The Frontier Fields: Survey Design and Initial Results

Exceptionally deep observations of the distant universe with the Hubble Space Telescope (HST) have consistently pushed the frontiers of human knowledge. A succession of observing programs with each generation of HST detectors, in concert with the other NASA Great Observatories (Spitzer Space Telescope and Chandra X-ray Observatory), have probed the star formation and assembly histories of galaxies through more than 95% of the universe's lifetime. These observations have been made publicly available to the greater astronomy community, enabling a wide range of science and ancillary observing programs. The study of HST deep fields has established a number of techniques now standard in extragalactic astronomy, including the Lyman break selection of distant star-forming galaxies; photometric redshift determinations; stellar population fitting to multi-band photometry; quantitative morphological analysis; and the detection of high-redshift transient phenomena. Here we present the new Frontier Fields, an HST and Spitzer Director's discretionary time campaign to observe six massive strong-lensing clusters and six parallel fields, designed to simultaneously detect the faintest galaxies ever observed and provide a statistical picture of galaxy evolution at early times.

The first Hubble Deep Field (HDF) observations with HST WFPC2 revealed thousands of galaxies to 28th magnitude, fainter than any seen before (Williams et al. 1996; Ferguson et al. 2000). Utilizing the Lyman break technique (Guhathakurta et al. 1990; Songaila et al. 1990), the HDF and subsequent HDF-South (HDF-S; Casertano et al. 2000; Ferguson et al. 2000; Williams et al. 2000) detected significant numbers of distant star-forming galaxies visible in the optical out to redshifts z ∼ 5 (e.g., Madau et al. 1996). HST's deep images with high spatial resolution showed that many of these distant galaxies were smaller with higher surface brightnesses and more irregular structures than local galaxy populations (e.g., Abraham et al. 1996).

Follow-up observations of the HDF and HDF-S in the infrared with HST's NICMOS camera (Dickinson 1999; Thompson et al. 1999; Franx 2003) enabled studies of the stellar mass of the z < 5 populations (e.g., Papovich et al. 2001; Dickinson et al. 2003; Fontana et al. 2003) as well as the detection of higher redshift galaxies at 5 < z < 7 (Bouwens et al. 2003; Thompson 2003) and intrinsically redder populations (Fernández-Soto et al. 1999; Stiavelli et al. 1999; Labbé et al. 2003). Combined with the spectroscopic confirmation of many of these faint galaxies (e.g., Steidel et al. 1996; Lowenthal et al. 1997; Cohen et al. 2000), it became possible to track the cosmic star formation (Madau et al. 1996; Lanzetta et al. 2002; Bouwens et al. 2003) and assembly history of stellar mass (Dickinson et al. 2003) over the majority of the universe's lifetime. HST NICMOS observations of the HDF in 1997 discovered the highest redshift supernova (SN) of Type Ia known at that time (z = 1.7), confirming the acceleration of the universe (Riess et al. 2001).

After the original HDFs, synergistic multi-wavelength deep observations with Great Observatories and new capabilities on the HST further expanded the boundaries of our understanding. The installation of the Advanced Camera for Surveys Wide Field Camera (ACS/WFC; Ford et al. 1998) on HST in 2002 greatly improved the depth and area of optical imaging possible within a fixed exposure time. The fields for the Great Observatories Origins Deep Survey (GOODS; Giavalisco et al. 2004) were chosen to overlap with existing X-ray deep fields from Chandra (HDF/Chandra Deep Field North and the new Chandra Deep Field South (CDFS); Hornschemeier et al. 2000; Giacconi et al. 2001). New HST and Spitzer imaging produced high-quality and deep multi-wavelength photometry, revealed new distant galaxy populations, measured photometric redshifts, improved stellar mass estimates, and could be matched to faint X-ray sources in the Chandra Deep Fields (e.g., Mobasher et al. 2004; Treister et al. 2004; Barger et al. 2005; Fontana et al. 2006; Grazian et al. 2006; Pope et al. 2006). The cadence of the HST GOODS observations was designed to perform a systematic search for high-redshift supernovae (Riess et al. 2004a, 2004b). The location of the HST Ultra Deep Field (HUDF; Beckwith et al. 2006) within GOODS-S/CDFS was chosen to leverage this existing data with an additional 400 orbits (268 hr) to reach optical depths of ∼29th magnitude, fainter than original HDF WFPC2 limits. Deep NICMOS IR imaging was also obtained in the HUDF (Thompson et al. 2005). The resulting "wedding cake" survey of the combined GOODS and HUDF observations proved to be an important strategy for spanning the depth and area needed to constrain both the bright and faint ends of the luminosity function of galaxies approaching the epoch of reionization (e.g., Bouwens et al. 2007).

With the success of the HST servicing mission SM4 in 2009 and the installation of the Wide Field Camera 3 (WFC3; MacKenty et al. 2008) with its IR channel, HST greatly improved the efficiency of its high-spatial-resolution near-infrared imaging. The WFC3 Early Release Science near-infrared observations of GOODS-S (Windhorst et al. 2011) and deep imaging in HUDF and parallels revealed new populations of galaxies at z ∼ 8 (Bouwens et al. 2010; Illingworth & Bouwens 2010). The UDF 2012 WFC3/IR campaign (Ellis et al. 2013; Koekemoer et al. 2013) added the F140W filter and deeper observations in F105W and F160W filters to increase the detection efficiency of highest redshift candidates (8.5 < z < 12) to ∼30th magnitude. (See also Illingworth et al. 2013 for a separate reduction of all HUDF data.) Wider field near-infrared imaging with the HST Multi-Cycle Treasury Cosmic Assembly Near-infrared Deep Extra-galactic Legacy Survey (CANDELS; Grogin et al. 2012; Koekemoer et al. 2011) built upon the previous HST ACS/WFC and Spitzer observations of the GOODS, GEMS (Rix et al. 2004), COSMOS (Scoville et al. 2007), EGS (Davis et al. 2007), and UDS (Lawrence et al. 2007) extragalactic legacy fields. Thanks to WFC3, detections of z ∼ 8 candidates are now relatively commonplace (e.g., Finkelstein et al. 2010; Labbé et al. 2010; McLure et al. 2011; Yan et al. 2011; Bradley et al. 2012). The current measurement of the cosmic star formation history extends to less than 500 Myr after the Big Bang (e.g., Ellis et al. 2013; Oesch et al. 2013, 2016; Bouwens et al. 2015a; Finkelstein et al. 2015; but see Brammer et al. 2013; Pirzkal et al. 2013), albeit with very small numbers of candidates at z > 9. HST's observations of high-redshift galaxies have placed important constraints on cosmological measures of reionization (e.g., Bouwens et al. 2015b; Finkelstein et al. 2015; Robertson et al. 2015).

With the launch of the James Webb Space Telescope (JWST) still several years away, and no new servicing missions to HST planned, significant progress on understanding the first billion years of the universe with the remaining years of HST poses a major challenge. The HST and Spitzer projects proposed supporting a new joint Deep Fields program supported with Director's discretionary time in their 2012 NASA Senior Review proposals. The Hubble Deep Fields Initiative science working group (HDFI SWG) was convened by STScI Director M. Mountain in 2012. They recommended a new strategy to "go deep": use massive clusters of galaxies as cosmic telescopes, combined with very deep HST and Spitzer observations.¹¹ Very massive clusters of galaxies are the most massive structures in the universe, bending spacetime to create efficient gravitational lenses (e.g., Kneib & Natarajan 2011). The light from galaxies behind these natural telescopes experiences magnification factors of a few within a few arcminutes of the cluster cores, and magnifications ∼10 or greater within smaller windows along the critical curves. Therefore, HST observations of these strongly lensed fields can probe galaxies as intrinsically faint or fainter than those detected in the HUDF in a much shorter exposure time—provided those galaxies fall within the high-magnification windows.

The advantages of this strategy had already been demonstrated by the Cluster Lensing and Supernova Survey (CLASH; Postman et al. 2012), a 524-orbit HST Multi-Cycle Treasury Program to study the gravitational lensing properties of 25 galaxy clusters. CLASH targeted each cluster with shallow observations in 16 ultraviolet–near-infrared HST bandpasses, in order to obtain precise photometric redshift constraints on background lensed galaxies. Within only a few orbits of HST time in the reddest filters, CLASH discovered several z > 9 galaxy candidates highly magnified by intervening massive clusters at z ∼ 0.5 (Zheng et al. 2012; Coe et al. 2013; Bouwens et al. 2014).

