prior decelerating phase, (3) that SNe Ia at a look-back time of 10 Gyr appear both spectroscopically and photometrically similar to those seen locally, and (4) no rapid change is seen in w(z) and thus no departure is yet seen from the cosmological constant, though the constraint on the time variation remains an order of magnitude worse than on the w₀.

SNe Ia play a central role, not only as distance indicators for cosmography, but also as major contributors to cosmic metal production and distribution. The measurement of high-redshift SN Ia rates is therefore integral to understanding the history of chemical enrichment. The high-redshift rate cannot easily be predicted from the star formation rate because the nature and, hence, the timescales of the process behind the growth of the white dwarf toward the Chandrasekhar mass are not known. The two leading, competing scenarios are accretion from a close binary companion—the single-degenerate scenario (Whelan & Iben 1973; Nomoto 1982), or merger with another white dwarf, following loss of orbital energy and angular momentum by emission of gravitational waves—the double-degenerate scenario (Iben & Tutukov 1984; Webbink 1984). One way to constrain the different progenitor scenarios is to measure the delay-time distribution (DTD) of SNe Ia. This is the distribution of times that elapse between a brief burst of star formation and the subsequent SN Ia explosions. Observations have suggested various different forms for the DTD (e.g., Dahlen et al. 2004, 2008; Mannucci et al. 2006; Pritchet et al. 2008). However, a number of more recent measurements and analyses point to a DTD that is a power law of index ≈ − 1 (Totani et al. 2008; Maoz et al. 2010, 2011; Maoz & Badenes 2010; Brandt et al. 2010; Horiuchi & Beacom 2010; Graur et al. 2011). Specifically, Graur et al. (2011) have shown that such a DTD fits well the measured SN Ia rate out to z ≈ 2. Small sample sizes are the major limiting factor at high redshifts. Large high-z SN samples are therefore needed to resolve the issue, and control possible biases in cosmological studies (due to the evolving SN Ia channel mix).

The CLASH survey uses ACS in parallel with the cluster program to continue the discovery of SNe Ia at 1 < z < 1.5, the objects which tell us about the variation in w. With WFC3 in parallel, CLASH will yield SNe Ia at 1.5 < z < 2.5. Observations in this fully matter-dominated epoch provide the unique chance to test SN Ia distance measurements for the deleterious effects of evolution independent of our ignorance of dark energy (Riess & Livio 2006). Because the SNe Ia are detected when these cameras are in parallel, they are far from the cluster core (∼2 Mpc at the median cluster redshift of z = 0.4) and, hence, the effects of lensing are small (and correctable), making the SNe usable for improving the limits on the redshift variation of the dark energy equation of state. At z < 1, SN Ia distance measurements are most sensitive to the static component of dark energy, w₀. At 1 < z < 1.5, the measurements are most sensitive to the dynamic component, w_a. By z > 1.5, the measurements are most sensitive to evolution if present (e.g., the changing C/O ratio of the donor star), providing the means to diagnose and calibrate the degree of SN Ia evolution in dark energy measurements. Figure 4 shows how variations in the DE equation of state or an evolution in the white dwarf C/O ratio can change the SN Ia distance modulus as a function of redshift.

Zoom In Zoom Out Reset image size

Accounting for the systematic uncertainty introduced by the cosmic star formation history, Graur et al. (2011) predicted the SN Ia rate out to higher redshifts (their Figure 13), which we use here, in conjunction with the CLASH observational parameters, to estimate the number of z ⩾ 1 SNe Ia that will be discovered over the course of this program. These estimates are presented in Table 2 and are the 68% confidence intervals, accounting for both statistical and systematic uncertainties.

One of the most important goals in observational cosmology is to find the first generation of galaxies and understand their roles in the reionization of the intergalactic medium (IGM) between z ∼ 6 and z ∼ 12. Substantial progress has been made in recent years in finding objects at z ∼ 6–7 (Bouwens et al. 2006, 2009; Bunker et al. 2010; Oesch et al. 2010a, 2010b; McLure et al. 2010; Capak et al. 2011), which are, at most, ∼700–900 Myr of age. We wish to find objects at even earlier epochs, in order to understand (1) how the first galaxies were formed through a hierarchical merging process; (2) how the chemical elements were generated and redistributed through the galaxies; (3) how the central black holes exerted influence over the galaxy formation; and (4) how these objects contributed to the end of the "dark ages."

The majority of galaxies at z = 6–7 have been discovered via two methods: (1) deep pencil-beam surveys, such as the Hubble Ultra Deep Field (HUDF; Beckwith et al. 2006) and the Great Observatories Origins Deep Survey (Giavalisco et al. 2004) and (2) degree-size surveys with 10 m class ground-based telescopes, such as the Subaru Deep Field. Both use a color selection to search for "dropout" candidates—galaxies with deep IGM absorption at the wavelengths shortward of the redshifted Lyα break. These efforts have proven successful at finding objects at z < 7, but there is growing evidence for a rapid decrease in the number of candidates at higher redshifts (Iye et al. 2006; Bouwens et al. 2006; Castellano et al. 2010). HST/WFC3 data yield 73 z ∼ 7 and 59 z ∼ 8 candidates (Bouwens et al. 2011). These results are based on ∼500 orbits of HUDF observations. In comparison, >600 galaxies are found at z ∼ 6 (Bouwens et al. 2007; Su et al. 2011). To date, there have been only two or three spectroscopically confirmed galaxies at z > 7 (Lehnert et al. 2010; Ono et al. 2011; Schenker et al. 2011) and one spectrum for a gamma ray burst at z ∼ 8.2 (Salvaterra et al. 2009; Tanvir et al. 2009). The reason, in part, is a lack of photons: at redshift z = 8 and 10, an L* galaxy would have an apparent magnitude in the first NIR detection band of 28.2 and 29.6, respectively.

