Optically dark massive galaxies

A substantial population of previously unknown massive dusty galaxies during the first two billion years after the Big Bang have been identified with submillimetre observations. They may solve some outstanding puzzles related to the formation and evolution of most massive galaxies in the Universe today.

H ow did massive galaxies present in the Universe today come to be? The current paradigm of galaxy formation and evolution relies on hierarchical growth 1 , with galaxy assembly expected to be a gradual process. Departing from this picture, massive galaxies seen in the earliest times of the Universe are now believed to have formed via gas fuelled through cold streams to dark matter haloes 2 . Such fuelling allows galaxies to assemble mass through new stars formed within, rather than through mergers with other galaxies. However, a proper understanding of the evolutionary pathways of massive galaxies throughout cosmic history is required to determine the relative role of mergers versus cold streams in assembling galaxies and their mass growth during various epochs of the Universe.
There is currently an apparent discrepancy between the number density of massive galaxies at redshift z = 2, considered 'red and dead' galaxies as they are no longer forming stars, and their presumed progenitors at z > 3. While a few of the latter have been observed in wide-area far-infrared and submillimetre surveys, their implied number density is at least an order of magnitude too low to explain the quiescent galaxy population at z = 2. In a recent issue of Nature, Tao Wang and collaborators report 3 the discovery of a population of massive and star-forming galaxies at a redshift of z > 3. These galaxies were first selected based on their optical to near-infrared colour, leading to an estimate of their stellar mass 4 . In this study, the dusty, star-forming nature of the galaxy sample is confirmed at submillimetre wavelengths with the Atacama Large Millimeter/submillimeter Array (ALMA). A combination of their number counts and spatial distribution suggests that these dustrich galaxies are consistent with those that grow to become the massive red galaxies we observe in the dense environments of the present-day Universe. We now have an observational picture connecting presentday massive elliptical galaxies to their progenitors that formed early in the history of the Universe (Fig. 1). Subsequent studies will help answer how and when these galaxies assembled most of their mass.
The identification of distant galaxy samples generally requires multiwavelength observations using some of the best observational facilities available. It is traditional to use rest-frame ultraviolet (UV) and optical light, or visible and nearinfrared light seen today, to locate distant galaxies. These identifications make use of features such as the Lyman break or 4,000 Å break in the spectra of galaxies. Lymanbreak emission is especially useful to locate distant galaxies as energetic UV photons below the Lyman-limit are absorbed by the neutral gas between galaxies that was abundant at very early cosmic times. In the case of dusty, star-forming galaxies, the large amount of dust also absorbs the rest-frame UV light, making them effectively invisible H-band dropout; dusty, star-forming galaxy 2 billion years BzK selection; red and dead galaxy 3 billion years Luminous red galaxy 13.6 billion years  3 show that the number and spatial distribution of dusty, star-forming galaxies appearing as H-band dropouts and confirmed with ALMA observations are compatible with a growth history that includes both quiescent galaxies that appear as 'red and dead' at z = 2 and elliptical galaxies at z = 0. Credit: luminous red galaxy, Australian Astronomical Observatory/David Malin Images; BzK galaxy, reproduced from ref. 10  in the visible light seen today and mimicking the signature of Lyman-break galaxies. Therefore, the optical and near-infrared selection alone is not adequate to distinguish the galaxies that harbour dust from those that are quiescent. The problem arises due to the inherent degeneracies among redshift, age and extinction, leading to uncertainties in the selection. Existing photometric selections (such as BzK and UVJ colours) end up in a mixed sample of both dusty and star-forming (hence red) and quiescent massive galaxies.
In 2016, using data from the Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey (CANDELS) 5 collected with the Hubble and Spitzer space telescopes, Wang et al. 4 identified a sample of massive galaxies at high redshifts. Spitzer infrared imaging is especially useful since those data capture the rest-frame optical and nearinfrared light that is mostly unaffected by the dust absorption. These galaxies would therefore appear as dropouts in the Hubble optical bands but they would emerge in the Spitzer bands. The true nature of these galaxies, however, remained unknown in 2016.