The Frontier Fields program is an ambitious multi-cycle observing campaign using Director's discretionary time with HST and Spitzer Space Telescope to peer deeper into the universe than ever before. The Frontier Fields combine the power of HST with the natural gravitational telescopes of six high-magnification clusters of galaxies to produce the deepest observations of clusters and their lensed galaxies ever obtained. The HST cluster images are obtained in parallel with six "blank" field images; the parallel field images are the second deepest images ever obtained, and triple the blank field area imaged to a depth of 29th ABmag. The Spitzer Space Telescope is also dedicating over 1000 hr of Director's discretionary time to obtain IRAC 3.6 and 4.5 μm imaging to depths of 26.5, 26.0 ABmag in the six cluster and six parallel Frontier Fields. In this paper, we describe the primary science goals in Section 2, the field selection criterion in Section 3, the Frontier Field clusters and parallel fields in Section 4, the HST and Spitzer observations in Section 5, and the public Frontier Fields lensing modeling effort in Section 6. In Section 7, we present the image depths achieved for the HST observations for the first two cluster and parallel fields, and review the current results relevant to primary science goals. Further details, the latest HST data releases, and Frontier Fields updates may be found at www.stsci.edu/hst/campaigns/frontier-fields/ and at the STScI MAST Archive (10.17909/T9KK5N). Details describing the Spitzer observations will be presented in P. Capak et al. (2017, in preparation) and more information is available at ssc.spitzer.caltech.edu/warmmission/scheduling/approvedprograms/ddt/frontier/.

The primary science goals of the Frontier Fields are to explore the high-redshift universe accessible only with deep HST observations, and to set the scene for JWST studies of the early universe. Studies of high-redshift quasar absorption lines have found that the epoch of reionization was completed by z ∼ 6 (Fan et al. 2006), while observations of the cosmic microwave background place the start of reionization before z ∼ 10 (e.g., Spergel et al. 2003; Hinshaw et al. 2013; Planck Collaboration et al. 2016). Including recent estimates of the optical depth from Planck data, the era between z ∼ 11 and z ∼ 6 probed by the deepest and reddest HST observations marks a critical transition in the universe's history (e.g., Bouwens et al. 2015b; Mitra et al. 2015; Planck Collaboration et al. 2016; Robertson et al. 2015).

The installation of the HST WFC3 camera with the near-infrared channel dramatically increased the number of galaxy candidates detected at z > 6. However, prior to the start of the Frontier Fields in 2013, astronomers' understanding of the galaxy populations during the epoch of reionization was based largely on those detected in direct HST WFC3/IR imaging surveys (HUDF, CANDELS, BORG) and handfuls of lensed objects in shallow HST observations from CLASH. The detected unlensed galaxies are the most luminous objects of their era, and thus significantly more massive and rare than the progenitors of today's Milky Way galaxies (e.g., Behroozi et al. 2013; Boylan-Kolchin et al. 2014). High-redshift galaxies are barely resolved by HST (Oesch et al. 2010; Ono et al. 2013), with lensed z > 8 galaxies yielding intrinsic sizes less than a few hundred parsecs across (Coe et al. 2013). Because such high-redshift galaxies are often observable only in the reddest HST bandpasses, limited information about their rest-frame ultraviolet slopes, stellar populations, and dust content can be inferred from their observed colors (e.g., Bouwens et al. 2012; Finkelstein et al. 2012). Unseen z > 6 dwarf galaxies well below HST's nominal direct detection limit are needed to produce the number of ionizing photons required to disassociate the universe's reservoir of intergalactic neutral hydrogen (e.g., Finkelstein et al. 2015; Robertson et al. 2015). Very few candidates at z ∼ 9 and above were identified (Zheng et al. 2012; Coe et al. 2013; Ellis et al. 2013; Oesch et al. 2013), resulting in vigorous debate about how quickly the first star formation proceeded and how many z > 9 objects future JWST observations might see (Oesch et al. 2012). (The role of early black holes in terms of their contribution to the reionization budget is currently unknown and this will be revealed by JWST.)

In order to address many of these unknowns, the Frontier Fields program was designed with the following science aims:

• 1.  
  To reveal populations of z = 5–10 galaxies that are >10 times fainter than any currently known, the key building blocks of ∼L* galaxies in the local universe.
• 2.  
  To characterize the stellar populations of faint galaxies at high redshift and solidify our understanding of the stellar mass function at the earliest times.
• 3.  
  To provide, for the first time, a statistical morphological characterization of star-forming galaxies at z > 5.
• 4.  
  To find z > 8 galaxies stretched out enough by foreground clusters to measure sizes and internal structure and/or magnified enough for spectroscopic follow-up.

The Frontier Fields (FF) combines several previous high-redshift galaxy observing strategies to achieve these aims: very deep multiband HST imaging to identify very faint distant galaxy candidates by their color, and strong gravitational lensing by massive clusters of galaxies to probe galaxies fainter than those accessible with direct "blank" field HST imaging. Deep imaging with the Spitzer IRAC 3.6 and 4.5 μm bands is also required to improve photometric redshifts, measure stellar masses and specific star formation rates, and rule out low-redshift interlopers (e.g., Labbé et al. 2013; Bradac̆ et al. 2014). The clusters and their exact pointings were selected to optimize the number of detectable z ∼ 10 objects within the HST WFC3/IR field of view magnified by factors of ∼1.5–100, depending on their positions relative to the critical curves of the clusters. The HST exposure times were chosen to probe intrinsic depths up to and more than 10 times fainter than the HUDF in the highest magnification regions of the lensed fields (intrinsic magnitudes ∼32, with a magnification factor of 20–30), but with significantly less observing time. The volumes probed at the highest magnifications are very small (see Coe et al. 2015) with significant cosmic variance for rare and high-redshift galaxies (e.g., Robertson et al. 2015), thus the program observes multiple clusters to improve the statistical likelihood of capturing the light from the faintest and most distant galaxies. While color, redshift, and other relative measures such as specific star formation rates and emission-line equivalent widths are immune to errors in the magnification estimates, measurements of the intrinsic luminosities and sizes of individual objects depend directly on the inferred lensing magnifications. (Integrated quantities such as galaxy luminosity functions are less susceptible to magnification uncertainties.) In concert with the observing campaigns using Director's discretionary time, a unified effort to create high-fidelity public maps of the lensing properties of each FF cluster is an integral part of the FF (see Section 6).

Because each cluster is observed at a fixed HST roll angle for an extended period, we also obtain simultaneous deep parallel field observations at a single pointing centered ∼6 arcmin from the cluster core (>1.8 projected co-moving Mpc for a z > 0.3 lensing cluster). These six new "blank fields" are comparable in depth to the HUDF parallel fields (Oesch et al. 2007), and triple the area of unlensed fields observed by HST to a depth of ∼29th ABmag. The background volumes lensed by the clusters are much smaller than those probed by unlensed fields. Thus, while the cluster pointings allow us to see intrinsically fainter objects than the HUDF within small volumes, the parallel fields provide a dramatic improvement in the volume and statistical counting of distant galaxies brighter than 29th magnitude. This is particularly important for understanding the biases associated with cosmic variance, i.e., the fact that every single sightline through the universe is unique (e.g., Robertson et al. 2014).

The Frontier Fields will set the stage for the JWST to study first-light galaxies at z > 10 and to understand the assembly of galaxies over cosmic time. JWST is a 6.5 m cold telescope sensitive at 0.7–27 μm, to be launched at the end of 2018 with a limited lifetime requirement of 5 yr and a goal of 10 yr. Because JWST's lifetime is short relative to HST's, it is important for the astronomical community to be prepared for JWST observations early on. The high-redshift galaxy candidates detected by the Frontier Fields are likely to be among the first spectroscopic targets for JWST, and current studies will produce a better understanding of the high-redshift galaxy luminosity functions, spectral energy distributions, and sizes needed to effectively plan for JWST surveys. The HST Frontier Fields (HFF) high-resolution optical imaging shortward of 0.7 μm in ACS F435W and F606W reaches depths comparable to those achieved by JWST NIRCam within 1–2 hr, and hence provides an important legacy data set for future JWST extragalactic work. Finally, direct observations of the faintest first galaxies and the dwarf galaxies and early accreting black holes expected to be responsible for reionization will be challenging even with JWST. Development of cluster lens modeling techniques now will enable future JWST studies of strong-lensing clusters and their lensed galaxies.

The Frontier Fields data offer the opportunity to do ground-breaking science in a number of fields other than the highest redshift universe. Several complementary HST General Observer (GO) observing programs have been awarded to obtain deep WFC3/UV imaging (GO 13389, 14209; PI B. Siana; Alavi et al. 2016), WFC3/IR grism spectroscopy (GO 13459; T. Treu), and target-of-opportunity follow-up of transient events (GO 13386, 13790, 14208; S. Rodney). Hundreds of multiply imaged background galaxies at all redshifts have permitted the construction of dark-matter maps of the clusters at unprecedented resolution to probe cluster substructure (e.g., Jauzac et al. 2014, 2015; Wang et al. 2015; Hoag et al. 2016; Limousin et al. 2016; Mohammed et al. 2016; P. Natarajan 2017 in preparation), and will enable new cosmological constraints via angular scaling relations (e.g., Kneib & Natarajan 2011). At the recommendation of the HFF review committee, an exercise comparing the various independent lens modeling methodologies and their fidelity has been on-going and results where more than 10 independent research groups participated are now in preparation (Meneghetti et al. 2016). Detailed studies of galaxies observed both at high magnification and in deep parallel imaging will probe their internal structures, stellar populations, and luminosity functions (e.g., Livermore et al. 2012; Alavi et al. 2014, Jones et al. 2015; Castellano et al. 2016a; Pope et al. 2016) These deepest-ever images of massive galaxy clusters have detected intracluster light, ram-pressure stripping, and tidal streams at z > 0.3 (e.g., Montes & Trujillo 2014; McPartland et al. 2016), probing the dynamic processes impacting galaxy evolution within these unique environments. The new HST Frontier Fields observations have detected a number of transients (e.g., Rodney et al. 2015), including the light curves from the first multiply imaged supernova (Kelly et al. 2015; discovered in GLASS).

