Hubble Boldly Goes: The Frontier Fields Program

Jennifer Lotz,, I. Neill Reid,, and Ken Sembach,


The Hubble Frontier Fields (HFF) program is a Director’s Discretionary Time (DDT) campaign to observe 4–6 strongly lensing galaxy clusters and, in parallel, adjacent “blank” fields with the aim of detecting and characterizing high-redshift galaxies. The program will capitalize on the lensing capabilities of the clusters to probe redshifts beyond z ~ 10, reaching galaxies with intrinsic luminosities thirty times fainter than those detected in the HUDF12. The associated blank fields will have sensitivity comparable to the parallel fields of the HUDF09, and will expand the areal coverage by a factor of three. Together, these measurements will enable Hubble to provide a preliminary peek at the distant universe accessible to the James Webb Space Telescope.


Deep-field imaging designed to probe galaxy formation and evolution in the early universe is now well established as a key theme of the Hubble science program. Such was not always the case. Support for the original HDF DDT program was far from unanimous within the astronomical community. Some felt that the observations would offer little insight into galaxy formation, which many contemporary theoretical models predicted to be a sedate, gradualist process. Others worried that, so soon after the correction of Hubble’s vision at considerable tax-payer expense, there might be Congressional repercussions from investing ten days of Hubble time on a single project with a dubious prospect of returns.  Those fears proved unfounded; the HDF revealed galaxy assembly to be an active, dynamic process—and provided iconic images that have now permeated the public consciousness. The original HDF together with Keck 10-m spectroscopy of the brightest 125 galaxies in the field validated the concept of photometric redshifts that have now succeeded in opening distant objects to analysis via multi-bandpass imaging.

The HDF lies at northern declination. Following its success, Hubble compiled matching observations of a field in the southern sky, the HDF-S. Since then, as each new servicing mission enhanced its scientific capabilities, Hubble has devoted considerable time and resources to deep-imaging programs that span a range of depth and areal coverage. Those programs include GOODS (the southern field within the CDF-S), the HUDF (within the CDF-S), COSMOS, HUDF09, CANDELS, UDF12, and 3D-HST. As with the HDF, each Hubble deep-imaging program has been supported by extensive complementary observations from other space observatories, notably Spitzer, Chandra and Herschel, together with substantial photometric and spectroscopic contributions from ground-based facilities. All told, over 3,000 orbits of Hubble observations (approximately one cycle) have been invested in major deep-field survey programs, with more than 800 orbits devoted to the 15-square-arcminute UDF alone. This wide panoply of multi-wavelength observation has revolutionized our view of the universe: z ~ 1 galaxies are demoted to “low-redshift” systems, the z ~ 2–3 peak in star formation now lies at “moderate redshifts,” and detection limits have been pushed through the era of reionization to the brink of cosmic dawn at z ~ 10–12.


The Hubble Deep Field Initiative

Given the substantial advances made in this field over the past 15 years—and the costs—can Hubble offer the prospect of further transformative science from another deep-field survey? Or should we sit back, focus on consolidating our gains, and wait for Webb to bring the next breakthrough? That, in essence, was the question that prompted the Hubble Deep Field Initiative (HDFI). Following discussions with the Space Telescope Users Committee and the community, the Institute Director, Matt Mountain, chartered a science working group (SWG) chaired by James Bullock (UC Irvine) to examine how Hubble might extend our knowledge of the cosmic frontiers at high redshift. The SWG members were Mark Dickinson (NOAO), Steve Finkelstein (UT Austin), Adriano Fontana (INAF), Ann Hornschemeier-Cardiff (GSFC), Jennifer Lotz (STScI), Priya Natarajan (Yale), Alexandra Pope (UMass), Brant Robertson (UA), Brian Siana (UC Riverside), Jason Tumlinson (STScI), and Michael Woods-Vasey (Pittsburgh). As a secondary goal, the committee was asked to consider observations that would lay the groundwork for future Webb observations of the early universe.

Formally, the HDFI SWG was charged with the following tasks:

  • Define the science case and a set of science goals for a new set of ultra-deep-imaging fields with sensitivity depths comparable to those of the HUDF and the HUDF09 infrared follow-up. Provide an assessment of the urgency of pursuing this science.
  • Assess the prospects for near-field science that can be achieved with these deep-field observations.
  • Recommend the locations and number of fields that should be obtained to meet the science goals defined for the HDFI.
  • Recommend the suite of filters and exposure times necessary to accomplish the science goals defined for the HDFI.
  • Solicit input from the astronomical community in defining the science goals and recommendations described in the above tasks.
  • Produce a short (10–15 page) white paper describing the results of the above tasks by October 1, 2012.

The HDFI SWG was also constrained to identify a program that required no more than 800–1,000 orbits, which (like the HDF) could therefore be accommodated through DDT without impinging on GO allocations in Cycle 21 and succeeding cycles.

