Laboca survey of_extended_chandra_deep_field_south

Mon. Not. R. Astron. Soc. 000, 1–13 (2011) Printed 5 December 2011 (MN L TEX style ﬁle v2.2)
A

The LABOCA Survey of the Extended Chandra Deep Field South:
Clustering of submillimetre galaxies

Ryan C. Hickox1,2,3⋆ , J. L. Wardlow1,4 , Ian Smail5 , A. D. Myers6, D. M. Alexander1 ,
A. M. Swinbank5 , A. L. R. Danielson5 , J. P. Stott1 , S. C. Chapman7 , K. E. K. Coppin8,
J. S. Dunlop9 , E. Gawiser10, D. Lutz11 , P. van der Werf12, A. Weiß13
arXiv:1112.0321v1 [astro-ph.GA] 1 Dec 2011

1 Department of Physics, Durham University, South Road, Durham DH1 3LE
2 STFC Postdoctoral Fellow
3 Department of Physics and Astronomy, Dartmouth College, 6127 Wilder Laboratory, Hanover, NH 03755, USA
4 Department of Physics & Astronomy, University of California, Irvine, CA 92697, USA
5 Institute for Computational Cosmology, Durham University, South Road, Durham DH1 3LE
6 Department of Physics and Astronomy, University of Wyoming, Laramie, WY 82071, USA
7 Institute of Astronomy, Madingley Road, Cambridge CB3 0HA
8 Department of Physics, McGill University, Ernest Rutherford Building, 3600 Rue University, Montreal, Quebec H3A 2T8, Canada
9 Institute for Astronomy, University of Edinburgh, Royal Observatory, Edinburgh EH9 3HJ
10 Department of Physics and Astronomy, Rutgers, The State University of New Jersey, Piscataway, NJ 08854, USA
11 Max-Planck-Institut f¨ r extraterrestrische Physik, Postfach 1312, 85741 Garching, Germany
u
12 Leiden Observatory, Leiden University, NL 2300 RA Leiden, The Netherlands
13 Max-Planck-Institut f¨ r Radioastronomie, Auf dem H¨ gel 69, 53121, Bonn, Germany
u u

5 December 2011

ABSTRACT
We present a measurement of the spatial clustering of submillimetre galaxies (SMGs) at
z = 1–3. Using data from the 870 µm LABOCA submillimetre survey of the Extended Chan-
dra Deep Field South, we employ a novel technique to measure the cross-correlation between
SMGs and galaxies, accounting for the full probability distributions for photometric redshifts
of the galaxies. From the observed projected two-point cross-correlation function we derive
the linear bias and characteristic dark matter halo masses for the SMGs. We detect cluster-
ing in the cross-correlation between SMGs and galaxies at the > 4σ level. Accounting for
the clustering of galaxies from their autocorrelation function, we estimate an autocorrelation
+1.8
length for SMGs of r0 = 7.7−2.3 h−1 Mpc assuming a power-law slope γ = 1.8, and derive
+0.3
a corresponding dark matter halo mass of log(Mhalo [h−1 M⊙ ]) = 12.8−0.5 . Based on the
evolution of dark matter haloes derived from simulations, we show that that the z = 0 descen-
dants of SMGs are typically massive (∼ 2–3 L∗ ) elliptical galaxies residing in moderate- to
+0.3
high-mass groups (log(Mhalo [h−1 M⊙ ]) = 13.3−0.5 ). From the observed clustering we esti-
mate an SMG lifetime of ∼100 Myr, consistent with lifetimes derived from gas consumption
times and star-formation timescales, although with considerable uncertainties. The clustering
of SMGs at z ∼ 2 is consistent with measurements for optically-selected quasi-stellar ob-
jects (QSOs), supporting evolutionary scenarios in which powerful starbursts and QSOs oc-
cur in the same systems. Given that SMGs reside in haloes of characteristic mass ∼ 6 × 1012
h−1 M⊙ , we demonstrate that the redshift distribution of SMGs can be described remarkably
well by the combination of two effects: the cosmological growth of structure and the evolution
of the molecular gas fraction in galaxies. We conclude that the powerful starbursts in SMGs
likely represent a short-lived but universal phase in massive galaxy evolution, associated with
the transition between cold gas-rich, star-forming galaxies and passively evolving systems.
Key words: galaxies: evolution – galaxies: high-redshift – galaxies: starburst – large-scale
structure of the Universe – submillimetre.