The six Frontier Field clusters and parallel fields (Table 1) were selected to meet the primary scientific goals outlined in the HDFI SWG recommendations, as well as to optimize the HST and Spitzer observing campaigns. A list of 25 cluster candidates was suggested by the HDFI SWG, and additional candidates were suggested by the community during the selection process. Each cluster was evaluated using the following criteria.

Table 1.  The Frontier Fields Locations

Cluster	Cluster Center (J2000)		Parallel Center (J2000)		Epoch1	Epoch2	Zodiacal ${H}_{F160W}$ a	${E}_{(B-V)}$ b
	α	δ	α	δ	HST	HST	(ABmag/$\square ^{\prime\prime}$)
Abell 2744	00:14:21.2	−30:23:50.1	00:13:53.6	−30:22:54.3	10/2013–12/2013	5/2014–7/2014	22.2/21.9	0.012
MACSJ0416.1−2403	04:16:08.9	−24:04:28.7	04:16:33.1	−24:06:48.7	1/2014–2/2014	7/2014–9/2014	22.4/22.3	0.036
MACSJ0717.5+3745	07:17:34.0	+37:44:49.0	07:17:17.0	+37:49:47.3	9/2014–12/2014	2/2015–3/2015	21.8/22.0	0.068
MACSJ1149.5+2223	11:49:36.3	+22:23:58.1	11:49:40.5	+22:18:02.3	11/2014–1/2015	4/2015–5/2015	21.9/22.0	0.020
Abell S1063	22:48:44.4	−44:31:48.5	22:49:17.7	−44:32:43.8	10/2015–11/2015	4/2016–6/2016	22.2/20.6	0.010
Abell 370	02:39:52.9	−01:34:36.5	02:40:13.4	−01:37:32.8	12/2015–2/2016	7/2016–9/2016	21.8/21.9	0.028

Notes.

^aTypical zodiacal background in ${H}_{F160W}$ for HST Epoch1/Epoch2 observations; computed using HST exposure time calculator and median observing date. ^bSchlafly & Finkbeiner (2011), courtesy of the NASA/IPAC Extragalactic Database.

Download table as:  ASCIITypeset image

Lensing properties. The primary consideration for selecting each of the Frontier Fields was the lensing strength of the cluster. Each cluster's lensing strength was evaluated by calculating the likelihood of observing a z = 9.6 galaxy magnified to ${H}_{F160W}\leqslant 27$ ABmag within the HST WFC3/IR field of view, ignoring corrections for incompleteness or sky brightness (Table 2). This metric encapsulates several constraints into a single number, making it straightforward to compare different clusters. It was calculated based on cluster lens models and a nominal blank-field luminosity function. Preliminary lens models were provided by two independent modelers, J. Richard and A. Zitrin. We adopted a luminosity function with ${\phi }^{* }=4.27\times {10}^{-4}$, ${M}_{{UV}}^{* }=-19.5$, and α = −1.98, extrapolated from z ∼ 8 (Bradley et al. 2012) by assuming ${{dM}}^{* }/{dz}=0.46$ (Coe et al. 2015). Actual expectations at z = 9.6 might be lower (e.g., Oesch et al. 2014; Bouwens et al. 2014), but the choice of luminosity function does not affect the cluster rankings.

Table 2.  Frontier Fields: Cluster Properties and Ancillary Data

Cluster	za	Mvira	LXa	P(z = 9.6)b	Parallel μc	Spitzer	Herscheld	ALMAe
		(M⊙)	(erg s−1)	H ≤ 27		MIPS 24 μm	PACS/SPIRE
Abell 2744	0.308	1.8 × 1015	3.1 × 1045	0.69 ± 0.07	1.14–1.21	yes	100/250/350/500	yes
MACSJ0416.1−2403	0.396	1.2 × 1015	1.0 × 1045	0.63 ± 0.12	1.09–1.16	no	100/250/350/500	yes
MACSJ0717.5+3745	0.545	2–3 × 1015	3.3 × 1045	0.84 ± 0.05	1.07–1.42	yes	100/250/350/500	no
MACSJ1149.5+2223	0.543	2.5 × 1015	1.8 × 1045	0.60 ± 0.10	1.02–1.07	no	70/100/250/350/500	yes
Abell S1063	0.348	1.4 × 1015	1.8 × 1045	0.69 ± 0.08	1.02	yes	70/100/250/350/500	yes
Abell 370	0.375	∼1 × 1015	1.1 × 1045	0.90 ± 0.08	1.2–1.3	yes	100/250/350/500	yes

Notes.

^aSee the text for references for each cluster. ^bMedian probability of lensing a z = 9.6 background galaxy to apparent ${H}_{F160W}$ ABmag ≤ 27 within the WFC3/IR field of view, calculated using the pre-HFF v1.0 lensing models. ^cMedian magnification factor μ in the parallel fields within the WFC3/IR field of view; based on the weak-lensing estimates from pre-HFF v1.0 Merten models. Note that magnification factors may be larger at locations closer to the cluster. ^dSee Rawle et al. (2016) for a summary of Herschel and Spitzer cryogenic observations. Note that the Herschel SPIRE 250/350/500 mm field of view covers both cluster and parallel fields for all but MACSJ0416.1-2403. ^eVisibility from ALMA.

Download table as:  ASCIITypeset image

Table 2 provides updated versions of these estimates based on the publicly available version 1 lens models (see Section 6). The expectations are similarly high for all clusters. Clusters with lower expectations were generally excluded from our selection (e.g., MACSJ0329.6-0211 and MACSJ1423.8+2404). We note in practice that cosmic variance along each lensed line of sight is likely to be an equally important factor in detection rates of high-redshift galaxies; this cannot be known a priori for a given cluster.

Because we based our selection upon the results of the lensing model predictions, our selection was biased toward better studied clusters with existing imaging and spectroscopic data from which lensing models could be constructed. Some otherwise promising clusters (e.g., El Gordo; Menanteau et al. 2012) could not be evaluated, because insufficient lensing model constraints were available at the time of selection.

We excluded several lower-redshift z < 0.3 strong-lensing clusters (e.g., Abell 1689) because we could not adequately sample the critical curves of the low-redshift clusters within a single WFC3/IR 22 × 20 pointing. We note that although the critical curves of the z = 0.545 merging cluster MACSJ0717.5+3745s could not be covered by a single WFC3/IR pointing, the probability of observing a lensed z ∼ 9.6 galaxy was among the highest of all cluster candidates.

Sky brightness and Galactic extinction. Observations of the very faint extragalactic universe are limited by the brightness of the sky and by foreground Galactic extinction. Zodiacal light can have a significant impact on the depths obtained by HST and Spitzer imaging within a given exposure time. This background depends upon the angular distance of the target from the Sun and the ecliptic (Figure 1). Targets observed with high zodiacal backgrounds have near-infrared sky brightnesses several magnitudes brighter than the lowest zodiacal backgrounds, resulting in images with significantly lower signal-to-noise ratio within a given exposure time. Given the highly constrained roll angles required to obtain observations of a fixed parallel field with both the WFC3 and ACS cameras, we have a limited ability to mitigate the impact of the zodiacal background by constraining the solar avoidance angle (Table 1). Therefore, strong preference was given to clusters at high ecliptic latitudes. This selection criterion excluded a number of strong-lensing clusters at low ecliptic latitudes. Additionally, clusters at high Galactic latitude with low extinction were strongly preferred (Figure 1, Table 1). MACS0717.5+3745 has relatively high Galactic extinction, with ${E}_{B-V}=0.068$ (Schlegel & Finkbeiner 2011). However, this cluster was the second strongest potential lenser on our list of candidates (Table 2). Estimates of the ${H}_{F160W}$ zodiacal background at the epoch of observation and Galactic extinction for each cluster are given in Table 1.