Gravitational lensing by clusters amplifies the flux of background sources considerably. These cosmic telescopes improve the efficiency of searches for relatively bright high-redshift galaxies (Bouwens et al. 2009; Maizy et al. 2010). In fact, the majority of m < 25.5 AB, z ≳ 6.5 galaxy candidates have been found in cluster fields (e.g., Kneib et al. 2004; Bradley et al. 2008; Zheng et al. 2009; Bradley et al. 2012; Schenker et al. 2011). Interestingly, most of these candidates are found in regions with amplification of ∼10. Furthermore, lens models can be used to help discriminate between highly reddened objects and truly distant, high-redshift objects, as the projected positions of the lensed images are strong functions of the source redshifts.

Luminous z > 7 galaxies are extremely valuable as their spectra can be used to determine the epoch of the IGM reionization. This is because only a tiny fraction of neutral hydrogen is needed to produce the high opacity of Lyα observed at z ∼ 6. The damped Lyα absorption profile that results from a completely neutral IGM (Miralda-Escude 1998) can be measured even at low-spectral resolution. Furthermore, direct measurements of the early star formation rate (via Lyα and Hα emission; Iye et al. 2006) can be derived from the spectra of bright high-z galaxies. CLASH may detect dozens of relatively bright (magnified to m < 26.7 AB) z > 7 galaxies, including some bright enough for spectroscopic follow-up. An NIR survey of a comparable area (100 arcmin²) in unlensed fields would have a ∼10 × lower efficiency. An estimate, albeit a highly uncertain one, of the CLASH high-z detection efficiency enhancement is shown in Figure 5.

Zoom In Zoom Out Reset image size

ΛCDM provides a robust theoretical framework for the evolution of DM halos. However, the formation of galaxies within these halos is governed by complicated baryonic interactions that are difficult to simulate on cosmological scales. The gap in our understanding of the connection between the formation of galaxies and that of their parent halos is illustrated by observations of galaxy "downsizing" (Cowie et al. 1996). ΛCDM predicts that structures form hierarchically, with small halos forming early and later assembling into larger halos. In contrast, massive galaxies appear to have formed at early times, as evidenced by the redshift at which their star formation peaks (e.g., Cowie et al. 1996; Guzman et al. 1997; Brinchmann & Ellis 2000; Juneau et al. 2005) and the old ages of their stellar populations at z = 0 (e.g., Faber et al. 1995; Worthey 1996; Proctor & Sansom 2002; Thomas et al. 2005). The stellar mass density of the universe is also increasing with cosmic time, as galaxies form stars, then shut down their star formation and, ultimately, accumulate on the red sequence (Bell et al. 2004; Faber et al. 2007).

Several open questions remain about the origin and nature of this stellar mass growth. Is it dominated by in situ star formation, or through mergers of existing stellar systems? How does the balance of these two processes change with galaxy mass and cosmic time? In a system dominated by hierarchical assembly, this is akin to asking whether most of the mass accreted by galaxies comes in as gas (e.g., the "cold flows" of Kereš et al. 2005; Dekel & Birnboim 2006) or as stars (e.g., "dry mergers," Khochfar & Burkert 2003; Bell et al. 2004). Theoretical expectations predict that the in situ population should be very centrally concentrated, while an accreted stellar component should extend to a larger radius (Oser et al. 2010). Thus, massive galaxies are predicted to have substantial gradients in the origin of their stars, with the innermost stars having formed in situ and the outer stars largely accreted through merging.

Recent work has demonstrated that massive, passive galaxies already exist at z ∼ 2 (e.g., Trujillo et al. 2006; Kriek et al. 2009) and that these systems are much more compact than local massive galaxies (Trujillo et al. 2006; Longhetti et al. 2007; van Dokkum 2008), suggesting that they must grow significantly in size between z ∼ 2 and z ∼ 0. This process likely happens through "dry" (dissipationless) merging, since such massive galaxies are observed to have old stellar populations locally. There are also substantial populations of massive star-forming galaxies over a similar redshift range. These galaxies tend to exhibit large star-forming "clumps" (Cowie et al. 1995; van den Bergh 1996; Elmegreen et al. 2004a, 2004b), although at least some of these galaxies may form disks with coherent rotation (Genzel et al. 2011).

There are several observational challenges to measuring and characterizing these two modes of growth. It is difficult to resolve substructure in high-redshift galaxies, even from space. Both resolution and signal-to-noise (S/N) considerations limit observations to the upper end of the mass function. Searches that use a small number of imaging filters struggle to identify the redshifts (and therefore the masses) of the targets and can only poorly constrain their spectral energy distributions (SEDs) and stellar populations. Studies that rely on emission lines probe only the assembly of star-forming galaxies.

The CLASH program is well suited for studying galaxy assembly at z < 3. The high magnification and spatial stretching of strongly lensed distant galaxies, coupled with the broad 16-band photometry and resulting photometric redshifts, enables the measurement of star formation rates and stellar ages in each lensed galaxy over many resolution elements. The magnification makes it possible to resolve substructures in high-redshift lensed galaxies further down the galaxy mass function than is possible in unlensed galaxies at similar redshifts. In addition, with CLASH, we can contrast the relative properties of galaxy cores versus their outer regions, and (where they exist) the properties of substructures such as clumps, spiral arms, and ongoing mergers. In the full CLASH sample of 25 clusters, we expect to find several strongly lensed background galaxies per cluster that are suitable for such analyses, providing a sample of 50–100 galaxies with 1 < z < 3.