⋆ E-mail: ryan.c.hickox@dartmouth.edu

c 2011 RAS

2 Ryan C. Hickox et al.
1 INTRODUCTION
-27.5
Submillimetre galaxies (SMGs) are a population of high-redshift
ultraluminous infrared galaxies (ULIRGs) selected through their -27.6
redshifted far-infrared emission in the submillimetre waveband
(e.g., Smail, Ivison & Blain 1997; Barger et al. 1998; Hughes et al.
1998; Blain et al. 2002). The redshift distribution of this popu- -27.7
lation appears to peak at z ∼ 2.5 (e.g., Chapman et al. 2003,

Dec (deg)
2005; Wardlow et al. 2011), so that SMGs are at their common- -27.8
est around the same epoch as the peak in powerful active galactic
nuclei (AGN) and specifically quasi-stellar objects (QSOs) (e.g.,
Richards et al. 2006; Assef et al. 2011). This correspondence may -27.9
indicate an evolutionary link between SMGs and QSOs, similar
to that suggested at low redshift between ULIRGs and QSOs by
-28.0
Sanders et al. (1988). However there is little direct overlap (∼ a
few percent) between the high-redshift SMG and QSO popula-
tions (e.g., Page et al. 2004; Chapman et al. 2005; Stevens et al. -28.1 SMGs galaxies
2005; Alexander et al. 2008; Wardlow et al. 2011). The immense
far-infrared luminosities of SMGs are widely believed to arise from 53.5 53.4 53.3 53.2 53.1 53.0 52.9 52.8
intense, but highly-obscured, gas-rich starbursts (e.g., Greve et al. RA (deg)
2005; Alexander et al. 2005; Pope et al. 2008; Tacconi et al. 2006,
2008; Ivison et al. 2011), suggesting that they may represent the Figure 1. Two-dimensional distribution of the 50 LESS SMGs and
formation phase of the most massive local galaxies: giant ellipti- ∼ 50,000 IRAC galaxies in the ECDFS that are used in our analysis. The
cals (e.g., Eales et al. 1999; Swinbank et al. 2006). SMGs shown represent the subset of the 126 SMGs in the full LESS sam-
ple (Weiß et al. 2009) that are in the redshift range 1 < z < 3 and are
SMGs and QSOs may thus represent phases in an evolu-
in regions of good photometry, and so are used in this analysis. The IRAC
tionary sequence that eventually results in the population of lo-
galaxies are chosen to reside at 0.5 < z < 3.5. The SMGs are shown here
cal massive elliptical galaxies. This is a compelling picture, but individually, while the density of galaxies is given by the grayscale. The
testing the evolutionary links is challenging due to the lack of an blank areas represent regions which are excluded from the analysis, includ-
easily-measured and conserved observable to tie the various pop- ing areas of poor photometry (for example around bright stars) or additional
ulations together. For example, the stellar masses of both QSOs sources identified by eye in the vicinity of SMG, as discussed in §2. The
and SMGs are difficult to measure reliably due to either the high density of IRAC galaxies in the field enables an accurate measurement
brightness of the nuclear emission in the QSOs (e.g., Croom et al. of the SMG-galaxy cross-correlation function.
2004; Kotilainen et al. 2009) or strong dust obscuration and po-
tentially complex star-formation histories for the SMGs (e.g.,
Hainline et al. 2011; Wardlow et al. 2011; but see also Dunlop
2011; Michałowski et al. 2011), while the details of the high-
14 SMGs (1 < z < 3)
redshift star formation that produced local massive elliptical galax- All IRAC galaxies 4000
ies are likewise poorly constrained (e.g., Allanson et al. 2009). De- 12 Galaxies for angular
autocorrelation
riving dynamical masses for QSO hosts from rest-frame optical