Zoom In Zoom Out Reset image size

Suitability of available parallel fields. The HST observing strategy requires the simultaneous observation of the cluster field and a blank parallel field with WFC3/IR and ACS cameras. Ideally, these parallel fields would complement the existing deep unlensed images of the HUDF and its two parallel fields, without the complication of magnification by the clusters. However, as we discuss below, this observing requirement limits the range of available roll angles, and hence locations for the parallel fields. The potential parallel field locations were selected to avoid bright stars and extended cluster structures when possible. The weak-lensing signal for each of the parallel fields was also examined where possible (J. Merten, E. Medezinski, K. Umetsu 2012, private communication). The weak-lensing signal within the parallel fields has median magnification factors between 1.02 and 1.30 for background galaxies in the range 1 < z < 9; see discussion of each cluster for detailed estimates.

Suitability for ground-based follow-up. Follow-up of interesting objects detected in the Frontier Fields requires access to those fields from the major ground-based facilities. ALMA in particular has the potential to spectroscopically confirm the redshift of very high-redshift (z > 6) galaxy candidates via the [C ii] 158 μm line and other atomic emission lines (e.g., da Cunha et al. 2013). Additionally, spectroscopic redshifts of multiply imaged galaxies add strong constraints to the lensing models for the clusters. Thus, access to the telescopes on Maunakea, in addition to southern facilities like ALMA and VLT, was a major consideration. Five out of the six selected clusters are visible from ALMA, with MACS0717.5+3745 as the exception (Tables 1, 2). Five out of the six clusters are visible from Maunakea, with Abell S1063 as the exception.

Existing ancillary data. The existence of supporting data was a key consideration recommended by the HDFI SWG. Many of the candidate clusters have been studied previously by space missions, including HST, the Spitzer cryo-mission with MIPS and IRAC (including 5, 8 μm channels), Herschel, XMM, and Chandra (see the discussion of each cluster for details). Additionally, ground-based spectroscopic and wide-field imaging survey data were evaluated from the literature. Four of the chosen clusters were drawn from the CLASH survey (Postman et al. 2012), with supporting multi-band shallow HST imaging, wide-field ground-based imaging (Subaru), spectroscopy (VLT), as well as archival Herschel and Chandra data. Since the announcement of the Frontier Field selection, the community has responded with additional observations with Chandra (PI S. Murray, C. Jones-Forman), VLA (PI E. Murphy), XMM (PI J. P. Kneib, Eckert et al. 2015), ALMA (PI F. Bauer), LMT (PI A. Pope), Gemini GeMS/GSAOI K_s imaging (e.g., Schirmer et al. 2015), VLT Hawk-I K_s imaging (PI D. Marchesini & G. Brammer), VLT MUSE spectroscopy (PIs K. Caputi & Clément, F. Bauer, J. Richard, C. Grillo, e.g., Karman et al. 2015, Grillo et al. 2016), as well as the release of previously unpublished data on these fields (e.g., Ebeling et al. 2014, Gruen et al. 2014). Additional HST WFC3 ultraviolet imaging (Alavi et al. 2016) and grism spectroscopy (Schmidt et al. 2014) are also being obtained for all the Frontier Field clusters. We continue to maintain a clearing-house website for public data links and Frontier Fields-related publications: www.stsci.edu/hst/campaigns/frontier-fields/FF-Data.

In addition to the science-driven considerations given above, we optimized the cluster selection for a number of practical issues.

HST observability. The Frontier Fields are observed with HST at a fixed roll angle and its 180° offset in order to obtain deep observations in the cluster field and parallel field with both WFC3/IR and ACS. These observations consist of 70 orbits at each orientation. Each field was evaluated to determine the ability to hold a fixed roll angle for more than 30 days and the availability of guide stars at these orientations. For optimal stability, HST requires two guide stars brighter than 15th magnitude. Our initial evaluation of MACSJ1149.5+2223 found only one acceptable guide star; however, a second guide star with a magnitude slightly fainter than the nominal limit was available. This new guide star was tested in early observations and found to be suitable.

Spitzer observability. Each cluster and parallel field was evaluated by the Spitzer implementation team. Spitzer observations are sensitive to bright stars in the field, because saturation above ∼35,000 DN can result in "column pull-down" impacting the data quality along the affected column. MACSJ0647.7+7015 (e.g., Coe et al. 2013) in particular was found to have unacceptably bright stars in the vicinity, and was excluded.

Scheduling. Each set of cluster/parallel field observations constitutes a considerable investment of HST time, with 70 orbits at each orientation and 140 orbits total per field. The optimal scheduling of these observations is a challenge. We also anticipated that the Frontier Fields would be popular fields for ancillary HST observing programs. Therefore to avoid schedule collisions with the main Frontier Field program, supporting Frontier Field programs, and other popular HST fields (e.g., the UDF/GOODS-S), the Frontier Fields were selected to span a range in R.A. The order in which the fields are observed was determined primarily by the desire to prevent overlapping epochs of HST observations.

JWST observability. Each of the selected Frontier Fields positions was run through preliminary JWST scheduling software and confirmed to have extended periods of JWST visibility.

In 2013 February, the six Frontier Field clusters and their parallel field locations were finalized and announced prior to the HST Cycle 21 proposal deadline. The Frontier Field clusters are Abell 2744, MACSJ0416.1-2403, MACSJ0717.5+3745, MACSJ1149.5+2223, Abell S1063 (also known as RXCJ2248.7-4431), and Abell 370 (Table 1). These clusters are at redshifts between 0.3 and 0.55, and are among the most massive known clusters at these redshifts (Table 2). All of the clusters had previous (shallow) HST imaging, with four clusters previously observed as part of the CLASH HST MCT survey (MACSJ0416.1-2403, MACSJ0717.5+3745, MACSJ1149.5+2223, and Abell S1063) and all but Abell 370 were part of the Massive Clusters Survey (Ebeling et al. 2001).

4.1. Abell 2744

Abell 2744 (Figure 2) is a massive X-ray-luminous merging cluster at z = 0.308, (Couch & Newell 1982; Abell et al. 1989), also known as AC118 or "Pandora's Cluster." It has a total X-ray luminosity of L_X = 3.1 × 10⁴⁵ erg s⁻¹ at 2–10 keV (Allen 1998), with X-ray emission concentrated on the southern compact core and extending to the northwest (Owers et al. 2011; Eckert et al. 2015). Its virial mass within the central 1.3 Mpc is $\sim 1.8\times {10}^{15}\,{M}_{\odot }$ (Merten et al. 2011). The velocity dispersion is σ = 1497 ± 47 km s⁻¹ (Owers et al. 2011), but shows two distinct structures, with the northern substructure offset in velocity by −1600 km s⁻¹ and σ ∼ 800 km s⁻¹ (Boschin et al. 2006; Braglia et al. 2007). Abell 2744's complicated velocity structure and lensing properties suggest that it is a merging system with at least three separate substructures (Cypriano et al. 2004; Braglia et al. 2007; Merten et al. 2011). Weak-lensing analysis by Merten et al. (2011) identified four mass concentrations labeled core, N, NW, W of 2.2, 0.8, 1.1, 1.1 × 10¹⁴ M_⊙ respectively, with the NW structure showing evidence for spatially separated dark matter, gas, and galaxies. Abell 2744 is also host to a powerful extended radio halo with P_{1.4 GHz} = 1.5 × 10²⁵ W Hz⁻¹ (Giovannini et al. 1999).

Zoom In Zoom Out Reset image size

Despite its obviously complicated geometry, Abell 2744 was one of the strongest Frontier Field cluster candidates based on its lensing strength, sky location, and pre-existing ancillary data. The pre-FF lensing model by Merten et al. (2011; using the Light-traces-mass modeling method of Zitrin et al. 2009a) found 34 strong-lensed images of 11 galaxies in HST F814W imaging of the core of Abell 2744 (HST GO 11689, PI R. Dupke), giving a core mass $\sim 2\times {10}^{14}\,{M}_{\odot }$. This core region is ∼100'' × 100'', and therefore fits within the HST WFC3/IR field of view of 22 × 21. Analysis of preliminary models constructed by Zitrin and Richard separately suggested a very high probability of magnifying a z ∼ 10 galaxy to H = 27 ABmag within the WFC3/IR field of view. This high lensing probability has been confirmed by subsequent models provided by the lensing map effort and independent teams (e.g., Atek et al. 2014; Johnson et al. 2014; Lam et al. 2014; Richard et al. 2014; Zitrin et al. 2014; Coe et al. 2015; Ishigaki et al. 2015; Jauzac et al. 2015; Wang et al. 2015; Table 2).

Abell 2744 has one of the darkest skies and lowest Galactic extinctions, ${E}_{(B-V)}=0.012$ (Schlafly & Finkbeiner 2011), of all the cluster candidates. The typical zodiacal background in ${H}_{F160W}$ during the cluster IR epoch (10/2013–12/2013) and the parallel IR epoch (5/2014–7/2014) is ∼22.2 and 21.9 ABmag per $\square ^{\prime\prime}$ respectively. At a decl. of −30, it is easily observable with ALMA and the VLT but also within reach of Maunakea and the Very Large Array. It has been extensively studied by the Chandra X-ray Observatory (e.g., Kempner & David 2004; Merten et al. 2011; Owers et al. 2011). Abell 2744 was also observed during the Spitzer cryo-mission, with MIPS 24 μm and IRAC 3.6–8 μm observations (PI G. Rieke). This cluster is part of the Herschel Lensing Survey (Egami et al. 2010), with deep Herschel Space Observatory PACS 100/160 μm and SPIRE 250/350/500 μm imaging.