To date, robust joint SL+WL analyses have only been performed on a small, highly biased sample of five to ten clusters (Table 1 and Figure 3). These clusters were selected for study primarily based on their gravitational lensing strength, which tends to preferentially select systems with higher central matter concentrations. To establish a sample that is largely free of lensing bias, we selected 20 massive clusters from X-ray-based compilations of dynamically relaxed systems. Sixteen of these 20 clusters were taken from the Allen et al. (2008) compilation of massive relaxed clusters. Clusters in our X-ray-selected subsample all have T_x ⩾ 5 keV and exhibit a high degree of dynamical relaxation as evidenced by Chandra X-ray Observatory images that show well-defined central surface brightness peaks and nearly concentric isophotes. The clusters, in general, also show minimal evidence for departures from hydrostatic equilibrium in X-ray pressure maps. Most of the 20 clusters have smooth and only mildly elliptical (〈〉 = 0.19) X-ray emission and a BCG within a projected distance of 23 kpc of the X-ray centroid. No lensing information was used a priori to select them in order to ensure that they are not preferentially aligned along the line of sight, in contrast with a purely lensing-selected sample, where the surface mass density is, on average, biased upward along the line of sight by intrinsic triaxiality (Hennawi et al. 2007; Oguri et al. 2009; Meneghetti et al. 2010, 2011). A handful of the clusters in the CLASH X-ray-selected subset have some evidence for departures from symmetric X-ray surface brightness distributions. These systems are briefly discussed in Section 3.3.

These 20 clusters comprise the primary sample for our DM distribution studies. Existing X-ray data will be used to measure the gas density profile for subtraction from the total mass profile (as applied by Lemze et al. 2008 for A1689), which is necessary when making detailed comparison with the predictions of DM-only simulations. Although these clusters are X-ray selected, one or more lensed arcs are indeed visible in all of the clusters for which sufficiently high-quality imaging was available. This indicates that the X-ray-selected clusters in our sample have Einstein radii in the range 15''–35''. X-ray images from the ACCEPT²⁵ database for these 20 CLASH clusters are shown in Figure 6.

Zoom In Zoom Out Reset image size

In addition, CLASH includes five clusters selected based solely on their exceptional strength as gravitational lenses (large Einstein radii, θ_Ein > 35'' for z_s = 2). Analysis of these clusters will allow us to further quantify the lensing selection bias toward high concentrations. The excellent strong-lensing data yielded by these clusters will enable derivation of some of the highest resolution DM maps which may be obtained (as in Coe et al. 2010). The primary motivation for selecting these five "high-magnification" clusters, however, was to significantly increase the likelihood of discovering very highly magnified high-redshift galaxies (Section 2.3). While most of the CLASH clusters lie behind regions of low Galactic extinction (median E(B − V) = 0.026), there is one cluster, MACS2129.4−0741, with significant amounts of Galactic cirrus located northeast of the cluster core. Because this cluster has one of the larger Einstein radii known, its inclusion in the "high-magnification" sample was deemed worthwhile in spite of the presence of this cirrus.

Figure 7 shows, respectively, the distribution of the projected separation between the BCG and the peak of the X-ray surface brightness and the distribution of the ellipticity of the X-ray emission for the CLASH sample. Ellipticity measurements are from Maughan et al. (2008). The X-ray-selected cluster subsample exhibits, on average, a very small offset between the BCG and peak X-ray flux. The X-ray-selected cluster subsample also exhibits a lower average ellipticity than our high-magnification (Einstein-radii selected) sample but is consistent with the mean ellipticity of the larger Maughan et al. (2008) sample.

Zoom In Zoom Out Reset image size

The 25 clusters in the CLASH program are presented in Table 3 and some of their key X-ray properties are given in Table 4, including the bolometric luminosity (defined for convenience to be between 0.1 and 100 keV), temperature, and cluster-to-solar [Fe/H] ratio, where the solar abundance reference is from Anders & Grevesse (1989). Table 4 also lists the source of the X-ray selection in Column 6. The CLASH sample is drawn heavily from the Abell and MACS cluster catalogs (Abell 1958; Abell et al. 1989; Ebeling et al. 2001, 2007, 2010). The CLASH clusters span almost an order of magnitude in mass (∼5 to ∼30 × 10¹⁴ M_☉). These clusters were also selected to cover a wide redshift range (0.18 < z < 0.90 with a median z_med = 0.40) allowing us to probe the full c(M, z) relations expected from simulations.

