Abell 370. Credit: NASA, ESA, J. Lotz, M. Mountain, A. Koekemoer, and the HFF Team (STScI).

Previews of the James Webb Space Telescope: The Frontier Fields Program

Dan Coe, dcoe@stsci.edu


The Frontier Fields program is the latest chapter in Hubble’s hallowed tradition of deep-field initiatives. This time, by combining deep Hubble imaging with gravitational lensing, astronomers have observed the faintest sources ever studied, even fainter than those revealed in the Ultra-Deep Field. This three-year Director’s Discretionary program began in October 2013 and will conclude this coming September. Deep optical and near-infrared imaging of six galaxy clusters and six blank fields with both Hubble and Spitzer is almost complete. Supporting data have also been obtained with Chandra, Subaru, VLT, and other observatories. To date, over 60 papers have studied the observed clusters, supernovae, and distant galaxies out to z ∼ 10, within the universe’s first 500 Myr. The ultra-faint objects revealed in the Frontier Fields are giving us a preview of the universe we will observe with the James Webb Space Telescope.

The Frontier Fields program was described in previous Newsletter articles (Lotz et al. 2013; Ogaz et al. 2015). More details may be found via http://www.stsci.edu/hst/campaigns/frontier-fields/. Meetings to discuss the Frontier Fields have been organized in Yale and Honolulu. Here we provide an update on some of the science results to date, focusing on searches for distant galaxies.

Bold Hubble Deep Fields

The original Hubble Deep Field North (HDF-N) was a bold undertaking by the second Institute director, Bob Williams. It was unclear we would learn much by staring at a blank patch of sky for 10 days. But in 1995, those deep images revealed about 3,000 galaxies, including distant galaxies that were clearly very different from the galaxies we see around us today.

After the success of the HDF-N, Hubble followed with the HDF-S, and then the Ultra-Deep Field (UDF) in ultraviolet, optical, and infrared wavelengths. Each deep image of a blank patch of sky with successively upgraded cameras (WFPC2, NICMOS, ACS, and WFC3) yielded more insights into fainter and more distant galaxies.

However in 2012, with no further servicing missions planned to upgrade Hubble, it was unclear how best to improve on these deep-field programs. Significantly deeper imaging would be prohibitive, requiring too much telescope time. But additional UDFs (similarly deep) would be very useful to increase statistics and overcome cosmic variance.

Then-Institute Director Matt Mountain solicited ideas from the community and convened a committee to deliberate over them. One idea was gravitational lensing. The Cluster Lensing And Supernova survey with Hubble (CLASH; Postman et al. 2012) had recently delivered a candidate for the most distant galaxy known at z ∼ 11, observed 420 Myr after the Big Bang (Coe et al. 2013). This distant galaxy is extremely compact, with a half-light radius less than 100 pc, or roughly the size of giant molecular clouds—star-forming regions in our universe today.

Much deeper imaging of galaxy clusters could reveal lensed populations of dwarf galaxies that were smaller and fainter than any observed before. These dwarf galaxies were likely the dominant source of reionizing flux in the early universe. They were also the building blocks of modern-day galaxies, including our Milky Way. Understanding these distant dwarf galaxies is the frontier of extragalactic research.

Ultimately, the unanimous recommendation from the Hubble Deep-Fields Initiative Science Working Group was the Frontier Fields. Six additional blank deep fields would be observed in parallel with six deep galaxy cluster observations (Figure 1). The directors of both Hubble and Spitzer (Matt Mountain and Tom Soifer) dedicated significant Director’s Discretionary observing time to this program: 840 orbits and 1,000 hours, respectively. With 70 ACS orbits and 70 WFC3/IR orbits per cluster, the infrared Hubble imaging is 10 times deeper than any prior cluster-lensing WFC3/IR images.

Figure 1: MACS J0416.1-2403 and its blank parallel field 6' away. Six pairs of fields like this are being observed by the Frontier Fields program. Credit: NASA, ESA, J. Lotz, M. Mountain, A. Koekemoer, and the HFF Team (STScI).

This is a bold new direction for the Hubble deep-field programs. Staring deeply at massive galaxy clusters represents a significant departure from previous deep-field observations of blank patches of sky. Would ultra-faint distant galaxies get lost in the glare of the foreground galaxy clusters? Would lens-modeling uncertainties hinder any of our science goals?

Faintest Galaxies yet Known

Detecting the intrinsically faintest high-redshift galaxies is challenging. The intrinsically faintest galaxies are those that are highly magnified, yet still just faintly detected. Highly magnified galaxy images sometimes appear close to brighter galaxies in the lensing cluster, but modeling and subtracting these foreground cluster galaxies have revealed distant galaxies fainter than any previously observed.