In their current work, Wang et al. 3 break the degeneracy between dusty, star-forming populations and quiescent galaxies thanks to submillimetre-wave observations at 870 μm with ALMA. Dust-rich galaxies light up in the ALMA 870-μm band due to the re-emission of absorbed photons by the dust grains, while quiescent galaxies remain undetected. The addition of ALMA observations was therefore crucial, since without it the authors could not have made the identification of actively forming galaxies when the Universe was 2 billion years old or younger.
This population is of interest since it is a sample that was previously unknown and has properties that are consistent with progenitors of most massive galaxies today, including luminous red elliptical galaxies that are found in dense environments such as galaxy groups.
In total the authors identify 39 new dusty, star-forming galaxies in some of the deepest CANDELS fields. Extrapolating after accounting for their selection function, the authors estimate about 500 such dusty, star-forming galaxies per square degree on the sky. In comparison, the number density of previously known massive starburst galaxies [6][7][8] , selected in wide-area submillimetre continuum surveys, is about a few galaxies per square degree. The discovery and confirmation of this new population of H-band dropout galaxies as massive galaxies at z > 3 alleviates the tension between the small number of massive Lyman-break galaxies at z > 3 and the sample of massive (and quiescent) galaxies at z ≈ 2.
Previously, due to lack of other galaxy samples, a combination of massive Lymanbreak galaxies and dusty, starburst galaxies were considered to be the progenitors of the quiescent population at z = 2. The population identified with Hubble, Spitzer and ALMA provides a solution to the lack of extreme starbursting high-z systems that could serve as progenitors of the most massive systems -this new dusty bright H-dropout sample can fill the void to explain the massive galaxies present today. The argument is made stronger by the clustering measurements: the new galaxy population has a large-scale clustering bias factor that is consistent with the elliptical galaxy bias factor extrapolated to higher redshifts through a simple linear growth ansatz. The connection can be made stronger with future observations through chemical and other signatures that probe the star-formation and mass-assembly history of these galaxies.
This sample of ALMA-detected H-band dropout galaxies have lower flux densities at 870 μm (median S 870 = 1.6 mJy) on average than the rarer submillimetre-bright starburst galaxies, suggesting that while they contain dust and are actively star-forming, their star formation rates are significantly lower than the typical star-formation rates of 1,000 M ⊙ yr -1 or more usually associated with starburst galaxies at these redshifts. Indeed, their star formation is at the level of 100 M ⊙ yr -1 , which, given that their stellar masses are estimated from the spectral energy distribution, puts them right on the 'main sequence' of typical star-forming galaxies at that redshift. In comparison, rare dusty starbursts lie off the main sequence, suggesting that the sample selected by Wang et al. and the previously known dusty, starbursts may have different physical processes at play during their assembly.
While the new sample has, on average, a lower star-formation rate for the rarer starbursting population due to their higher number density, it dominates the cosmic star-formation rate density in the Universe at z > 3. Previously, through surveys with the Herschel Space Observatory, it was identified that dusty, star-forming galaxies at z = 1-2 dominate the cosmic starformation rate density in the Universe at those redshifts. While Herschel lacked the sensitivity to select star-forming galaxies at z > 3 in its blind surveys, these results show that even at higher redshifts cosmic star-formation rate density is dominated by galaxies that are bright in the infrared, especially long infrared wavelengths that are sensitive to the dust emission from these galaxies.
While observationally there is consistency between the number of massive galaxies today and the sample of Wang et al., existing numerical simulations also do not find adequate numbers of massive galaxies at z > 3.
Future studies will be needed to properly understand how these massive galaxies assembled and the role of mergers versus cold streams. To address the processes at play, future observations that can trace the underlying physical properties of these galaxies must be pursued. These could target the assembly via cold inflows (through study of the gas inflowing to these galaxies) or via the merger history (by searching for companions around these galaxies). The required observations could come from additional observations with ALMA and other facilities such as the upcoming James Webb Space Telescope, while in the longer term, sensitive far-infrared spectroscopic measurements with the Origins Space Telescope 9 could be essential to understand the chemical abundances of these massive galaxies, the physical properties of their star-forming gas and the initial mass function of their stars, among other things.