This document describes the discovery of a massive protocluster of galaxies located approximately 1 billion years after the Big Bang (redshift of z=5.3). The protocluster contains overdense regions of massive galaxies extending over 13 megaparsecs. It contains an extremely luminous starburst galaxy with large molecular gas reserves and a luminous quasar. Together, these objects place a minimum total mass of over 4×1011 solar masses in this early cluster, consistent with cosmological simulations of the earliest galaxy clusters. This discovery provides evidence for the hierarchical formation of massive structures in the early universe.

1. LETTER doi:10.1038/nature09681
A massive protocluster of galaxies at a redshift of
z < 5.3
Peter L. Capak1, Dominik Riechers2, Nick Z. Scoville2, Chris Carilli3, Pierre Cox4, Roberto Neri4, Brant Robertson2, Mara Salvato5,
Eva Schinnerer6, Lin Yan1, Grant W. Wilson7, Min Yun7, Francesca Civano8, Martin Elvis8, Alexander Karim6, Bahram Mobasher9
& Johannes G. Staguhn10
Massive clusters of galaxies have been found that date from as early Multi-Object Spectrograph (W. M. Keck Observatory, Hawaii) to mea-
as 3.9 billion years1 (3.9 Gyr; z 5 1.62) after the Big Bang, contain- sure redshifts.
ing stars that formed at even earlier epochs2,3. Cosmological simu- We found a grouping of four major objects at z 5 5.30 (Fig. 1). The
lations using the current cold dark matter model predict that these most significant overdensity appears near the extreme starburst galaxy
systems should descend from ‘protoclusters’—early overdensities COSMOS AzTEC-3, which contains .5.3 3 1010M[ of molecular gas
of massive galaxies that merge hierarchically to form a cluster4,5. and has a dynamical mass, including dark matter, of .1.4 3 1011M[
These protocluster regions themselves are built up hierarchically (ref. 8). The far-infrared (60–120-mm) luminosity of this system is
and so are expected to contain extremely massive galaxies that can estimated to be (1.7 6 0.8) 3 1013 solar luminosities (L[), corres-
be observed as luminous quasars and starbursts4–6. Observational ponding to a star formation rate of .1,500M[ per year18, which is
evidence for this picture, however, is sparse because high-redshift .100 times the rate of an average galaxy (with luminosity LÃ ) at
protoclusters are rare and difficult to observe6,7. Here we report a z 5 5.3 (ref. 19). The value and error given are the mean estimate
protocluster region that dates from 1 Gyr (z 5 5.3) after the Big and scatter derived from empirical estimates based on the sub-
Bang. This cluster of massive galaxies extends over more than 13 millimetre flux, radio flux limit, and CO luminosity, along with model
megaparsecs and contains a luminous quasar as well as a system fitting. The models predict a much broader range in total infrared
rich in molecular gas8. These massive galaxies place a lower limit of (8–1,000-mm) luminosities, ranging from 2.2 3 1013L[ to 11 3 1013L[.
more than 4 3 1011 solar masses of dark and luminous matter in The large uncertainty results from the many assumptions used in the
this region, consistent with that expected from cosmological simu- models, combined with a lack of data constraining the infrared emis-
lations for the earliest galaxy clusters4,5,7. sion at wavelengths less than rest-frame 140 mm. However, the
Cosmological simulations predict that the progenitors of present-
day galaxy clusters are the largest structures at high redshift4,5,7
350
(Mhalo . 2 3 1011 solar masses (M[) and Mstars . 4 3 109M[ at
z < 6). These protocluster regions should be characterized by local Ly Si II O I/Si II C II Si IV S III C IV
300
overdensities of massive galaxies on co-moving distance scales of
2–8 Mpc that coherently extend over tens of megaparsecs, forming a
250
structure that will eventually coalesce into a cluster4,5,7,9. Furthermore,
owing to the high mass densities and correspondingly high merger
200 Quasar
rates, extreme phenomena such as starbursts and quasars should
Counts
z = 5.305
preferentially exist in these regions4–7,9,10. Although overdensities have
been reported around radio galaxies on ,10–20-Mpc scales6,7 and 150
Cluster LBG
large gas masses around quasars11,12 at redshifts greater than z 5 5, z = 5.300
the available data is not comprehensive enough to constrain the mass 100
of these protoclusters and hence provide robust constraints on cos-