Without the aid of gravitational lensing, the blank parallel Frontier Fields reveal galaxies as faint as 29th magnitude AB (5-sigma detections). This is roughly one magnitude shallower than the UDF (AB mag 30). However, the lensing power of the Frontier Fields clusters boosts the depth of those images by at least three magnitudes to AB mag 32 within small, highly magnified regions. This roughly matches the expected depth of ultra-deep Webb images of blank fields, incredibly giving us a sneak preview of Webb’s universe!

The Frontier Fields data show that z ∼ 6–8 luminosity functions remain relatively steep all the way down to AB mag 32 (Figure 2; Atek et al. 2015; Livermore et al. 2016). This confirms there will be plenty of faint galaxies for Webb to observe at these redshifts. The Frontier Fields have also revealed galaxies among the most distant yet known at z ∼ 10, in the first 500 Myr (Figure 3). One of these z ∼ 10 candidates is lensed to form three multiple magnified images (Zitrin et al. 2014). Another is the faintest z ∼ 10 candidate yet known, intrinsically about AB mag 32, magnified by a factor of ∼20 to AB mag 28.5 (Infante et al. 2015). This dwarf galaxy (dubbed “Tayna”) is about the size of the LMC, with a star-formation rate 10 times higher. These galaxies will be prime targets for detailed follow-up study with Webb.

Figure 2: The z ∼ 7 luminosity function (galaxies observed 750 Myr after the Big Bang) constrained ten times (2.5 magnitudes) fainter than the UDF, based on Frontier Fields imaging of the first two clusters (Livermore et al. 2016). The Frontier Fields images are revealing the faintest galaxies yet known.

Figure 3: Candidate z ∼ 10 galaxies revealed by the Frontier Fields. Left: three magnified images of a z ∼ 10 candidate galaxy strongly lensed by Abell 2744 (Zitrin et al. 2014). Right: The faintest z ∼ 10 candidate known, magnified by a factor of ∼20 by MACS J0416.1-2403 (Infante et al. 2015).

One key unresolved question is whether the numbers of z > 9 candidates are consistent with expectations. Previous Hubble programs suggested a deficit of z ∼ 9–11 galaxies compared to extrapolations from lower redshifts z ∼ 4–8 (e.g., Oesch et al. 2013). This suggests that galaxies formed and evolved extremely rapidly in the first 500 Myr. Data from the Frontier Fields should strongly confirm or refute this claim (Coe et al. 2015). Based on the data obtained so far, some authors report confirmation of this rapid evolution (Laporte et al. 2016), while others have reported a less rapid, smoother evolution (McLeod et al. 2016). Further scrutiny of the full Frontier Fields dataset will be required to resolve this issue.

Lens Models Validated by Supernovae

Studying the properties of these lensed galaxies requires accurate magnification estimates. Lensing does not affect galaxy colors or any properties derived from them, such as redshift, age, or metallicity, but lensing uncertainties do propagate directly to estimates of galaxy mass, size, and star-formation rate.

Fortunately, the Frontier Fields clusters have the best lens models ever produced. The Institute initially coordinated the efforts of five teams as they shared the best available observational constraints and then delivered lens models of all six clusters to the community. (See the Newsletter article by Priyamvada Natarajan in Vol. 32, Issue 1.) These “version 1” models were made publicly available before the deep imaging began. Now based on the deeper Frontier Fields imaging and additional spectroscopy, improved models are being produced and delivered to the community. Figure 4 shows the improvement in strong lensing constraints for one cluster, MACS J0416.1-2403.

Figure 4: Lens models of the Frontier Fields clusters are the best ever produced, thanks to the exquisite data and collaborative modeling efforts. Left: Multiple images of galaxies lensed by MACS J0416.1-2403, as identified in CLASH imaging (Zitrin et al. 2013; Johnson et al. 2014). Right: Multiple images identified in deeper Frontier Fields imaging (Jauzac et al. 2014; Diego et al. 2015; Kawamata et al. 2016). Red circles indicate galaxies with spectroscopic redshifts, primarily from VLT X-Shooter (Christensen et al. 2012), CLASH-VLT (Grillo et al. 2015), and HST GLASS (Hoag et al. 2016).

Supernovae are providing direct empirical tests of the lens model accuracies. Lensed by Abell 2744, supernova Tomas is a Type Ia at z = 1.3457 (Rodney et al. 2015). This “standard candle” is observed to be about twice as bright as other Type Ia’s at similar redshifts. Lens model predictions for the magnification ranged from about two to three; on average, they were high by about 25%. This level of accuracy is roughly consistent with results from tests being performed on simulated data to quantify lens-model accuracies in more detail (Meneghetti et al., in prep.).