The choice of parallel field (Figure 3, Table 1) was particularly challenging in this case. HST roll angles with >30 day observing windows at both orientations placed the observable parallel field 6' either east or west of the core of Abell 2744. However, the eastern parallel field location was undesirable because of the presence of an unavoidable bright star. Therefore the western parallel field location (α = 00:13:53.6, δ = −30:22:54.3, J2000) was chosen. The parallel field is ∼1'–2' west of the NW and W substructures identified in Merten et al. (2011). The weak-lensing magnification boost from the cluster is therefore predicted to be significant, with median magnification factors ∼1.14–1.21 and maximum magnification factors 1.5–1.85 for 1 < z < 9 within the WFC3/IR pointing based on the pre-HFF v1.0 Merten model (Table 2).

Zoom In Zoom Out Reset image size

4.2. MACSJ0416.1-2403

MACSJ0416.1-2403 (Figure 4(a)) is a massive elongated X-ray-luminous cluster at z = 0.397 (Ebeling et al. 2007, 2014).¹² Its bolometric X-ray luminosity is L_x = 1.02 × 10⁴⁵ erg s⁻¹, with a double-peaked profile suggestive of a merging cluster (Mann & Ebeling 2012). The velocity dispersions for each of these components are σ = 779${}_{-20}^{+22}$ and 955${}_{-22}^{+17}$ km s⁻¹ (Ebeling et al. 2014; Jauzac et al. 2014), and the total mass enclosed within 950 kpc ∼1.2 × 10¹⁵ M_⊙ (Jauzac et al. 2014; Grillo et al. 2015). MACSJ0416.1-2403 was selected as one of five strong-lensing clusters for the CLASH HST MCT survey (Postman et al. 2012) based on its large Einstein radius (θ_E > 035 at z = 2). Prior to the Frontier Fields observations, Zitrin et al. (2013) found a high number of multiple images relative to its critical area in the CLASH HST images, likely due to its highly elongated and irregular structure.

Zoom In Zoom Out Reset image size

Preliminary evaluation of MACSJ0416.1-2403's lensing models yielded moderate to high probabilities of detecting a $z\sim 10\,H\leqslant 27$ mag galaxy within the WFC3/IR field of view (Table 2). MACSJ0416.1-2403 is at a high ecliptic latitude with a Galactic extinction $E(B-V)$ = 0.036 (Schlafly & Finkbeiner 2011). The typical zodiacal background in ${H}_{F160W}$ during the cluster IR epoch (7/2014–9/2014) and the parallel IR epoch (1/2014–2/2014) is ∼22.3 and 22.4 ABmag per $\square ^{\prime\prime}$ respectively. At decl. ∼ −24, this field is easily observable with ALMA, and also available to Maunakea. A significant amount of data was collected on this cluster as part of MACS and CLASH, including shallow multi-band HST data, Chandra imaging, Spitzer warm-mission IRAC (PI R. Bouwens), and VLT spectroscopy (e.g., Grillo et al. 2015). Additional Chandra imaging has since been obtained by C. Jones-Forman and S. Murray (Ogrean et al. 2015). However, there are no legacy Spitzer cryogenic observations.

MACSJ0416.1-2403 is notable for having a J = 10, V = 13 magnitude star within 1' of the cluster core. This star has a high proper motion, with DSS and 2MASS imaging from the mid-1990s showing a position a few arcseconds north of its current (2014) HST ACS position. This star is included in the Frontier Fields ACS pointing and lies just off the WFC3/IR pointing, resulting in scattered light and saturated diffraction spikes in the Frontier Field images. However, this star is bright enough to act as a guide star for adaptive optics, and therefore provides a unique opportunity to obtain adaptive optics imaging (e.g., Schrimer et al. 2015; Gemini-GEMS) and spectroscopy of the critical curves surrounding a strong-lensing cluster.

The MACSJ0416.1-2403 parallel field (Figure 4(b), Table 1) was chosen to lie westward of the cluster pointing in order to avoid the bright eastern stars in the Spitzer Frontier Field observations. This orientation is perpendicular to the elongation of the cluster on the sky, and therefore we expect minimal contamination of the parallel field from the cluster. The parallel field is predicted to have median magnification factors ∼1.09–1.16 and maximum magnification factors 1.2–1.4 for 1 < z < 9 within the WFC3/IR pointing based on the pre-HFF v1.0 Merten model (Table 2).

4.3. MACS0717.5+3745

MACSJ0717.5+3745 (Figure 5(a)) is an extremely massive X-ray-luminous merging cluster at z = 0.545 (Edge et al. 2003). The X-ray luminosity between 0.1 and 2.4 keV is (3.3 ± 0.2) × 10⁴⁵ erg s⁻¹ (Edge et al. 2003). The cluster's velocity dispersion is ${1660}_{-130}^{+120}$ km s⁻¹ (Ebeling et al. 2007). Its optical and X-ray morphology shows a double peak and a lack of center cluster core, with a filament extending southeast (Ebeling et al. 2004; Kartaltepe et al. 2008). This cluster also hosts the most powerful known radio source (P_{1.4 GHz} ∼ 5 ×10²⁵ W Hz⁻¹) with a radio relic significantly offset from the cluster center to the north (van Weeren et al. 2009). MACSJ0717.5+3745 was also chosen as one of the CLASH strong-lensing clusters (Postman et al. 2012). It has the largest known Einstein radius (∼350 kpc, Zitrin et al. 2009b) and an estimated virial mass ≥2–$3\times {10}^{15}\,{M}_{\odot }$ (Zitrin et al. 2009b; Limousin et al. 2012). Several pointings of HST ACS imaging were obtained previously by Ebeling in Cycle 12 (GO 9722). Weak-lensing analyses of the pre-Frontier Fields HST imaging and ground-based Subaru imaging have confirmed the presence of the southeast filament, with a projected length of ∼4.5 Mpc and true length of ∼18 Mpc (Jauzac et al. 2012; Medenski et al. 2013).

Zoom In Zoom Out Reset image size

Independent preliminary lensing models from Zitrin and Richard ranked MACS0717.5+3745 as the strongest lenser of all the clusters considered (see Table 2). However, MACSJ0717.5+3745 has the highest zodiacal background of all the Frontier Fields, as well as a relatively high Galactic extinction ${E}_{(B-V)}=0.068$ (Schlafly & Finkbeiner 2011). It has an ecliptic latitude of 154, with a typical zodiacal background in ${H}_{F160W}$ during the cluster IR epoch (2/2015–3/2015) and the parallel IR epoch (9/2014–12/2014) of ∼22.0 and 21.8 ABmag per $\square ^{\prime\prime}$ respectively. It is also our northernmost cluster at decl. >30°, placing it just out of reach of ALMA and other southern observatories. As a CLASH cluster, significant shallow HST imaging, ancillary wide-field Subaru imaging, and a photometric redshift catalog are available. This cluster was also observed with the Spitzer cryogenic mission with both IRAC and MIPS (PI D. Kocevski) and the Spitzer warm-mission SURFSUP program (PI M. Bradac̆; Bradac̆ et al. 2014), as well as with the Herschel Space Observatory (Egami et al. 2010). A spectroscopic redshift catalog was recently published by Ebeling et al. (2014).

The MACSJ0717.5+3745 parallel field (Figure 5(b)) was chosen to lie northwest of the cluster pointing in order to avoid the long cluster filament extending to the southeast. The parallel field is predicted to have median magnification factors ∼1.07–1.15 and maximum magnification factors 1.17–1.42 for 1 < z < 9 within the WFC3/IR pointing based on the pre-HFF v1.0 Merten model (Table 2).

4.4. MACS1149.5+2223

MACSJ1149.5+2223 (Figure 6(a)) at z = 0.543 was discovered as part of the MACS survey as one of the most X-ray-luminous clusters known at z > 0.5 (Ebeling et al. 2001, 2007). Its 0.1–24 keV X-ray luminosity is L_x = (1.76 ± 0.04) × 10⁴⁵ erg s⁻¹ and it has a velocity dispersion of ${1840}_{-170}^{+120}$ km s⁻¹ (Ebeling et al. 2007). Its optically selected galaxy population and X-ray morphology are elongated within the cluster core, but do not show evidence of extended filaments (Kartaltepe et al. 2008). Spectroscopic studies and lensing analysis of previous HST ACS imaging (PI H. Ebeling; GO 9722) suggest four or more large-scale dark-matter subhaloes and a complex merger history (Zitrin & Broadhurst 2009; see also Smith et al. 2009). A CLASH strong-lensing cluster (Postman et al. 2012), it has a large Einstein radius (∼170 kpc, Zitrin & Broadhurst 2009) and an estimated total mass $\sim 2.5\times {10}^{15}\,{M}_{\odot }$ (Zheng et al. 2012). Based on the CLASH imaging, Zheng et al. (2012) reported a singly imaged z = 9.6 galaxy candidate with a magnification ∼14.5 and observed F160W magnitude ∼26.5.