Table 4. X-ray Properties of the CLASH Cluster Sample

	kT	(LBol)a	[Fe/H] Ratio	Ellipticityb	Centroid Shiftb	X-ray Morphology
Cluster	(keV)	(1044 erg s−1)	(Cluster/Solar)c		<w >(10−2R500)	Reference
X-ray Selected Clusters:
Abell 209	7.3 ± 0.54	12.7 ± 0.3	0.18 ± 0.09	0.21 ± 0.01	0.55 ± 0.05	1
Abell 383	6.5 ± 0.24	6.7 ± 0.2	0.59 ± 0.10	0.04 ± 0.01	0.18 ± 0.02	2
MACS0329.7−0211	8.0 ± 0.50	17.0 ± 0.6	0.48 ± 0.10	0.15 ± 0.03	1.38 ± 0.13	2,3
MACS0429.6−0253	6.0 ± 0.44	11.2 ± 0.5	0.34 ± 0.11	0.21 ± 0.02	0.38 ± 0.03	2,3
MACS0744.9+3927	8.9 ± 0.80	29.1 ± 1.2	0.37 ± 0.10	0.11 ± 0.02	1.41 ± 0.13	2,3
Abell 611	7.9 ± 0.35	11.7 ± 0.2	0.30 ± 0.07	⋅⋅⋅	⋅⋅⋅	2
MACS1115.9+0129	8.0 ± 0.40	21.1 ± 0.4	0.32 ± 0.06	⋅⋅⋅	⋅⋅⋅	2,3
Abell 1423	7.1 ± 0.65	7.8 ± 0.2	0.23 ± 0.13	⋅⋅⋅	⋅⋅⋅	4
MACS1206.2−0847	10.8 ± 0.60	43.0 ± 1.0	0.25 ± 0.09	⋅⋅⋅	⋅⋅⋅	3,4
CLJ1226.9+3332	13.8 ± 2.80	34.4 ± 3.0	0.36 ± 0.25	0.10 ± 0.03	1.30 ± 0.11	2
MACS1311.0−0310	5.9 ± 0.40	9.4 ± 0.4	0.42 ± 0.10	0.08 ± 0.02	0.39 ± 0.03	2,3
RXJ1347.5−1145	15.5 ± 0.60	90.8 ± 1.0	0.20 ± 0.06	0.26 ± 0.01	0.63 ± 0.05	2
MACS1423.8+2404	6.5 ± 0.24	14.5 ± 0.4	0.35 ± 0.06	0.17 ± 0.03	0.25 ± 0.02	2,3
RXJ1532.9+3021	5.5 ± 0.40	20.5 ± 0.9	0.52 ± 0.12	0.18 ± 0.02	0.07 ± 0.01	2
MACS1720.3+3536	6.6 ± 0.40	13.3 ± 0.5	0.29 ± 0.09	0.17 ± 0.02	0.57 ± 0.05	2,3
Abell 2261	7.6 ± 0.30	18.0 ± 0.2	0.31 ± 0.06	0.10 ± 0.01	0.71 ± 0.06	5
MACS1931.8−2635	6.7 ± 0.40	20.9 ± 0.6	0.26 ± 0.08	0.30 ± 0.01	0.28 ± 0.02	2,3
RXJ2129.7+0005	5.8 ± 0.40	11.4 ± 2.0	0.33 ± 0.10	0.26 ± 0.02	0.55 ± 0.05	2
MS2137−2353	5.9 ± 0.30	9.9 ± 0.3	0.41 ± 0.08	⋅⋅⋅	⋅⋅⋅	2
RXJ2248.7−4431 (Abell 1063S)	12.4 ± 0.60	69.5 ± 0.1	0.39 ± 0.06	0.20 ± 0.01	0.74 ± 0.06	6
High Magnification Clusters:
MACS0416.1−2403	7.5 ± 0.80	16.0 ± 0.9	0.40 ± 0.14	⋅⋅⋅	⋅⋅⋅	3
MACS0647.8+7015	13.3 ± 1.80	32.5 ± 2.1	0.30 ± 0.19	0.36 ± 0.02	0.62 ± 0.06	3
MACS0717.5+3745	12.5 ± 0.70	55.8 ± 1.1	0.22 ± 0.07	0.30 ± 0.01	2.11 ± 0.18	3
MACS1149.6+2223	8.7 ± 0.90	30.2 ± 1.2	0.24 ± 0.11	0.25 ± 0.02	1.21 ± 0.10	3
MACS2129.4−0741	9.0 ± 1.20	22.6 ± 1.5	0.40 ± 0.17	0.19 ± 0.03	1.56 ± 0.14	3

Notes. ^aThe X-ray bolometric luminosity covers the energy range 0.1 − 100 keV. ^bThe X-ray ellipticity, , and centroid shift, <w >, values are from Maughan et al. (2008). ^cThe solar abundance reference values used here are from Anders & Grevesse (1989). References. (1) Maughan et al. (2008); (2) Allen et al. (2008); (3) Ebeling et al. (2007); (4) Cavagnolo et al. (2008); (5) Mantz et al. (2010); (6) CLASH team selected cluster.

Download table as:  ASCIITypeset image

The X-ray parameters in Table 4 are derived by us using CIAO v4.3 and CALDB v4.4.3. We filtered the Chandra data for flares and reprojected the deep Chandra background fields to match the cluster observations. The background data were also filtered for "status =0" events to be suitable for use with the VFAINT mode data. To obtain X-ray luminosities and temperatures uniformly across the sample, we extracted spectra from apertures with radii of 714 kpc (500 h⁻¹₁₀₀ kpc) and excluded the central 71.4 kpc (50 h⁻¹₁₀₀ kpc). These "core-excised" spectra provide X-ray temperature estimates that are relatively unaffected by the presence or absence of a cool core (e.g., Markevitch 1998). Figure 8 shows the X-ray temperature, T_x, as a function of the scaled bolometric luminosity, L_x/E(z), for CLASH clusters using the data in Table 4. The X-ray luminosities are scaled assuming self-similar evolution, $E(z) = \sqrt{\Omega _{m}(1+z)^3 + \Lambda }$, to allow direct comparison to the low-redshift luminosity temperature relationships from Markevitch (1998) and Pratt et al. (2009) (L₂ (measured within 0.15–1.00 R₅₀₀) versus T₃ (0.15–0.75 R₅₀₀), to be specific). We verified that the deep background particle event rates matched that seen in the cluster spectra between 9 and 12 keV. The X-ray counts were binned to a minimum of 20 counts per energy bin. Since the extracted spectra were dominated by source counts, the results were not very sensitive to the background scaling. To estimate the cluster temperature, we used XSPEC 12.6.0q to fit the X-ray spectra between 0.7 and 7.0 keV. We assumed a single temperature plasma model (the XSPEC apec model) absorbed by a Galactic hydrogen column fixed to the value obtained from the Bell survey (Dickey & Lockman 1990). The metallicity, temperature, and normalization were allowed to be free. The best-fitted temperatures and 1σ error bars, and the bolometric X-ray luminosities are reported in Table 4.