At the outset of the process, the community was asked to weigh in on the HDFI, and 32 white papers were received outlining a variety of techniques, programs, and constraints that the community felt should be taken into account. The SWG engaged in lively discussions through the summer and fall of 2012, thoroughly covering the pros and cons of a wide variety of programs, including duplicating the UDF at a different location, probing deeper within the UDF itself, adding deep grism observations in selected fields, obtaining blue/ultraviolet data in selected fields as a precursor for Webb, and using galaxy clusters as telescopes to probe the high-redshift universe. Those discussions resulted in a unanimous recommendation to the Director, and were summarized in the SWG’s report, submitted on November 21, 2012. The HDFI SWG report is available here.

After due deliberation, the Director decided to go ahead with the program, renaming it “Hubble Frontier Fields.” The observations of four galaxy clusters will be executed in Cycles 21 and 22, and, dependent on the results of an interim review, with observations of the final two clusters in Cycle 23. The HFF decision was announced in the Cycle 21 Call for Proposals, issued on December 5, 2012.


The HFF program

The HDFI SWG recommended that the Institute director should pursue a joint strategy of deep imaging on six strongly lensing galaxy clusters, together with parallel observations of adjacent “blank” fields (Figure 1). Quoting from the report,

[The HFF program] combines proven techniques for studying high-redshift galaxies in blank fields with the potentially revolutionary use of natural gravitational telescopes to exploit their magnification of the faintest galaxies in the distant universe. The blank fields will increase three-fold the area covered at comparable depth by the HUDF09 and its parallel fields, tracing the history of star formation and the growth of stellar mass with improved statistics and reduced cosmic variance. In the cluster fields, HST can reach high redshift galaxies as faint intrinsically as those that JWST can detect in blank fields, even down to the dwarf galaxies thought to be the progenitors of typical L_* galaxies in the modern universe.

The proposed observations reach magnitude limits comparable to the HUDF09 parallel fields, matching the second deepest images that we currently have for the high-redshift universe. The blank-field observations will therefore provide important constraints on cosmic variance at redshifts z ~ 6–8. Observations in the field have pushed to galaxies at redshifts z ~ 8–10. Nevertheless, it is important to recognize that those galaxies represent the tip of the luminosity function at those redshifts. The cluster fields imaged by the HFF program will not necessarily reveal more high-redshift galaxies, but they offer the potential to detect the intrinsically fainter galaxies that are the progenitors of galaxies like the Milky Way in the local universe. Moreover, intrinsically luminous galaxies might be amplified to the point where spectroscopic observations with Hubble or Webb become possible.

The HDFI WG made a number of recommendations regarding the development of the observing program. In particular, they identified the following criteria for selecting the target clusters:

  • The clusters must be massive and among the strongest lenses known, with high magnification caustics for objects at z ~ 4–10 that fit within the WFC3-IR field of view. This requirement is most easily met by clusters at z > 0.37.
  • The clusters must be observable with Hubble, Spitzer, and Webb.
  • The cluster fields should have low zodiacal background and low Galactic extinction.
  • To the extent possible, the blank fields should avoid bright stars and outlying structure due to the galaxy cluster.
  • Every effort should be made to choose clusters that are observable by the Atacama Large Millimeter/submillimeter Array and the observatories on Mauna Kea.
  • If possible, ancillary data should be available from Hubble, Spitzer-MIPS, Herschel, Chandra, and ground-based telescopes.

The HDFI WG recognized that satisfying every criterion for every cluster may not be possible.

They noted that analyzing the cluster data depends critically on understanding the cluster magnification maps, and how those translate to the volume probed at high redshift. This is a specialized area of research, and they urged the Institute to provide the appropriate reference data and analysis tools to level the playing field for all potential participants in this community program.


Implementing the HFF program

The responsibility for executing the HFF program rests with a core implementation team of Institute staff led by one of the authors (JL). That team is receiving advice and guidance from a number of external science advisors. From the outset, Spitzer researchers have been closely involved in the HDFI. The Spitzer Director, Tom Soifer, has committed up to 1,000 hours for observations by the InfraRed Array Camera of the target fields, and Spitzer staff members are working closely with the Hubble team to coordinate and implement the observations. In addition, several key members of the community are serving as external science advisors to the program, providing advice throughout.

The Institute has taken steps to involve the community in every aspect of the HFF implementation process. Immediately following the program’s announcement and the release of the HDFI WG report, the community was solicited for additional input on the observing strategy, filter selection, and cluster selection. In addition, members of the community with appropriate expertise were contacted individually and invited to comment on various aspects of the program. Finally, a website and blog are being maintained to keep the community informed of developments as they happen.

The HDFI WG report outlines a potential observing scheme coupling 70 orbits of red/far-red optical observations using the Advanced Camera for Surveys (ACS) with 70 orbits of imaging with the WFC3-IR camera. This scenario achieves magnitude limits matching those in the HUDF09 parallel fields.  Following discussion with the community, those recommendations were modified slightly—adding F140W observations in the blank field—to give the filter selection illustrated in Figure 2. The resulting sensitivities are summarized in Table 1, where the AB magnitude limits represent 5σ detections of a point source, as measured in an aperture 0.4 arcsecond in diameter and corrected to total magnitude. The observations will be taken at two epochs approximately six months apart, with the telescope rotated by 180o between epochs to switch cameras between cluster and blank field. Orientation constraints typically limit the scheduling window at each epoch to 30–50 days. Further details on the observations, including the dithering strategy that will be adopted, can be found at the Frontier Fields website.