Zoom In Zoom Out Reset image size

Preliminary lensing models from Zitrin and Richard ranked MACSJ1149.5+2223 as a moderate lenser (Table 2). Its Galactic extinction is fairly low, E(B − V) = 0.020 (Schlafly & Finkbeiner 2011), and it has a zodiacal background $\sim 22{H}_{F160W}$ ABmag per $\square ^{\prime\prime}$ during the epochs of observation (cluster IR: 11/2014–1/2015; parallel IR: 4/2015–5/2015). Initially, this cluster was not considered an ideal HST target because only one bright guide star was known at the required orientation. However, further investigation revealed a second guide star slightly fainter than the nominal magnitude cutoff, and early observations of MACSJ1149.5+2223 in Cycle 21 confirmed the suitability of this guide star pair. At decl. +22°, this cluster is barely observable with ALMA but easily observed from Maunakea and other northern observatories like the Very Large Array. This cluster is part of the Herschel Lensing Survey (Egami et al. 2010) and a GT Cycle 1 program (PI D. Lutz), and was targeted by the Spitzer warm-mission SURFSUP IRAC imaging program (Bradac̆ et al. 2014).

The southern position for the MACSJ1149.5+2223 parallel field (Figure 6(b), Table 1) was chosen to avoid a particularly bright star at the northern position. The parallel field is predicted to have median magnification factors ∼1.02–1.07 and maximum magnification factors 1.1–1.3 for 1 < z < 9 within the WFC3/IR pointing based on the pre-HFF v1.0 Merten lensing model (Table 2).

4.5. Abell S1063

4.6. Abell 370

4.7. Other Cluster Candidates

Deep optical and near-infrared imaging achieving 5σ depths of ∼29th AB magnitude in seven HST bandpasses (ACS/WFC ${B}_{F435W}$, ${V}_{F606W}$, ${I}_{F814W}$, WFC3/IR ${Y}_{F105W}$, ${J}_{F125W}$, ${{JH}}_{F140W}$, ${H}_{F160W}$) from 0.4–1.6 μm is used to identify high-redshift galaxies (z > 4) using the Lyman break drop-out technique (Table 3, Figures 2–6). Deep Spitzer IRAC imaging at 3.6 and 4.5 μm places additional constraints on galaxy redshifts (Table 3, Figure 7). Fitting of the spectral energy distribution of the multi-wavelength photometry from the combined HST and Spitzer imaging (e.g., Castellano et al. 2016; Merlin et al. 2016) provides photometric redshifts and estimates of the galaxies' stellar masses and recent star formation histories (e.g., Castellano et al. 2016).

Zoom In Zoom Out Reset image size

Table 3.  Frontier Fields Target Depths

Camera/Filter	Exposure Timea	5σb
HST ACS/WFC F435W	45 ks	28.8
HST ACS/WFC F606W	25 ks	28.8
HST ACS/WFC F814W	105 ks	29.1
HST WFC3/IR F105W	60 ks	28.9
HST WFC3/IR F125W	30 ks	28.6
HST WFC3/IR F140W	25 ks	28.6
HST WFC3/IR F160W	60 ks	28.7
Spitzer IRAC 3.6 μm	180 ks	26.5
Spitzer IRAC 4.5 μm	180 ks	26.0

Notes.

^aAssuming 2500 s per HST orbit. Spitzer depths include previous archival observations. ^bCalculated for a point source within a 04 diameter aperture for HST.

Download table as:  ASCIITypeset image

The Frontier Field cluster observations have the same exposure times as the parallel fields, and similar observed depths. However, the intrinsic depths for background galaxies lensed by the clusters are deeper than the parallel fields (considering the contribution to the foreground by the intracluster light and galaxies in the cluster; see Livermore et al. (2017) and Merlin et al. (2016) for strategies for subtracting the intracluster light), with typical magnifications across the cluster pointings of ∼1.5–2 and small areas magnified by factors as large as 10–100.

5.1. HST Observing Strategy

5.2. HST Data Reduction

5.3. Spitzer Observations

In order to enable study of background lensed galaxies by a broad cross section of the extragalactic community, the HST Frontier Fields team has also supported the development and public release of lensing maps for each selected cluster. In particular, magnification estimates for the high-redshift lensed galaxies are critical for determining their intrinsic luminosities and estimating the faint end of the UV luminosity function. The initial lensing models were based on data taken before the Frontier Fields observing campaign to ensure that the community could make use of the Frontier Fields data as soon as possible (Table 4). Five independent teams (Bradac̆; Clusters As Telescopes, PI J.P. Kneib & P. Natarajan; A. Zitrin & J. Merten; K. Sharon; L. Williams), using a diversity of approaches (Bradac̆ et al. 2005; LENSTOOL: Jullo & Kneib 2009; Zitrin et al. 2009; Merten et al. 2009; GRALE: Mohammed et al. 2014), coordinated to adopt the same input archival HST and ground-based data sets, the same redshifts, and multiple image identifications (Tables 4, 5). These models were made public on MAST prior to the HST Frontier Fields observations in autumn 2013.¹⁷ The initial pre-FF model predictions for the galaxy numbers and volumes probed at high redshift are described in Coe et al. (2015).

However, these first pre-FF models have been rapidly superseded. The deep HST data have resulted in an unprecedented set of strong-lensed arcs and multiple images for constraining the cluster potentials (e.g., Jauzac et al. 2014, 2015; Lam et al. 2014; Diego et al. 2015, 2016; Wang et al. 2015; Kawamata et al. 2016), doubling or tripling the number of multiple images (Table 5, Figure 8). Subsequent observations with the GLASS HST WFC3/IR grism GO program (Schmidt et al. 2014; Treu et al. 2015; see 10.17909/T9KG60) and new ground-based spectroscopic campaigns have greatly increased the number and accuracy of the redshifts for the background lensed FF galaxies (Ebeling et al. 2014; Johnson et al. 2014; Richard et al. 2014; Balestra et al. 2015; Grillo et al. 2015; Hoag et al. 2016). The detection of a lensed SN Ia in MACSJ0416.1-2403 has also provided a strong constraint on its true magnification (Rodney et al. 2015). The discovery of the multiply imaged SN Refsdal in MACSJ1149.5+2223 (Kelly et al. 2015) sparked an independent coordinated effort to predict the time delays and reappearance of this supernova in another image of the host galaxy (Kelly et al. 2016; Treu et al. 2016; also Rodney et al. 2016).

Zoom In Zoom Out Reset image size

The newly discovered arcs and new spectroscopic redshifts have been incorporated into updated HFF+ versions of the Abell 2744 and MACSJ0416.1-2403 lensing models (Table 5); many of these have ${z}_{\mathrm{phot}}\gt 3$ (see Figure 8 for a comparison of the arc redshift distributions adopted by the pre-HFF and new HFF+ lensing models; Cypriano et al. 2004; Okabe & Umetsu 2008; Zitrin et al. 2009; 2013; Okabe et al. 2010a, 2010b; Merten et al. 2011; Christensen et al. 2012; Mann & Ebeling 2012; Jauzac et al. 2014; Lam et al. 2014; Richard et al. 2014; Balestra et al. 2015; Diego et al. 2015; Grillo et al. 2015; Jauzac et al. 2015; Rodney et al. 2015; Wang et al. 2015; Kawamata et al. 2016). The incorporation of these new multiple image systems often results in a reduction in the statistical uncertainty in the galaxy magnifications for a given model. All of the public HFF lensing models provide a range of possible realizations from which the statistical uncertainty of a given model set may be calculated (typically 100 but no fewer than 30). We plot the cumulative distribution of the galaxy magnification uncertainties σ(model)/$\langle \mu (\mathrm{model})\rangle$, for the galaxies and photometric redshifts provided by the ASTRODEEP catalogs (Merlin et al. 2016; Castellano et al. 2016a) for Abell 2744 (Figure 9) and MACSJ0416.1-2403 (Figure 10). Generally, the statistical uncertainties are reduced for the models computed with the new HFF data sets, with more dramatic reductions for the methods that rely strongly upon the strong-lensing constraints. The parametric methods (CATS, Sharon, Zitrin, GLAFIC) report median statistical magnification errors of 0.2%–5%, while the non-parametric methods (Bradac̆ Williams, Diego) report median statistical magnification errors of 2%–11% for the post-HFF calculations (green curves), versus 2%–22% and 2%–17% respectively for pre-HFF models (blue curves). (We note that the statistical errors for the MACSJ0416.1-2403 Bradac̆ post-HFF models (Hoag et al. 2016) included additional uncertainties due to the photometric redshift uncertainties of the multiple images. These were not included in the pre-HFF Bradac̆ model, and thus may explain why the post-HFF statistical errors are larger for this model.)