mological models6,7,9. 50 COSMOS AzTEC3
z = 5.298
We used data covering the entire accessible electromagnetic spectrum
in the 2-square-degree Cosmological Evolution Survey (COSMOS) 0
field13 (right ascension, 10 h 00 min 30 s; declination, 2u 309 0099) to 1,200 1,250 1,300 1,350 1,400 1,450 1,500 1,550 1,600
Rest wavelength (Å)
search for starbursts, quasars and massive galaxies as signposts of poten-
tial overdensities at high redshift. This deep, large-area field provides the Figure 1 | Spectra of confirmed cluster members. These spectra were taken
multiwavelength data required to find protoclusters on scales .10 Mpc with the Keck II telescope and correspond to the extreme starburst (COSMOS
(59). Optically bright objects at redshifts greater than z 5 4 were iden- AzTEC3), a combined spectrum of two Lyman-break galaxies at 95 kpc
(Cluster LBG) and the Chandra-detected quasar at 13 Mpc from the extreme
tified through optical and near-infrared colours. Extreme star formation
starburst. The galaxy spectra show absorption features indicative of interstellar
activity was found using millimetre-wave14,15 and radio16 measurements, gas (Si II, O I/Si II and C II) and young massive stars (Si IV and C IV) indicative of
and potential luminous quasars were identified by X-ray measure- a stellar population less than 30 Myr old26. The quasar shows broad Lyman-a
ments17. Finally, extreme objects and their surrounding galaxies were (Lya) emission absorbed by strong winds, with a narrow Lyman-a line seen at
targeted with the Keck II telescope and the Deep Extragalactic Imaging the same systemic velocity as absorption features in the spectra.
1
Spitzer Science Centre, 314-6 California Institute of Technology, 1200 East California Boulevard, Pasadena, California 91125, USA. 2Department of Astronomy, 249-17 California Institute of Technology,
1200 East California Boulevard, Pasadena, California 91125, USA. 3National Radio Astronomy Observatory, PO Box O, Socorro, New Mexico 87801, USA. 4Institut de Radio Astronomie Millimetrique, 300
´
rue de la Piscine, F-38406 St-Martin-d’Heres, France. 5Max-Planck-Institute fur Plasma Physics, Boltzmann Strasse 2, Garching 85748, Germany. 6Max-Planck-Institute fur Astronomie, Konigstuhl 17,
` ¨ ¨ ¨
Heidelberg 69117, Germany. 7Department of Astronomy, University of Massachusetts, Lederle Graduate Research Tower B, 619E, 710 North Pleasant Street, Amherst, Massachusetts 01003-9305, USA.
8
Harvard Smithsonian Center for Astrophysics, 60 Garden Street, MS, 67, Cambridge, Massachusetts 02138, USA. 9Department of Physics and Astronomy, University of California, Riverside, California
92521, USA. 10Johns Hopkins University, Laboratory for Observational Cosmology, Code 665, Building 34, NASA’s Goddard Space Flight Center, Greenbelt, Maryland 20771, USA.
0 0 M O N T H 2 0 1 0 | VO L 0 0 0 | N AT U R E | 1
©2011 Macmillan Publishers Limited. All rights reserved

2. RESEARCH LETTER
Starburst
z = 5.298
z = 5.298
T star
z = 5.300
z = 5.300
M star
z = 5.300
z ≈ 1 galaxies
10 1
Figure 2 | Image of the region around the protocluster core. This area Figure 3 | Detail of the protocluster core. This area corresponds to a
corresponds to a 29 3 29 region around the starburst (COSMOS AzTEC-3). 22.599 3 22.599 area on the sky at a co-moving distance of 0.865 Mpc, or a
The z < 5.3 candidates are marked in white and a 2-Mpc co-moving radius is proper distance of 0.137 Mpc, at z 5 5.298. The six optically bright objects with
marked with a green circle. The boxed area is shown larger in Fig. 3, where the spectral energy distributions consistent with z 5 5.298 are marked and
optical counterpart of the submillimetre source COSMOS AzTEC-3 is labelled spectroscopic redshifts are indicated. The optical counterpart of the
‘Starburst’. Spectroscopic redshifts and other red objects that have been submillimetre source COSMOS AzTEC-3 is labelled ‘Starburst’.
identified as galactic stars or low-redshift galaxies by their spectral energy
distribution are also labelled. Eddington rate5. Assuming the final black hole/stellar mass relation to
be MBH < 0.002Mstars (ref. 5) implies that this object will eventually
observed limit on the submillimetre spectral slope favours models with have a stellar mass of .1010M[–1011M[, placing it among the most
colder dust and, hence, lower luminosities. luminous and massive objects at this redshift19,23.