In November 2014, supernova Refsdal was lensed to form four multiple images by a cluster galaxy in MACS J1149.5+2223 (Figure 5; Kelly et al. 2015). The lens models predicted the supernova would reappear a year later in a fifth multiple image. Based on these predictions, astronomers proposed additional Hubble imaging to monitor Refsdal. The proposal was approved, and in December 2015, Refsdal reappeared right on schedule—as predicted by the lens models (Kelly et al. 2016).

Figure 5: Refsdal is the first-ever-predicted supernova image. It appeared first as four multiple images in the red circles (Kelly et al. 2015) and then one year later in the cyan circle, as predicted by the lensing models (Kelly et al. 2016).

Frontier Fields Finale

The Frontier Fields Hubble and Spitzer imaging is nearly complete. All that remains is the deep Hubble ACS imaging of the final two blank fields, and deep WFC3/IR imaging of the final two clusters: Abell S1063 and Abell 370.

Abell 370 has historical significance. The long, bright, colorful arc (Figure 6) was the first to be confidently identified as a gravitationally lensed galaxy (Soucail et al. 1988). A new field of research was born. Hubble has since imaged Abell 370 multiple times with WFPC2, ACS, and WFC3. The deeper Frontier Fields ACS imaging was completed in February and can be seen in Figure 6 alongside the earlier ground-based imaging. This September, the deep WFC3/IR imaging will be completed, and the cluster that started it all will be the final Frontier Field.

Figure 6: Abell 370, as observed in 1985 by CFHT (left) and today by Hubble (right). Left: The long arc at the bottom of this image was the first to be confidently identified as a gravitationally lensed galaxy (Soucail et al. 1988; image from Lynds & Petrosian 1989). Right: Deep ACS Frontier Fields imaging and shallow archival WFC3/IR imaging. Much deeper WFC3/IR imaging upcoming in July through September will conclude the Frontier Fields program.

We expect Webb to revisit the Frontier Fields; we selected the targets with that in mind. The Hubble images confirm an abundance of ultra-faint distant galaxies for Webb to examine in more detail. How many galaxies will Webb find within the first 400 Myr (z > 11)? Or the first 200 Myr (z > 18)? What will they look like? If the relative abundance of faint galaxies continues to climb with redshift, then the power of gravitational lensing will be further magnified. Lensing may prove to be the key to seeing the first galaxies with Webb.


Atek, H., et al. 2015, ApJ, 814, 69

Coe, D., et al. 2013, ApJ, 762, 32

Coe, D., et al. 2015, ApJ, 800, 84

Diego, J. M., et al. 2015, MNRAS, 447, 3130

Grillo, C., et al. 2015, ApJ, 800, 38

Hoag, A., et al. 2016, arXiv:1603.00505

Infante, L., et al. 2015, ApJ, 815, 18

Jauzac, M., et al. 2014, MNRAS, 443, 1549

Johnson, T. L., et al. 2014, ApJ, 797, 48

Kawamata, R., et al. 2016, ApJ, 819, 114

Kelly, P. L., et al. 2015, Sci, 347, 1123

Kelly, P. L., et al. 2016, ApJ, 819, L8

Laporte, N., et al. 2016, ApJ, 820, 98

Livermore, R. C., et al. 2016, arXiv:1604.06799

Lotz, J., et al. 2013; https://blogs.stsci.edu/newsletter/2013/04/12/hubble-boldly-goes-the-frontier-fields-program/

Lynds, R., & Petrosian, V. 1989, ApJ, 336, 1

McLeod, D. J., et al. 2016, arXiv:1602.05199

Oesch, P. A., et al. 2013, ApJ, 773, 75

Ogaz, S., et al. 2015; https://blogs.stsci.edu/newsletter/files/2015/03/FFCalibration.pdf

Postman, M., et al. 2012, ApJS, 199, 25

Rodney, S. A., et al. 2015, ApJ, 811, 70

Soucail, G., et al. 1988, A&A, 191, L19

Zitrin, A., et al. 2013, ApJ, 762, L30

Zitrin, A., et al. 2014, ApJ, 793, L12

Matt Mountain, Kathy Flanigan, Ken Sembach – Hubble PI
Jennifer Lotz – Program Lead
Anton Koekemoer – Hubble Image Processing Lead
Tom Soifer – Spitzer PI
Peter Capak – Spitzer Program Lead
Dan Coe – Lens Modeling Coordinator