Zoom In Zoom Out Reset image size

The required size of our "relaxed" cluster sample is derived from the goal to measure "average" cluster concentrations to ∼10% (after accounting for variations in mass and redshift) and to detect a ∼15% deviation from the concentrations of simulated clusters at 99% confidence.

Statistically, if we assume measurement errors are independent and normally distributed, then the number of clusters, N_clus, needed to measure the average concentration to a fractional accuracy of f is just

$$\begin{equation} N_{{\rm clus}} = (\sigma _{{\rm tot}}/f)^2, \end{equation} \tag{ 1 }$$

where the total scatter in an individual measurement of the concentration, σ_tot, is

$$\begin{equation} \sigma _{{\rm tot}}^2 = \sigma _{{\rm int}}^2 + \sigma _{{\rm LSS}}^2 + \sigma _{{\rm meas}}^2 \end{equation} \tag{ 2 }$$

and includes contributions from intrinsic variation (including the effects of orientation averaging of triaxial halo shapes), variations due to intervening large-scale structure (LSS), and measurement uncertainties. Concentrations of relaxed simulated cluster halos have intrinsic scatters of ∼30% (e.g., Neto et al. 2007; Macciò et al. 2008; Duffy et al. 2008). LSS can scatter concentration measurements by ∼13% (Hoekstra et al. 2002). And we conservatively estimate measurement uncertainties to be ∼37% as follows based on empirical data (though analysis of simulated data suggests ∼11% accuracy is possible; Meneghetti et al. 2010).

Analysis of CL0024+17 yielded a concentration measurement uncertainty of ∼22% (Umetsu et al. 2010), which we will use here as a baseline. We assume here that our precisions will be dominated by the quality of our SL data and that our measurement uncertainty will scale roughly as $\sigma \propto 1/\sqrt{N_{{\rm arc}}} \propto 1/R_{E}$, where for CL0024+17 with R_E ∼ 35'', N_arc = 33 arcs were detected. Note that we have also assumed N_arc∝R²_E. Based on archival WFPC2 and ACS images, we measured a mean R_E ≈ 20'' for our sample of 20 relaxed clusters, which should conservatively yield N_arc ∼ 12 arcs per cluster. This scales the ∼22% uncertainty for CL0024+17 with 33 arcs up to a conservative ∼37% for an average cluster with just 12 arcs.

Adopting σ_int = 0.30, σ_LSS = 0.13, and σ_meas = 0.37, we find σ_tot = 0.49, and hence, if f = 0.10 then N_clus ∼ 24. This empirical sample size estimate is consistent with one derived from numerical simulations of strong lensing, as shown in Figure 9. The approximately log-normal distribution of DM halo concentrations seen in such simulations (Meneghetti et al. 2010) indicates that ∼20 massive clusters (M_vir > 5 × 10¹⁴ M_☉) are required to detect 15% deviations from expectations in the standard cosmological model at 99% confidence.

Zoom In Zoom Out Reset image size

Note that we may expect deviations to be greater than 15% if results from lensing-selected clusters are any indication. Simulated 10¹⁵ M_☉ h⁻¹ clusters at z = 0 typically have c_vir ∼ 5 (e.g., Duffy et al. 2008). For this fiducial cluster, Oguri et al. (2009) instead find c_vir ∼ 12 for observed clusters (based on a fit to values observed for 10 lensing clusters accounting for the various masses and redshifts). Even after including a 50% lensing bias, these observed concentrations are still ∼70% greater than expectations. Being more conservative and assuming a factor of two (100%) lensing bias (Meneghetti et al. 2010), the observed concentrations would still be ∼20% greater than expectations. Although as discussed in Section 2.1, more recent simulations (Prada et al. 2011) may alleviate these discrepancies.

While our X-ray selection criteria favor the inclusion of highly relaxed clusters in the CLASH sample, for eight of our clusters the dynamical state is somewhat ambiguous. Some researchers have reported evidence for substructure in the X-ray surface brightness profiles of these eight clusters. The presence of substructure may suggest that a cluster is not fully dynamically relaxed. We summarize this evidence below. We note that the presence of a minor degree of substructure does not inhibit our ability to determine a cluster's mass profile characteristics. In simulations, relaxed clusters have DM concentrations that are, on average, ∼20% higher than the general cluster population (Duffy et al. 2008; Prada et al. 2011).