The significance of the overdensity around the starburst is imme- We estimated the stellar mass of the protocluster system by fitting
diately apparent in Figs 2 and 3. In the 1-square-arcmin area stellar population models to the rest-frame ultraviolet–optical pho-
(2.3 3 2.3 Mpc2 at z 5 5.3) around the starburst, we would expect to tometry of the individual galaxies in the protocluster. The redshift
find 0.75 6 0.04 bright (z850 , 26) galaxies with colours consistent was fixed at z 5 5.298, and models with a single recent burst of star
with a Lyman break in their spectra at z 5 5.3 (ref. 19), but instead formation24 were used, allowing for up to ten visual magnitudes of
we find eight. This is an 11-fold overdensity, assuming the redshift extinction25. [O II] and Ha emission lines were added to the templates
range 4.5 , z , 6.5 probed by typical broadband colour selections19,20. with fluxes proportional to the ultraviolet continuum of the template18.
Within a 2-Mpc radius of the starburst, we find 11 objects brighter The accuracy of the stellar mass estimate is limited by the sensitivity of
than LÃ whose intermediate-band colours21 are consistent with being the 0.9–2.5-mm photometry. The present data are insufficient to fully
at z 5 5.3. This represents a .11-fold overdensity in both the mea- break the degeneracy between stellar age and dust obscuration.
sured and the expected density of luminous galaxies. Estimates of the However, the age of 10 Myr derived from the photometric fitting is
typical variance from clustering and cosmological simulations suggest consistent with the features seen in the Keck spectra26. Given the range
that this is significant at the .9s level even if we only consider the of acceptable fits and the concordance with the Keck spectra, the
spectroscopically confirmed systems. Of these 11 objects, three resulting stellar mass is probably accurate to a factor of ,2 (0.3 dex).
(including the optical counterpart of COSMOS AzTEC-3) are within Using the described procedure, we conservatively estimate that the
proper distance of 12.2 kpc (299) of COSMOS AzTEC-3, and two starburst AzTEC-3 has a stellar mass of (1–2) 3 1010M[, implying that
additional spectroscopically confirmed sources are found 95 kpc the baryonic matter is .70% gas, nearly twice that found in typical star-
(15.599) away. burst systems27 but in agreement with the dynamical estimates8,28. The 11
X-ray-selected (0.5–10-keV band) z . 5 quasars are extremely objects in the protocluster core have a total stellar mass of .2 3 1010M[,
rare22 owing to the high luminosities required for detection, yet one with individual galaxies weighing between 0.06 3 109M[ and
is found17 within 13 Mpc of the starburst at the same spectroscopic 10 3 109M[. With this stellar mass and gas fraction, a lower limit
redshift as COSMOS AzTEC-3. The distance between these objects is can be placed on the total mass of this system, assuming a global dark
comparable to the co-moving distance scale expected for protoclusters matter/baryon ratio of 5.9 (ref. 1). The resulting total halo mass is
at z < 5 (refs 5, 7). The optical spectrum of the quasar has deep, blue- .4 3 1011M[, with the starburst residing in a halo of mass
shifted gas absorption features indicative of strong winds driven by the .1011M[, comparable to the halo masses predicted for galaxies that
energy dissipated from the rapid black-hole growth. The object has an will eventually merge into present-day galaxy clusters7. However, we
X-ray luminosity of 1.9 3 1011 L[ and a bolometric luminosity esti- note that the actual mass is probably much higher because much of the
mated from its spectral energy distribution of $8.3 3 1011L[ (H. Hao baryonic mass is probably in unobserved hydrogen gas, and the star-
et al., manuscript in preparation), implying a black-hole mass of burst object alone accounts for .37% of the total mass. Furthermore,
$3 3 107M[ if it is accreting at the Eddington rate, with a more likely the contribution of significantly more numerous, fainter (luminosity,
mass of ,3 3 108M[ for the typical accretion rate of one-tenth the ,LÃ ) galaxies19 are not counted in this mass estimate.
