The CANDELS Multi-Cycle Treasury Program

H. Ferguson,, & the CANDELS team

The Cosmic Assembly Near-IR Deep Extragalactic Legacy Survey (CANDELS) is one of three Multi-Cycle Treasury (MCT) programs approved in 2010. The 900-orbits’ worth of observations finished in August 2013. The observations were designed to document the first third of galactic evolution from z = 8 to 1.5 (lookback times of 13 to 9 billion years ago) via deep imaging of more than 250,000 galaxies with the Wide Field Camera 3 infrared channel (WFC3/IR) and the Advanced Camera for Surveys (ACS). Together with the CLASH cluster observations, another aim has been to find and characterize Type Ia supernovae (SNe Ia) beyond z > 1.5, and to establish their accuracy as standard candles for cosmology. The survey targets five premier multi-wavelength sky regions; each has multi-wavelength data from Spitzer and other facilities, and has extensive spectroscopy of the brighter galaxies. The use of five widely separated fields mitigates cosmic variance and yields statistically robust and complete samples of galaxies down to 109 solar masses out to redshift z ~ 8. The data are immediately public, and high-level science products have been appearing regularly in the Mikulski Archive for Space Telescopes (MAST). Further information on the observations and data reduction can be found in the Institute Newsletter (Vol. 28, Issue 1) and on the CANDELS website, as well as in two overview papers (Grogin et al. 2011 and Koekemoer et al. 2011).

Even while the data have been pouring in, both the CANDELS team and the wider community have been busy trying to make progress on some of the science goals with only a portion of the data. By the end of June 2013, the team published 28 papers and submitted 11 others. While harder to identify, there are roughly 20 papers using CANDELS data from outside the team, either published or available as preprints.

We focus here on a few of the science highlights, typically from about a third of the full survey data.

Galaxies and AGN Hosts at Cosmic High Noon

CANDELS-1The CANDELS survey provides high-resolution data in the rest-frame optical for galaxies at z ~ 2, when cosmic star formation and active galactic nuclei were at their peak. A solid—but perhaps not particularly surprising—result comes from studies of the relation between star-formation activity and galaxy morphology: passive galaxies at z ~ 2 already look like spheroids, by and large. This result is seen clearly using different sample selection and different kinds of morphological analysis in CANDELS papers by Wuyts et al. (2011), Wang et al. (2012), Bell et al. (2012), Lee et al. (2013) and others. Perhaps the most visually striking illustration of this is in Figure 1, from Wang et al., which looks specifically at extremely red objects selected on the basis of the ratio of their Spitzer 3.6 micron fluxes relative to their Hubble 0.9 micron fluxes, and then separates these red galaxies into star forming and quiescent based on their 24-to-3.6 micron flux ratio. The inset at the upper left shows that the quiescent galaxies generally lie in the spheroid portion of the Gini-M20 diagram (Lotz et al. 2004). Indeed, the morphology of galaxies appears to be the best predictor of whether or not they are passive. It is a better predictor than stellar mass, for example (Bell et al. 2012).

CANDELS-2In addition to being spheroidal in shape, passive galaxies at z ~ 2 are generally quite small: their half-light radii are much smaller than spheroids of the same mass today, and they are smaller than star-forming galaxies of the same stellar mass at the same epoch.Ongoing work from CANDELS is providing better constraints on the size evolution of such galaxies (e.g., Cassata et al. 2011, 2013) and attempting to identify the compact star-forming progenitors of these “red nuggets” (Barro et al. 2013; Williams et al. submitted). Barro et al. identify blue nuggets that have roughly the required stellar surface densities to become red nuggets if star formation is shut off. Comparing number densities of blue and red nuggets in the redshift range 1 < z < 3, it appears possible to match the evolution in the number density of red nuggets if the transition timescale from blue to red is about 1 Gyr and the population of blue nuggets is replenished.  This suggests a fast track for creation of early, compact spheroids, followed by slower mechanisms to produce the larger spheroids (and grow the compact ones) at later epochs, as illustrated schematically in Figure 2.

Bruce et al. (2012) measured a subset of the most massive galaxies in the ultra-deep survey (UDS) CANDELS field, carrying out bulge-disk decomposition using the WFC3 images in F160W. At z ~ 2, about half of the most massive galaxies are bulge dominated, albeit with bulges that are more compact than their present-day counterparts. As one pushes closer to z ~ 3, the bulge fraction decreases, and there are some disk-like galaxies that have the red colors characteristic of passive galaxies.