The HDF WG provided an initial set of 16 candidate strong-lensing clusters. The implementation team supplemented that list with other candidates suggested by the community and matched their individual properties against the selection criteria defined by the Working Group. Particular attention had to be paid to the availability of suitable offset fields, since even moderately bright stars can cause significant problems in Spitzer IRAC imaging.


After considerable deliberations, the final selections are shown in Figure 3 and Table 2. All are strongly lensing, massive clusters, easily accessible with Hubble, Chandra, Spitzer, and Webb; all have previous Hubble and Chandra observations; four have been observed as part of the Hubble CLASH program. Four of the clusters are accessible by both ALMA and the telescopes on Mauna Kea; MACS0717.5+3745 is too far north for ALMA, and RXJ2248.7-4431, one of the strongest SZ sources detected by the South Pole Telescope, is too far south for observations from Mauna Kea.  All HFF clusters lie in locations that are at least moderately dark, some very dark.

The selected clusters already have sufficient Hubble observations in hand to allow the construction of magnification maps. Making such tools generally available is a priority and, to that end, Hubble issued a Request for Proposals in January of this year. Several responses were received, and contracts are being issued to a number of teams with the requirement that magnification maps for all six clusters are delivered prior to the start of Cycle 21 in October 2013.

All Hubble and Spitzer data taken for the HFF will have no proprietary time and will be available immediately to the community. The core implementation team will combine the Hubble observations to provide higher-level data products, which will be released within ~1 month of the final data acquisition at each epoch. In addition to the new Hubble and Spitzer observations, ancillary data from other observatories will be collected and made available to the community at a central website.


The Cycle 21 program

As announced in the Cycle 21 Call for Proposals, the community was encouraged to submit archival proposals to analyze HFF data, develop supporting theoretical tools, and/or submit observing proposals to obtain supplementary data. Submitted proposals will be reviewed in competition with all other Cycle 21 proposals by the Cycle-21 Time Allocation Committee. By policy, Institute staff members on the core implementation team may not serve as principal investigators on Cycle 21 proposals related to the HFF program, nor could they apply for funding as co-investigators on any such proposals.

The two clusters selected for observation in Cycle 21 are Abell 2744 and MACSJ0416.1-2403 (Figure 4). Abell 2744, also known as Pandora’s Cluster, is a complex, massive system lying at redshift z = 0.308. The cluster has previous Hubble observations (PI R. Dupke: GO 11689) and has been identified as the product of a major merger involving up to four ~1014 MSun separate clusters. Thirty to forty lensed images of background galaxies have already been identified in the system (Merten et al. 2011). To date, this cluster has only ACS observations, and the WFC3-IR imaging offers great potential for discovery. The HFF observations will focus on the southeast component, which is the most massive and well matched in angular size to the WFC3-IR field of view.

The second cluster scheduled for observation in Cycle 21, MACSJ0416.1-2403, lies at redshift z = 0.396 and has also been identified as a merging cluster (Mann & Ebeling 2012). Targeted by the CLASH Multi-Cycle Treasury program, multiply lensed images of more than 20 background galaxies are already known (Zitrin et al. 2013). Both clusters have initial measurements of critical lensing curves and magnification maps, and both can be expected to provide high-magnification sampling of the high-redshift universe. The first HFF observations of both clusters are likely to be obtained in the late fall of this year.


The extensive investment of Hubble resources in deep imaging programs over the last two decades has pushed the high-redshift frontier from z ~ 2 to beyond z = 10. At the same time, the non-proprietary nature of DDT and Treasury programs has ensured rapid access to these data for the entire astronomical community, maximizing the scientific exploitation of the observations. The HFF program represents a continuation of this philosophy, and lays the foundations for future such programs with the James Webb Space Telescope.


Merten, J., et al. 2011, MNRAS, 417, 333
Schlegel, D. J., Finkbeiner, D. P., Davis, M., 1998, ApJ, 500, 525
Zheng, W., et al. 2012, Nature, 489, 406
Zitrin, A., et al. 2013, ApJ, 762, L30

Deep Field Surveys

HDFHubble Deep Field
HDF–SHubble Deep Field-South
GOODSGreat Observatories Origins Deep Survey
HUDFHubble Ultra-Deep Field
COSMOSCosmic Evolution Survey
HUDF09Hubble Ultra-Deep Field 2009
CANDELSCosmic Assembly Near-infrared Deep Extragalactic Legacy Survey
CLASHCluster Lensing And Supernova survey with Hubble
HUDF12Hubble Ultra-Deep Field 2012
CDF–SChandra Deep Field–South
HFFHubble Frontier Fields