• 1.  
  A209. This system is marginally unrelaxed according to Smith et al. (2005). They based their relaxed/unrelaxed characterization mainly on a mass ratio between the cluster main component and the total mass (both at r < 500 kpc). Clusters with mass ratio values below 0.95 are considered unrelaxed. Their value for A209 is 0.87 ± 0.06. Maughan et al. (2008, hereafter M08) derived an ellipticity of = 0.21 ± 0.01 for this cluster. Gilmour et al. (2009, hereafter G09) find it to be relaxed based on a visual examination of its X-ray morphology.
• 2.  
  MACSJ0329.7−0211 and CLJ1226.9+3332. These two clusters are included in the A08 compilation (as well as in Schmidt & Allen 2007, hereafter SA07), and were classified in these works as dynamically relaxed. However, M08 report that both exhibit evidence for substructure.
• 3.  
  MACSJ0744.9+3927. This cluster shows evidence for substructure (SA07; M08). On the other hand, G09 find it to be relaxed based on a visual examination of its X-ray morphology.
• 4.  
  MACSJ1206.2−0848. In both optical and X-ray images the cluster appears close to relaxed in projection, with a pronounced X-ray peak at the location of the BCG. However, some evidence of merger activity along the line of sight may be suggested by the very high velocity dispersion of 1580 km s⁻¹. G09 visually classify this cluster as relaxed.
• 5.  
  RXJ1347.5−1145. Significant substructure is observed in this system (SA07; M08). M08 measure an average ellipticity of the X-ray isophotes of = 0.26 ± 0.01.
• 6.  
  A2261. M08 found a small level of substructure and G09 classified it as disturbed.
• 7.  
  RXJ2248.7−4431. M08 found this cluster to be slightly elliptical = 0.2 ± 0.01 and with some level of substructure, while G09 found it to be relaxed.

Redshift estimates for multiply lensed images are crucial for breaking lensing degeneracies and tightening constraints on mass profiles (e.g., Broadhurst et al. 2005; Zitrin et al. 2009; Saha & Read 2009). However, most of the useful lensed images are much too faint for spectroscopy. Typical lensed source magnitudes are 23 < I < 28 (Figure 13), so that only the brightest arcs yield spectroscopic redshifts even when observed with the largest ground-based facilities. With continuous sampling of the broad wavelength range from the NUV to NIR (∼ 2000–17000 Å) that is enabled with WFC3 and ACS we can now obtain very accurate photometric redshifts (photo-z's) for most of the lensed objects down to an apparent F775W AB magnitude limit of 26.

Zoom In Zoom Out Reset image size

For a fixed total observing time, splitting observations into multiple (ideally overlapping) filters significantly improves photo-z precision (Benítez et al. 2009). We performed simulations to inform our filter selection and estimate our eventual photo-z precision. Galaxy magnitudes, redshifts, and SEDs were drawn from the UDF (Coe et al. 2006) and "re-observed" with our filter set, adding noise as appropriate given our proposed depths in each filter. Photo-z's were then re-estimated using BPZ (Benítez 2000; Benítez et al. 2004; Coe et al. 2006). In this simulation, we find that for 16 filters, we obtain very accurate (Δz ∼ 0.02(1 + z)) photo-z's for 80% of objects with F775W mag <26. Most importantly, we find that we will be able to acquire ∼6 times as many reliable photometric redshifts than spectroscopic redshifts for objects at z > 1, enabling a very substantial improvement in the number of unique constraints on the DM mass distributions (see Figure 13).

The coverage provided by the 16 WFC3 and ACS filters allows the Lyman-limit feature (rest frame 912 Å) to be photometrically traced to redshifts as low as z ∼ 1.5 and Lyα to be detected out to z ∼ 10. The inclusion of NUV photometry, for example, resolves one of the most common photo-z degeneracies between the Balmer break in z ∼ 0.2 galaxies and the Lyman break in z ∼ 3 galaxies (Rafelski et al. 2009).

The exposure times for the primary camera (cluster center position) are set primarily by the need to achieve 80% photometric redshift completeness down to F775W = 26 AB mag. This requires ∼18 orbits per cluster, as we need to achieve a 10σ limiting AB magnitude of 26 in each filter. We augment this by two orbits per cluster to extend the NIR depth to a 10σ limiting AB mag of F160W = 26.7 and F814W = 27.0 as needed for our lensed high-z galaxy search (Section 2.3 and see Table 5). The limiting magnitudes in Table 5 are for a 04 diameter circular aperture and a point source with a flat F_ν spectrum. The minimum exposure in any given filter is one orbit and observations in each filter are distributed over at least four exposures to ensure robust cosmic ray rejection.

The five NIR filters provide the ability to identify z > 7 galaxies with high confidence. Measured colors in these filters are essential for discriminating between such high-redshift galaxies and lower-redshift highly reddened galaxies (e.g., see Figure 14). The former are generally relatively flat in F_ν (or AB magnitudes) in the NIR while the latter generally increase in brightness as a function of wavelength. Quantitatively, the following selection criteria have proven effective: z − J > 0.8 (a flux decrement factor of 2.1) and J − H < 0.5 (Bradley et al. 2008, 2012; Zheng et al. 2009). To reliably measure the above colors for high-z lensed galaxy detection, our exposures are set to reach the following AB magnitude limits: F850LP: 26.7 (5σ), F110W: 27.8 (5σ), F125W: 26.5 (10σ), and F160W 26.7 (10σ).

Zoom In Zoom Out Reset image size

In each orbit we use a compact four-point dither pattern that provides half-pixel sampling along both detector axes. The dither pattern serves to both improve the spatial sampling of the point spread function (PSF), especially for the WFC3/IR detector with its large pixel scale of 0.128 arcsec pixel⁻¹, and to help remove hot pixels and other detector imperfections that may be unaccounted for in the calibration reference files. In the first epoch of each roll angle, we use a small-scale dither pattern to preserve, as best as possible in light of significant geometric distortions, half-pixel sampling across the detector. In subsequent epochs involving WFC3/IR observations, either in prime or parallel, we use a slightly larger dither pattern to help identify and remove persistence artifacts from compact sources, which, if uncorrected, could possibly be misidentified as SN candidates. While our small-scale dither patterns are much smaller than is needed to cover the WFC3/UVIS and ACS/WFC detector gaps, our cluster observations are obtained at two orientations, leaving only two small diamond-shaped regions (∼4.4 arcsec² each) in the central cluster area without data in all 16 filters.