2 | N AT U R E | VO L 0 0 0 | 0 0 M O N T H 2 0 1 0
©2011 Macmillan Publishers Limited. All rights reserved

3. LETTER RESEARCH
The three objects around COSMOS AzTEC-3 probably represent 14. Scott, K. S. et al. AzTEC millimetre survey of the COSMOS field - I. Data reduction
and source catalogue. Mon. Not. R. Astron. Soc. 385, 2225–2238 (2008).
the progenitor of a massive central cluster galaxy (type cD) at lower 15. Younger, J. D. et al. Evidence for a population of high-redshift submillimeter
redshift. These objects are already within the radius of a typical local galaxies from interferometric imaging. Astrophys. J. 671, 1531–1537 (2007).
cD galaxy and their dynamical timescale is ,60 Myr, assuming a 16. Schinnerer, E. et al. The VLA-COSMOS Survey. II. Source catalog of the large project.
Astrophys. J. Suppl. Ser. 172, 46–69 (2007).
velocity dispersion of 200 km s21. Even considering the objects at 17. Elvis, M. et al. The Chandra COSMOS Survey. I. Overview and point source catalog.
95 kpc, the dynamical timescale is less than 0.5 Gyr, providing several Astrophys. J. 184, 158–171 (2009).
dynamical times for a merger to occur by z < 2 (that is, 2 Gyr later). 18. Kennicutt, J. Star formation in galaxies along the Hubble sequence. Annu. Rev.
Astron. Astrophys. 36, 189–232 (1998).
However, the observed stellar mass in these galaxies is significantly less 19. Bouwens, R. J., Illingworth, G. D., Franx, M. & Ford, H. UV luminosity functions at
than the ,1011M[–1012M[ in a typical local cD galaxy7, indicating z , 4, 5, and 6 from the Hubble Ultra Deep Field and other deep Hubble Space
that the majority of the stars have yet to form. Telescope ACS fields: evolution and star formation history. Astrophys. J. 670,
928–958 (2007).
The properties of this protocluster are in qualitative and quantitat- 20. Hildebrandt, H. et al. CARS: the CFHTLS-Archive-Research Survey. II. Weighing
ive agreement with galaxy formation simulations4,5. The spatial extent, dark matter halos of Lyman-break galaxies at z 5 3–5. Astron. Astrophys. 498,
star formation rate per unit mass and gas properties of the core struc- 725–736 (2009).
ture around the extreme starburst are all similar to the predictions for 21. Capak, P. et al. The first release COSMOS optical and near-IR data and catalog.
Astrophys. J. Suppl. Ser. 172, 99–116 (2007).
massive-galaxy formation in simulations. Furthermore, the properties 22. Brusa, M. et al. High-redshift quasars in the COSMOS survey: the space density of
of the quasar are also in agreement with the models of the later phases z . 3 X-ray selected QSOs. Astrophys. J. 693, 8–22 (2009).
of massive-galaxy formation when the quasar becomes visible. Finally, 23. Stark, D. P. et al. The evolutionary history of Lyman break galaxies between redshift
4 and 6: observing successive generations of massive galaxies in formation.
unlike for previously described overdensities at z . 5 (ref. 6), we have Astrophys. J. 697, 1493–1511 (2009).
strong spectroscopic and photometric evidence for a range of objects 24. Maraston, C. Evolutionary synthesis of stellar populations: a modular tool. Mon.
including massive, heavily star forming and active galaxies. These Not. R. Astron. Soc. 300, 872–892 (1998).
25. Calzetti, D., Kinney, A. L. & Storchi-Bergmann, T. Dust obscuration in starburst
are found both in the core of the structure and over a much larger galaxies from near-infrared spectroscopy. Astrophys. J. 458, 132–135 (1996).
area, indicating that the effects of environment on galaxy formation 26. Pettini, M., Steidel, C. C., Adelberger, K. L., Dickinson, M. & Giavalisco, M. The
as early as z < 5 can be studied. We conclude that this region con- ultraviolet spectrum of MS 1512–CB58: an insight into Lyman-break galaxies.
Astrophys. J. 528, 96–107 (2000).
tains a large-scale baryonic overdensity in the very early Universe 27. Tacconi, L. et al. High molecular gas fractions in normal massive star-forming
that will evolve into a high-mass cluster like those observed at lower galaxies in the young Universe. Nature 463, 781–784 (2010).
redshifts. 28. Tacconi, L. J. et al. Submillimeter galaxies at z , 2: evidence for major mergers and
constraints on lifetimes, IMF, and CO-H2 conversion factor. Astrophys. J. 680,
Received 18 May; accepted 16 November 2010. 246–262 (2008).
Published online 12 January 2011. Supplementary Information is linked to the online version of the paper at
www.nature.com/nature.