A popular paradigm for explaining the transition from star-forming to quiescent galaxies has been to invoke mergers and feedback from AGN. The merging process drives gas to the center of the galaxy, fueling star formation and the AGN. This can help to explain the tight correlation between the stellar mass in galaxy bulges and the mass of central black holes observed locally. Because the merger triggers the AGN activity, we might expect the host galaxies of AGN to look disturbed, or at least not to have re-grown their disks. CANDELS observations of AGN hosts at z ~ 2 do not confirm this expectation. Using visual morphology classifications from the CANDELS team compared the hosts of X-ray-selected AGN at z ~ 2 to a mass-matched sample at the same redshifts (Kocevski et al. 2011). There was no indication that the AGN hosts were more disturbed than non-AGN galaxies of the same stellar mass, and about 50% of the AGN hosts had a disk component. The sample does not include the most luminous AGN, so perhaps the merger scenario works for them, but that remains to be seen.

Galaxies near Cosmic Dawn

Another theme of CANDELS has been to study galaxies close to the epoch of re-ionization, at z > 6. Compared to the Hubble Ultra-Deep Fields (HUDFs), CANDELS offers additional volume to constrain the shape of the bright end of the luminosity function (LF), as well as deep multi-wavelength data to help constrain stellar populations and AGN fractions. Indeed, the CANDELS data have already been used by a variety of groups to explore galaxy evolution at these high redshifts (e.g., Yan et al. 2012; Finkelstein et al. 2012; Oesch et al. 2013; Bouwens et al. 2013; Schenker et al. 2013; McLure et al. 2013). Galaxies are identified via their Lyman break, which for these redshifts occurs at ~1216 Å in the rest frame. Shorter-wavelength photons are scattered out of the line of sight by intervening hydrogen in the intergalactic medium. Different investigators use slightly different criteria for selecting high-z candidates, but generally derive similar number densities per unit redshift when they account for their selection biases. There is some disagreement at the bright end of the luminosity function, at z ~ 7–8, with Yan et al. (2012) finding a few more bright candidates than expected from the extrapolation of the evolution trend seen in the best-fit luminosity functions at lower redshift from the Bouwens group. Such disagreements should be easy to resolve as further data from CANDELS and the WFC3 parallel programs are incorporated.  In spite of such discrepancies, the consensus is that the faint-end slope is very steep, and that it is therefore likely that dwarf galaxies are responsible for re-ionization, provided the ultraviolet (UV) photons can escape.

CANDELS-3There has also been a lot of interest in investigating the evolution of the UV spectral-slopes of high-redshift galaxies, spurred by the tantalizing report of very steep slopes from the early HUDF analysis (Bouwens et al. 2010). More recent analysis suggests that the bluest galaxies are not significantly bluer than local metal-poor galaxies like NGC 1705, and hence do not require metal-free stars (Fig. 3). Nevertheless, there is systematic evolutionary trend at fixed UV luminosity such that higher redshift objects have bluer rest-frame colors. The evidence also tends to favor a mild slope-luminosity relation in the sense that lower-luminosity galaxies are bluer than higher-luminosity ones, but this needs to be confirmed with more data. Both trends are in this sense expected if galaxies are increasing in metallicity and dust content with both mass and redshift. If the red colors are indeed due to dust, then the fact that luminous star-forming galaxies at z ~ 8 have some dust suggests that supernovae and not just AGB stars are responsible, because not enough time would have elapsed for the AGB stars to appear.


Both the CANDELS and the CLASH MCT programs have scheduled their observations to facilitate finding high-redshift supernovae. SNe Ia are the best known standard candles for measuring cosmological distances.  Together, the programs have identified more than 100 high-redshift supernovae of all types, with a handful at z > 1.5. At these high redshifts, dark energy should have a negligible influence on the distance-luminosity relation. On the other hand, the progenitor of a typical SN Ia at z > 1.5 will be considerably more massive and probably less metal-rich than a typical SN Ia progenitor today. Therefore, any departures from the matter-dominated distance-luminosity relation are more likely to be due to supernova evolution than dark energy. Observations of supernovae at these high redshifts will provide a strong lever arm for constraining the evolution of the luminosity and luminosity-light-curve-shape relation for SNe Ia. Followup observations of high-z candidates have been aimed at getting host-galaxy redshifts, supernova light curves, and supernova spectra where feasible.