Parallel observations are being obtained with a primary science goal of detecting SNe Ia. While the cluster core is being imaged with ACS, we obtain parallels with WFC3 and vice versa. For each of the two roll angles, observations are obtained over the course of four epochs spread out over ∼30–45 days (∼10–14 days between observations). Image-differencing analysis enables us to detect SNe which have gone off between observations. As noted above, the exposure sequences are executed either sequentially or interleaved, depending on scheduling availability.

Upon detection of a promising high-z SN Ia in the parallel field around a cluster, we can reprogram the remaining parallels to provide the follow-up (light-curve imaging and grism spectroscopy) necessary to measure its distance. The ability to reprogram the later of the two orientations to follow-up an SN detected early in a cluster observing sequence sets the upper limit on the angular offset between the two orientations—both orientations must be accessible during the entire cluster sequence. This constraint means that the two orientations cannot be more than ∼30° apart. A lower limit on the angular separation between the two orientations is set to ∼20° to ensure there is not excessive overlap in the ACS parallel pointings (with the exception of RXJ1347 where an orientation shift of 15° was needed to avoid placing a bright star in the parallel fields). The ability to reprogram a downstream orientation shift means that a new target of opportunity (ToO) observation will not be required if the SN is discovered early in the cluster observing sequence. In this regard, the design of the CLASH SN program has part of the follow-up built into its implementation. For the flexibility required to follow SNe found at the end of their corresponding cluster observing sequence, a small number of ToOs (four) and follow-up reserve orbits are included in the CLASH orbit allocation. CLASH parallel fields are not observed in 16 filters. The ACS parallels are taken in F775W and F850LP and the WFC3 parallels are taken in F350LP (UVIS), F125W (IR) and F160W (IR) (see Table 6).

The CLASH and CANDELS (Grogin et al. 2011; Koekemoer et al. 2011) SN programs are tightly coordinated and, in fact, share a common pool of reserve orbits from which both programs can draw upon for follow-up. This common reserve has been allocated 200 orbits: 50 from the CLASH allocation and 150 from CANDELS. The coordinated program is led by A. Riess who is a co-I in both programs.

While ACS functionality was restored in SM4, it is now only a "single-string" instrument, meaning there is no redundant path if the main CCD electronics box experiences a failure. This was a constraint imposed by the nature of the ACS repair. The primary impact of such a failure would be the loss of our parallel observations for SN searches. SNe could still be detected in our primary cluster core observations assuming we were to continue to distribute the exposures over multiple epochs. However, the use of SNe Ia for cosmology in strongly lensed regions is fraught with difficulty (see Section 2.2). While a failure of ACS would be a significant blow to the survey's dark energy science objectives, all other components of the CLASH science program, including our prime objective of studying the distribution of mass profile properties in clusters, would be able to be pursued with WFC3 alone.

If ACS fails permanently, we would abandon the dual orientation strategy and would, most likely, abandon the multiple epoch exposures, allowing the observations of each cluster to be completed on a much shorter time frame. By observing each cluster at a single roll angle, we would slightly increase the area in which we obtain full 16-filter coverage. All of the ACS filters are available with WFC3/UVIS, which would allow us to continue our 16-filter observations with WFC3 alone. However, there would be noticeable reductions in S/N in the redder filters (F775W, F814W, F850LP). In order to continue to achieve the same depths, we would require exposure times in the redder filters to be increased by 80%, adding about four orbits of integration time to each cluster.

The WFC3/UVIS and WFC3/IR fields of view are large enough to contain and yield robust photometric redshifts for most of the strongly lensed arcs. We would lose the extra areal coverage provided by the ACS observations, slightly reducing the area of overlap with our Subaru images. However, we estimate the impact on our ability to measure the mass profiles over the full radial range will be negligible.

SExtractor (Bertin & Arnouts 1996) is used to detect objects and measure their photometry. We run SExtractor (version 2.5.0) in dual image mode, using a detection image created from a weighted sum of the ACS/WFC and WFC3/IR images. The weights come from the inverse variance images produced by Mosaicdrizzle. We do not use the WFC3/UVIS images in the construction of the detection image but, of course, run SExtractor on the UVIS data to compute source photometry. We also create a detection image solely from the WFC3/IR images to optimize the search for high-redshift (z > 6) objects.

For object detection in the ACS+ IR images, we require a minimum of nine contiguous pixels at the level of the observed background rms or higher. The detection phase background sky level is computed in 5 × 5 grids of cells, with each cell being 128 × 128 pixels. For photometry, the local sky is estimated from a 24 pixel wide rectangular annulus around each detected object. The deblender minimum contrast ratio and number of threshold levels are 0.0015 and 32, respectively. These parameters were chosen after a systematic investigation of possible values and yields reasonable performance in minimizing spurious detections and suppressing overdeblending of bright objects, while achieving reasonable completeness in detecting faint sources behind bright cluster galaxies. The segmentation map for the cluster MACS1149.6+2223 using the above parameters is shown in Figure 15.

Zoom In Zoom Out Reset image size

Our image processing effectively rejects cosmic rays from the central regions of our UVIS and ACS images where we have four or more overlapping exposures. However, such rejection is not possible in the corners and edges of our images, where we have fewer exposures due to our observing strategy (two roll angles and dithering). As each cosmic ray only affects a single observation, it will often only be detected in a single filter. We prune these detections from our catalog by rejecting any object with only a single 5σ detection in one UVIS/ACS filter, as measured by SExtractor. We also reject any object without any 5σ detections.