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We also examine the systematic uncertainties associated with the different model approaches by computing the method-to-method standard deviations for the CATS, Sharon, Zitrin NFW, Williams, and Bradac̆ "best" magnification values for each input galaxy (first panels in Figures 9, 10). This model subset was chosen because they encompass the same methodologies for both pre-HFF and HFF+ models. We find that the additional constraints from the HFF data have done little to reduce the variations in magnification values across methodologies. The median systematic magnification uncertainties are less than 26% for Abell 2744 and 15% for MACSJ0416.1-2403 for the post-HFF models.

Additional programs have sought to understand and improve the systematics inherent in the lensing models (e.g., Zitrin et al. 2015a; Harvey et al. 2016; Johnson & Sharon 2016; Meneghetti et al. 2016; Mohammed et al. 2016). Using two simulated Frontier Fields-like clusters and data sets, Meneghetti et al. (2016) study the systematic effects of the lensing modeling approaches used by the Frontier Fields lensing teams. They find that magnifications within the Einstein radii are well constrained, with the largest uncertainties in the magnifications arising near the critical curves and substructures, and best parametric results yielding uncertainties relative to the input simulation of ∼10% for true magnifications ∼3 and 30% for true magnifications ∼10. With this simulation, Johnson & Sharon (2016) explore the systematic errors of parametric LENSTOOL models (Jullo & Kneib 2009) as a function of the number of multiple image systems with and without spectroscopic redshifts. They find that increasing the number of multiple image systems to 20 improved the accuracy in the magnification estimates, but the accuracy was not improved beyond 2% for N > 20 systems. Rather, they find that for a fixed number of multiple images, increasing the fraction of systems with spectroscopic redshifts increased the accuracy of the magnification estimates without changing the precision. Given the small fraction of systems with spectroscopic redshifts for the current HFF multiple image systems (currently ∼15% for Abell 2744 and MACSJ0416.1; Table 5), we expect that additional improvements in the model accuracy could be made with additional spectroscopic follow-up.

The Frontier Fields lensing models will continue to be refined, as the HFF observing program proceeds through September 2016, new ancillary spectroscopic and weak-lensing data sets are acquired, and the modeling methods improve. This investment is critical for ensuring the Frontier Fields' legacy for JWST studies. To continue to provide the best models to the broader community, a renewed effort to update the existing lensing models and incorporate new FF and ancillary data began in May 2015 for Abell 2744 and MACSJ0416.1-2403 (Table 4). The resulting models were publicly released in autumn 2015. A second round of lensing coordination is set to begin in summer 2016, and will encompass the last four clusters. The delivery of MACSJ1149.5+2223 and MACSJ0717.5+3647 models is due in February 2017, with final delivery of Abell S1063 and Abell 370 models due in September 2017.

In this section, we demonstrate the exceptional sensitivity of the Hubble Frontier Field images to faint and distant galaxies. We summarize the key early results from the current literature that address the primary goals of the HFF (Section 2), including the detection of new deep samples of intrinsically faint galaxy candidates during the epoch of reionization, dramatically improved constraints on the faint end of the UV luminosity function at z > 6, initial estimates of the sensitivity of the HFF as a function of stellar mass and star formation rates, and new understanding of the spatially resolved properties of distant HFF galaxies important for planning observations with JWST. We focus primarily on the studies of the first two Frontier Fields, Abell 2744 and MACSJ0416.1-2403, and make extensive use of the public ASTRODEEP multi-band photometric catalog and the catalog fitted to the spectral energy distribution (Castellano et al. 2016a; Merlin et al. 2016).

7.1. HST Detection Limits

The HFF v1.0 images of the first two clusters and their parallel fields have achieved the target depths presented in Table 2. Merlin et al. (2016) compute the 5σ detection limits for a point source within a 04 diameter aperture using the image rms values for these four sets of 60 mas pixel-scale images in the seven HST  bands. These are 28.6–29.0 for ${B}_{F435W}$, 28.7–29.0 for ${V}_{F606W}$, 29.0–29.3 for ${I}_{F814W}$, 29.15–29.3 for ${Y}_{F105W}$, 28.8–29.0 for ${J}_{F125W}$, 28.9–29.1 for ${{JH}}_{F140W}$, and 29.0–29.1 for ${H}_{F160W}$ (see Merlin et al. 2016, Table 1). We note that the total exposure times for these images are generally longer than that assumed in Table 2. The HFF program images were combined with archival HST data from the Dupke (11689), Siana (13389), Rodney (13386), and Postman (12459) programs when available. An additional 10 orbits of MACSJ0416.1-2403 (WFC3/IR: parallel, ACS/WFC: cluster) were reobserved due to severe persistence in the IR images.

We compare the 5σ detections of the HFF MACSJ0416.1-2403 cluster and parallel images to the CLASH MACSJ0416.1-2403 images, and the deepest available versions of the HUDF (ACS/WFC: Illingworth et al. 2013; WFC3/IR: Koekemoer et al. 2013) in Figures 11 and 12. The total exposure times for each filter and image are provided in Table 4. We plot the number counts as a function of AB magnitude measured within 04 diameter apertures for each of the HST bandpasses; these magnitudes are corrected for the expected flux loss within this aperture for a point source (−0.18 mag for ${B}_{F435W}$, ${V}_{F606W}$, −0.19 mag for ${I}_{F814W}$, −0.37 for ${Y}_{F105W}$, −0.42 ${J}_{F125W}$, −0.47 for ${{JH}}_{F140W}$, −0.48 for ${H}_{F160W}$). (Note that these number counts of the aperture magnitudes do not reflect the total magnitude number counts; refer to Merlin et al. 2016 for this analysis.)

Zoom In Zoom Out Reset image size

The aperture photometry is computed with SExtractor v2.8.6 (Bertin & Arnouts 1996) in single-band mode, using a 1σ detection and analysis threshold and object deblending settings DEBLEND_NTHRESH = 64, DEBLEND_MINCONT = 0.0001. The images are convolved with a four-pixel Gaussian filter as part of the SExtractor detection. For the ACS images, we analyze the 30 mas pixel-scale images and the HFF self-calibrated versions and adopt DETECT_MINAREA = 50 pixels. For the WFC3 images, we analyze the 60 mas pixel-scale images and the HFF self-calibrated versions and adopt DETECT_MINAREA = 12.5 pixels. We use SExtractor for the background subtraction with BACKGROUND_SIZE set to 3.8 arcsec (128/64 pixels for 30/60 mas pixel scales), BACKGROUND_FILTERSIZE = 5, BACKPHOTO_TYPE = LOCAL, and the annulus for the photometric background correction BACKPHOTO_THICK = 1.44 arcsec (48/24 pixels for 30/60 mas pixel scales).

We find that the faint limit for 5σ detections is in good agreement with the target HFF depths (vertical dashed lines, Figures 11, 12). These are substantially deeper than the 1–2 orbit CLASH images, as expected given the factor of 10 difference in the exposure times. The HFF WFC3/IR images are ∼1 mag shallower than the HUDF12 WFC3/IR images; thus faint high-redshift HFF lensed galaxies require magnification factors >2.5 in order to probe intrinsic magnitudes fainter than unlensed galaxies in the HUDF12 images. The HFF ACS/WFC ${B}_{F435W}$ images are slightly shallower (0.1–0.3 mag) than the HUDF ACS/WFC ${B}_{F435W}$ image. The HFF ACS/WFC ${V}_{F606W}$ images are ∼0.5 magnitudes shallower than the HUDF ACS/WFC ${V}_{F606W}$ images. The HUDF ACS/WFC ${I}_{F814W}$ image has less than half the exposure time of the HFF images, thus is ∼0.5 magnitudes shallower. The 5σ limits seen here in the number counts for the HUDF and MACSJ0416.1-2403 parallel fields are in good agreement with the 5σ depths reported in Finkelstein et al. (2015, Table 1).