Number of galaxies
Number of SMGs

spectroscopy is difficult due to the very broad emission lines from 10 3000
the AGN, while dynamical mass measurements using CO emis-
8
sion in gas-rich QSOs are also challenging, due to the potential
non-isotropic orientation of the QSO hosts on the sky and the 2000
6
lack of high-resolution velocity fields necessary to solve for this
(Coppin et al. 2008), as well as the general difficulties in model- 4
1000
ing CO kinematics (e.g., Tacconi et al. 2006; Bothwell et al. 2010;
Engel et al. 2010). 2
Another possibility is to compare source populations via 0 0
the masses of their central black holes. For QSOs and the pop- 0.5 1.0 1.5 2.0 2.5 3.0 3.5
ulation of SMGs that contain broad-line AGN, the black hole z
mass can be estimated using virial techniques based on the
broad emission lines (e.g., Vestergaard 2002; Peterson et al. 2004; Figure 2. Redshift distributions for the IRAC galaxy sample in the redshift
Vestergaard & Peterson 2006; Kollmeier et al. 2006; Shen et al. range 0.5 < z < 3.5 (dotted line), and the SMG sample in the range
2008). Such studies generally find that SMGs have small black 1 < z < 3 (solid line). The histogram for galaxies has been scaled so
holes relative to the local black hole-galaxy mass relations (e.g., that the distribution can be directly compared to that of the SMGs. Also
shown is the redshift distribution for 11,241 galaxies (dashed line) selected
Alexander et al. 2008; Carrera et al. 2011), while the black holes in
to match the overlap in the redshift distributions of the SMGs and galaxies,
z ∼ 2 QSOs tend to lie above the local relation, with masses sim- as used in the galaxy autocorrelation measurement (§3.2). For the SMGs,
ilar to those in local massive ellipticals (e.g., Decarli et al. 2010; 44% have spectroscopic redshifts, while the remainder of the SMGs and all
Bennert et al. 2010; Merloni et al. 2010). These results suggest that the IRAC galaxies have redshift estimates from photometric redshift calcu-
SMGs represent an earlier evolutionary stage, prior to the QSO lations (Wardlow et al. 2011).
phase in which the black hole reaches its final mass. However, high-
redshift virial black hole mass estimates are highly uncertain (e.g.,