We perform a separate IR-based detection with more aggressive deblending (64 levels of 0.0001 minimum contrast) and background subtraction (3 × 3 grids of 64 × 64 pixels). This detection is slightly more sensitive to redder objects, including those at high-redshift. It also performs slightly better at deblending these fainter objects, including those at high-z as well as arcs (strongly lensed galaxies), from brighter nearby cluster galaxies. As all WFC3/IR observations are obtained with multiple readouts and cosmic ray rejection, and to preserve our sensitivity to faint high-redshift objects, we do not prune this catalog based on 5σ detections.

Automated detection of strongly lensed galaxies is often challenging due to crowding in the cluster core, especially for low-surface brightness galaxies stretched into long, thin arcs. Progress has been made (e.g., Seidel & Bartelmann 2007), and we are testing this method for use with CLASH data. Currently, manual intervention is still required for those arcs which elude detection. We compare the initial source list with arc candidates identified both visually and based on our strong lens modeling. For arcs that are either missing from the detection list or that are only partially detected, we construct manual photometric apertures. We force SExtractor to adopt these apertures using the software package SExSeg (Coe et al. 2006). Arc photometry is then derived from images in which light from the BCG has been modeled and subtracted.

Isophotal apertures are used as they have been shown to yield robust colors (Benítez et al. 2004). In the source list presented in Section 5.3, no aperture corrections have been applied to the magnitudes. The PSF FWHM's span ∼007–015 (Ford et al. 2003; Sirianni et al. 2003; Dressel 2010). The photometric corrections for PSF variation are small for extended objects for all but the faintest sources, in part because we use relatively large isophotal apertures.

Extinction corrections are derived from the Schlegel et al. (1998) IR dust emission maps, and the resulting A_λ coefficients, in AB mag per unit E(B − V), are given in Table 5. Flux uncertainties are derived by SExtractor using the inverse variance images. We have compared the flux errors so derived with the predicted uncertainties from the ACS and WFC3 exposure time calculator and find excellent agreement.

We derive photometric redshift estimates using two independent packages: BPZ (Benítez 2000; Benítez et al. 2004; Coe et al. 2006) and LePHARE (Ilbert et al. 2006, hereafter LPZ). Both software packages use χ² minimization and template fitting but differ in their specific templates and their assumed priors. LPZ uses the SED library optimized for the COSMOS survey (Ilbert et al. 2009) without template interpolation. BPZ currently uses PEGASE SED templates (Fioc & Rocca-Volmerange 1997) which have been heavily recalibrated based on the FIREWORKS spectroscopic and photometric catalog (Wuyts et al. 2008). BPZ allows for interpolation between adjacent templates and uses an empirically derived prior on redshift and type based on observed magnitude. A full analysis of the accuracy and precision of our photometric redshift estimates will be discussed in an upcoming paper (S. Jouvel et al. 2012, in preparation). The photo-z accuracy analyses will be based on over 1000 redshifts, obtained from the facilities described in Section 6.

We present here an initial version of a representative CLASH source list for the z = 0.544 cluster MACS1149.6+2223. The full list is available online (at MAST treasury program archive Web site) and has over 100 parameters per object. In Table 7, we present just a small subset of the full list: 100 galaxies within 1 arcmin of the cluster center. This source list is based on the ACS+WFC3/IR detection image (see Section 5.1). Column 1 gives the SExtractor object ID, Columns 2 and 3 give the equatorial coordinates (J2000), Column 4 gives the isophotal area of the detection (in 0065 pixels), Column 5 gives the ellipticity of the source as measured by SExtractor, Column 6 gives the number of passbands in which the object is detected at a significance of 5σ or greater, and Column 7 gives the total number of bands in which the object is detected. In the case of MACS1149.6+2223, there are 17 HST passbands available because the archival imaging included ACS F555W. Note that even bright sources will sometimes not be detected in all bands because the footprints of each detector on the sky are different (see Figure 11). Columns 8–10 give the BPZ photometric redshift estimate, the BPZ Odds parameter (Benítez et al. 2004), and the χ² measurement presented in Coe et al. (2006). The 95% confidence limits on the range of the photometric redshift estimate are given in the row just below the best-fit photometric redshift value. This range is an indicator of the uncertainty in the redshift value. In general, BPZ photo-z estimates with Odds >0.90 are the most reliable estimates. Figure 16 shows the histogram of the BPZ photometric redshifts for 232 galaxies in the cluster MACS1149.6+2223 with BPZ Odds ⩾0.90, F850LP magnitude ⩽26.5 and z_phot ⩽ 3. The histogram shows the cluster peak very clearly.

Zoom In Zoom Out Reset image size

The remaining columns in Table 7 give the isophotal magnitudes (corrected for Galactic extinction) for five of the CLASH passbands: F275W, F390W, F850LP, F125W, and F160W. The photometric error in the isophotal magnitude is given, in parentheses, in the row directly below the magnitude value. Photometry has been corrected for Galactic extinction using an E(B − V) = 0.02297, and all magnitudes given are on the AB photometric system (see Table 5 for extinction coefficients used). A magnitude value of 99 indicates a non-detection (negative flux) in that band. In this case, the magnitude error value gives the 1σ detection limit in that band. A magnitude value of −99 indicates that the source lies outside of the detector field of view or lies within a gap between detectors. The full source lists for each CLASH cluster are released, via MAST, ∼6 months after the final observation for that cluster is acquired.