Zoom In Zoom Out Reset image size

For the cluster WFC3/IR images, there is evidence for lower detection rates than in the parallel field at ${H}_{F160W}\sim 28.5$ ABmag (first reported for Abell 2744 by Ishigaki et al. 2015, Figure 1). The completeness of object detections in the cluster fields is a strong function of the algorithms used to deblend the cluster galaxies and correct for the intracluster light; the additional rms noise from the cluster also degrades the detection limit (see McLeod et al. 2016). Different groups have employed increasingly aggressive strategies to detect faint sources within the core of the HFF clusters. Early works (Atek et al. 2014; Ishigaki et al. 2015; Laporte et al. 2014; Zheng et al. 2014) used SExtractor for deblending, source detection, background correction, and galaxy photometry. In order to correct for the intracluster light, Oesch et al. (2014) subtracted a median-filter smoothed background image prior to SExtractor photometric calculations, and computed completeness estimates >60% for H = 25–28 high-redshift galaxies with magnification factors ≤10. More recently, Merlin et al. (2016) have used the galaxy profile fitting software GALFIT (Peng et al. 2010) to subtract cluster galaxies and enable the detection of underlying faint sources, and to compute completeness efficiencies as a function of profile type (point source, disk, bulge) and magnitude, with 50% completeness at ${H}_{F160W}\sim 28.5$, 27.7, and 27.3 respectively. Finally, Livermore et al. (2017) have devised a wavelet approach that decomposes the images into Fourier modes and detects objects on the small spatial scales; this method is able to detect compact simulated galaxies to 28.5–29th magnitude at the 5σ level near the cores of Abell 2744 and MACSJ0416.1-2403 (Livermore et al. 2017; Figure 1).

7.2. High-redshift Galaxy Selection

The deep multiband HST and Spitzer imaging has enabled the selection of high-redshift galaxy candidates using both traditional Lyman break color–color selections and photometric redshift estimates. Uncertainties in the magnifications by the cluster do not affect the observed colors or the photometric redshift uncertainties. Castellano et al. (2016a) compute photometric redshifts using the full HST, Spitzer, and VLT Hawk-I K-band imaging for galaxies and report typical $\delta z/(1+z)$ errors of 4%, based on the available spectroscopic samples. In Figure 13, we show the photometric redshift distributions of the HFF MACSJ0416.1-2403 cluster and parallel field from Castellano et al. (2016a) versus the photometric redshift distributions computed from the CLASH observations of the MACSJ0416.1-2403 cluster (Postman et al. 2012; Balestra et al. 2015; Grillo et al. 2015; see also Castellano et al. 2016a, Figures 2 and 5). While CLASH detected one z > 7.5 candidate in the MACSJ0416.1-2403 cluster field (Bradley et al. 2014, not reported by Castellano et al. 2016a), the HFF observations find at least 18 z > 7.5 candidates.

Zoom In Zoom Out Reset image size

The current numbers of galaxy candidates during the epoch of reionization (6 < z < 10) reported by various groups in the literature for MACSJ0416.1-2403 and Abell 2744 clusters and parallel fields are consistent with the predictions by Coe et al. (2015), using the pre-HFF lensing models (e.g., Atek et al. 2015; Finkelstein et al. 2015; Castellano et al. 2016b; Livermore et al. 2017; McLeod et al. 2016). The inclusion of the WFC3/IR ${{JH}}_{F140W}$ has confirmed the detection of at least 33 z > 8.5 candidates in the first four Frontier Field cluster and parallel fields (e.g., McLeod et al. 2015, 2016). However, few of the z > 6 HFF high-redshift candidates have been confirmed spectroscopically yet. The GLASS HST WFC3/IR grism observations report nine Lyα emitters at z > 7 in the HFF cluster fields (Schmidt et al. 2016). Ground-based observations have been more successful at detecting z < 6 lensed emission-line galaxies (e.g., Karman et al. 2017; Vanzella et al. 2016). Deep LBT observations have confirmed z = 6.4 Lyα emission for a multiply imaged, highly magnified ${M}_{{UV}}\sim -18.7$ system in MACSJ0717.5+3745 (Vanzella et al. 2014). However, deep Keck searches for C iii] emission in Abell 2744 z ∼ 7–8 have been unsuccessful thus far (e.g., Zitrin et al. 2015b); systematic follow-up of many high-redshift HFF candidates may require the capabilities of JWST and ALMA.

7.3. Intrinsically Faint Galaxy Populations at z > 6

We present the motivation and survey design for the Frontier Fields, a Director's discretionary time program with HST and Spitzer to see deeper into the distant universe than ever before. Six strong-lensing clusters and six parallel fields are observed, probing galaxies to observed optical/near-infrared magnitudes of ∼29 ABmag and 10–100 times fainter in regions of high magnification. We explain the primary scientific goals of the Frontier Fields, the selection criteria for the fields, and the detailed properties of each Frontier Field cluster and parallel. We describe the HST and Spitzer observing programs, and the coordinated Frontier Fields lensing model effort. We present the initial results from the first HFF observations, and demonstrate that the HFF is achieving its primary science goals regarding high-redshift galaxies.

The HST Frontier Fields observations of the last cluster (Abell 370) and its parallel field will be completed in September 2016, and the coordinated lensing models will be updated in 2017–2018. The full Spitzer Frontier Fields observations are complete and were publicly released in early 2016. The first Frontier Fields observations have already probed galaxies during the epoch of reionization to intrinsic luminosities fainter than any previously seen (e.g., Zitrin et al. 2014; Atek et al. 2015; Laporte 2015; Castellano et al. 2016; Livermore et al. 2017) and improved our statistical accounting of z > 8 galaxies (e.g., Kawamata et al. 2015; McLeod et al. 2015, 2016; Ishigaki et al. 2016). The full data set will place strong statistical constraints on the faint end of the luminosity function during this era (Robertson et al. 2015). At the time of publication of this article, over 80 refereed publications and four conferences have been devoted to or based in part on the Frontier Fields. These works include studies of high-redshift galaxies in the cluster and parallel fields, new cluster lensing models and dark-matter maps, supernovae/transient studies, studies of intracluster light and cluster evolution, and ancillary observations probing highly lensed background sources with major ground-based facilities. These data and associated models will provide a unique legacy for future studies of the high-redshift universe with the JWST.

The Frontier Fields program was initiated by STScI Director Dr. Matt Mountain using Director's discretionary time on the Hubble Space Telescope. We wish to acknowledge the Hubble Deep Fields Initiative science working group members for conceiving and recommending the Frontier Fields program: J. Bullock (chair), M. Dickinson, S. Finkelstein, A. Fontana, A. Hornschemeier Cardiff, J. Lotz, P. Natarajan, A. Pope, B. Robertson, B. Siana, J. Tumlinson, and M. Wood-Vasey. We also thank the mid-term Frontier Fields review committee for their service: J Bullock, M. Dickinson, R. Ellis, M. Kriek, S. Oey, S. Seitz, S. A. Stanford, and J. Tumlinson. We recognize the contributors to the current Frontier Field lensing models: M. Bradac̆, S. Allen, D. Applegate, B. Cain, A. Hoag, P. Kelly, P. Schneider, T. Schrabback, T. Treu, A. von der Linden, J.-P. Kneib, P. Natarajan, H. Ebeling, J. Richard, B. Clement, M. Jauzac, E. Jullo, M. Limousin, E. Egami, J. Merten, A. Zitrin, I. Balestra, M. Bartelmann, N. Benitez, A. Biviano, T. Broadhurst, M. Carrasco, D. Coe, N. Czakon, M. Donahue, T. Eichner, R. Ellis, C. Giocoli, S. Golwala, C. Grillo, O. Host, L. Infante, S. Jouvel, D. Lemze, A. Mercurio, E. Medezinski, P. Melchior, A. Molino, M. Meneghetti, A. Monna, J. Moustakas, L. Moustakas, T. Mroczkowski, M. Nonino, M. Okabe, M. Postman, J. Rhodes, P. Rosati, J. Sayers, S. Seitz, K. Umetsu, K. Sharon, T. Johnson, M. Bayliss, L. Wiliams, I. Mohammed, P. Saha, J. Liesenborgs, K. Sebesta, M. Ishigaki, R. Kawamata, M. Oguri, J. M. Diego, D. Lam, and J. Lim. Finally, we thank David Adler, George Chapman, Bill Workman, Ian Jordan, Alan Welty, Karen Levay, Scott Fleming, Brandon Lawton, Carol Christian, Tony Darnell, Frank Summers, Kathy Cordes, Bonnie Eisenhamer, Lisa Frattare, Ann Jenkins, Hussein Jirdeh, John Maple, Holly Ryer, Ray Villard, Tracy Vogel, and Donna Weaver for their contributions to the HST Frontier Fields effort. We acknowledge the ASTRODEEP team for their public Frontier Fields galaxy catalogs for Abell 2744, MACSJ0416.1-2403, and associated parallel fields.

Based on observations obtained with the NASA/ESA Hubble Space Telescope, retrieved from the Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute (STScI). STScI is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5-26555. This work is based in part on observations made with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA. This work utilizes gravitational lensing models produced by PIs Bradac̆, Natarajan & Kneib (CATS), Merten & Zitrin, Sharon, Williams, and the GLAFIC and Diego groups. This lens modeling was partially funded by the HST Frontier Fields program conducted by STScI. We acknowledge ESO VIMOS CLASH-VLT Large Programme (186.A-0798, PI P. Rosati). Based on data collected at Subaru Telescope by PI K. Umetsu and archival imaging obtained from Subaru–Mitaka–Okayama–Kiso Archive (SMOKA), which is operated by the Astronomy Data Center, National Astronomical Observatory of Japan.