1. Larson, D. et al. Seven-year Wilkinson Microwave Anisotropy Probe (WMAP)
observations: power spectra and WMAP-derived parameters. Astrophys. J. Suppl. Acknowledgements These results are based on observations with: the W. M. Keck
Ser. (in the press); preprint at Æhttp://arxiv.org/abs/1001.4635æ (2010). Observatory, the IRAM Plateau de Bure Interferometer, the IRAM 30-m telescope with
2. Papovich, C. et al. A Spitzer-selected galaxy cluster at z 5 1.62. Astrophys. J. 716, the GISMO 2-mm camera, the Chandra X-ray Observatory, the Subaru Telescope, the
1503–1513 (2010). Hubble Space Telescope, the Canada-France-Hawaii Telescope with WIRCam and
3. Mei, S. et al. Evolution of the color-magnitude relation in galaxy clusters at z , 1 from MegaPrime, the United Kingdom Infrared Telescope, the Spitzer Space Telescope, the
the ACS Intermediate Redshift Cluster Survey. Astrophys. J. 690, 42–68 (2009). Smithsonian Submillimeter Array Telescope, the James Clerk Maxwell Telescope with
4. Springel, V. et al. Simulations of the formation, evolution and clustering of galaxies the AzTEC 1.1mm camera, and the National Radio Astronomy Observatory’s Very
and quasars. Nature 435, 629–636 (2005). Large Array. D.R. and B.R. acknowledge support from NASA through Hubble Fellowship
5. Li, Y. et al. Formation of z , 6 quasars from hierarchical galaxy mergers. Astrophys. grants awarded by the Space Telescope Science Institute. P.L.C. and N.Z.S.
J. 665, 187–208 (2007). acknowledge grant support from NASA. G.W.W., M.Y. and J.G.S. acknowledge grant
6. Overzier, R. et al. Stellar masses of Lyman break galaxies, Lya emitters, and radio support from the NSF.
galaxies in overdense regions at z 5 4–6. Astrophys. J. 704, 548–563 (2009). Author Contributions P.L.C. led the spectroscopic effort, reduced the spectroscopic
7. Overzier, R. et al. LCDM predictions for galaxy protoclusters - I. The relation and photometric data, and led the scientific analysis including the optical and radio/
between galaxies, protoclusters and quasars at z , 6. Mon. Not. R. Astron. Soc. 394, millimetre fitting analysis and cluster properties. N.Z.S. led the spectroscopic and
577–594 (2009). photometric follow-up efforts. D.R., C.C., P.C. and R.N. assisted with the physical
8. Riechers, D. et al. A massive molecular gas reservoir in the z 5 5.3 submillimeter interpretation of the radio data. B.R. provided cosmological simulations to check the
galaxy AzTEC-3. Astrophys. J. Lett. 720, 131–136 (2010). significance of the protocluster and the likelihood of finding it. M.S., L.Y., M.E., F.C. and
9. Robertson, B. et al. Photometric properties of the most massive high-redshift B.M. carried out the Keck observations and assisted with the data reduction. E.S.
galaxies. Astrophys. J. 667, 60–78 (2007). reduced and analysed the radio data. G.W.W. and M.Y. assisted with the submillimetre
10. Miley, G. K. et al. A large population of ‘Lyman-break’ galaxies in a protocluster at data analysis. F.C. and M.E. assisted with the X-ray data analysis. A.K. coordinated the
redshift z < 4.1. Nature 427, 47–50 (2004). 2-mm observations. J.G.S. conducted and reduced the 2-mm observations.
11. Walter, F. et al. Molecular gas in the host galaxy of a quasar at redshift z 5 6.42.
Nature 424, 406–408 (2003). Author Information Reprints and permissions information is available at
12. Wang, R. et al. Molecular gas in z < 6 quasar host galaxies. Astrophys. J. 714, www.nature.com/reprints. The authors declare no competing financial interests.
699–712 (2010). Readers are welcome to comment on the online version of this article at
13. Scoville, N. Z. et al. The Cosmic Evolution Survey (COSMOS): overview. Astrophys. J. www.nature.com/nature. Correspondence and requests for materials should be
Suppl. Ser. 172, 1–8 (2007). addressed to P.L.C. (capak@astro.caltech.edu).
0 0 M O N T H 2 0 1 0 | VO L 0 0 0 | N AT U R E | 3
©2011 Macmillan Publishers Limited. All rights reserved