c 2011 RAS, MNRAS 000, 1–13

Clustering of SMGs 3
Marconi et al. 2008; Fine et al. 2010; Netzer & Marziani 2010) amplitude, along with their relationship to QSOs and ellipticals, re-
and may suffer from significant selection effects (e.g., Lauer et al. mains uncertain.
2007; Shen & Kelly 2010; Kelly et al. 2010), and so conclusions To make improved measurements of the clustering of SMGs,
about connections between populations are necessarily limited. we need either much larger survey areas (see Cooray et al. 2010 for
The difficulties discussed above lead us to take another route a wide-field clustering measurement for far-IR detected sources) or
to compare SMGs to high-redshift QSOs and low-redshift el- the inclusion of redshift information (to allow us to reduce the ef-
lipticals: through their clustering. Spatial correlation measure- fects of projection on our clustering measurements). To this end, we
ments provide information about the characteristic bias and hence have reanalysed the Weiß et al. (2009) survey of ECDFS using new
mass of the haloes in which galaxies reside (e.g., Kaiser 1984; spectroscopic and photometric redshift constraints on the counter-
Bardeen et al. 1986), and so provide a robust mass estimate that is parts to SMGs (Wardlow et al. 2011) as well as a large catalogue
free of many of the systematics in measuring stellar or black hole of “normal” (less-active) galaxies in the same field. We employ a
masses. The observed clustering of SMGs and QSOs can thus allow new clustering analysis methodology (Myers, White & Ball 2009)
us to test whether these populations are found in similar haloes and to calculate the projected spatial cross-correlation between SMGs
so may evolve into each other over short timescales. With knowl- and galaxies, to obtain the tightest constraint to date on the cluster-
edge of how haloes evolve over cosmic time (e.g., Lacey & Cole ing amplitude of SMGs.
1993; Fakhouri, Ma & Boylan-Kolchin 2010), we can also explore This paper is organised as follows. In § 2 we introduce the
the links to modern elliptical galaxies (e.g., Overzier et al. 2003), as SMG and galaxy samples, and in § 3 we give an overview of the
well as the higher-redshift progenitors of SMGs. Clustering mea- methodology used to measure correlation functions and estimate
surements can also provide constraints on theoretical studies that dark matter (DM) halo masses. In § 4 we present the results, explore
explore the nature of SMGs in a cosmological context. Recent mod- the effects of photometric redshift errors, compare with previous
els for SMGs as relatively long-lived (> 0.5 Gyr) star formation measurements, and discuss our results in the context of the physical
episodes in the most massive galaxies, driven by the early collapse drivers, lifetimes, and evolutionary paths of SMGs. In § 5 we sum-
of the dark matter halo (Xia et al. 2011), or powered by steady ac- marise our conclusions. Throughout this paper we assume a cos-
cretion of intergalactic gas (Dav´ et al. 2010), yield strong cluster-
e mology with Ωm = 0.3 and ΩΛ = 0.7. For direct comparison with
ing for bright sources (850 µm fluxes > a few mJy) with correlation other works, we assume H0 = 70 km s−1 Mpc−1 (except for co-
lengths r0 10 h−1 Mpc. In contrast, models in which SMGs are moving distances and DM halo masses, which are explicitly given
short-lived bursts in less massive galaxies, with large luminosities in terms of h = H0 /(100 km s−1 Mpc−1 )). In order to easily com-
produced by a top-heavy initial mass function, predict significantly pare to estimated halo masses in other recent works on QSO clus-
ˆ
tering (e.g., Croom et al. 2005; Myers et al. 2006; da Angela et al.
weaker clustering with r0 ∼ 6 h−1 Mpc (Almeida, Baugh & Lacey
2011). 2008; Ross et al. 2009), we assume a normalisation for the mat-
ter power spectrum of σ8 = 0.84. All quoted uncertainties are 1σ
Attempts to measure the clustering of SMGs from their (68% confidence).
projected two-dimensional distribution on the sky have for
the most part been ambiguous (Scott et al. 2002; Borys et al.
2003; Webb et al. 2003; Weiß et al. 2009; Williams et al. 2011;
Lindner et al. 2011). Weiß et al. (2009) used the largest, contigu-
ous extragalactic 870-µm survey (of the Extended Chandra Deep 2 SMG AND GALAXY SAMPLES
Field South; ECDFS), to derive the clustering of > 5-mJy SMGs
∼ Our SMG sample comes from the survey of the ECDFS using the
from their projected distribution on the sky. They estimated a cor- Large APEX BOlometer CAmera (Siringo et al. 2009, LABOCA)
relation length of 13 ± 6h−1 Mpc. Most recently, Williams et al. on the Atacama Pathfinder EXperiment (G¨ sten et al. 2006, APEX)
u
(2011) analysed a 1100-µm survey of a region of the COSMOS 12-m telescope (the LABOCA ECDFS Submillimetre Survey, or
field and placed 1-σ upper limits on the clustering of bright SMGs LESS; Weiß et al. 2009). LESS mapped the full 0.35 deg2 ECDFS
(with apparent 870-µm fluxes > 8–10 mJy) of > 6–12 h−1 Mpc.
∼ ∼ to a 870-µm noise level of ∼ 1.2 mJy beam−1 and detected 126
Other work has attempted to improve on angular correla- SMGs at > 3.7σ significance (Weiß et al. 2009, equivalent to a
tion measurements by including redshift information. Using the false-detection rate of ∼ 4%). Radio and mid-infrared counter-
spectroscopic redshift survey of 73 SMGs with 870-µm fluxes of parts to LESS SMGs were identified by Biggs et al. (2011) using
> 5 mJy spread across seven fields from Chapman et al. (2005), a maximum-likelihood technique. Spectroscopic and photometric
∼
Blain et al. (2004) estimated a clustering amplitude from the num- redshifts were obtained for a significant fraction of these counter-
bers of pairs of SMGs within a 1000-km s−1 wide velocity win- parts by Wardlow et al. (2011) and we refer the reader to that work
dow. They derived an effective correlation length of 6.9 ± 2.1 for more details. For this study, we restrict our analysis to the 50
h−1 Mpc, suggesting that SMGs are strongly clustered. How- SMGs that have secure counterparts at z = 1–3 and do not lie close
ever their methodology was subsequently criticised by (Adelberger to bright stars (as discussed below). The upper limit of z = 3 on the
2005), who suggested that accounting for angular clustering of sample is included to maximize overlap in redshift space with the
sources and the redshift selection function significantly increases galaxy sample, in order to obtain a significant cross-correlation sig-
the uncertainties. Using data from the Chandra Deep Field-North, nal, while the lower bound of z = 1 is included to prevent the SMG
Blake et al. (2006) computed the angular cross-correlation between sample from being biased toward low redshifts. Of the SMGs in the
SMGs and galaxies in slices of spectroscopic and photometric red- sample, 22 SMGs (44%) have spectroscopic redshifts (Danielson
shift. They obtained a significant SMG-galaxy cross-correlation et al., in preparation) and the remainder have photometric redshifts
signal, with hints that SMGs are more strongly clustered than the with a typical precision of σz /(1+z) ∼ 0.1 (Wardlow et al. 2011).
optically-selected galaxies, although with only marginal (∼ 2σ) The 870-µm flux distribution for the SMGs having secure counter-
significance. Previous work has therefore pointed toward SMGs parts (Biggs et al. 2011) is consistent with that for all LESS SMGs
being a strongly clustered population, but their precise clustering Weiß et al. (2009), indicating that the requirement that SMGs have

c 2011 RAS, MNRAS 000, 1–13

secure counterparts does not strongly bias the fluxes of our SMG
sample. 0.005

galaxy 1

galaxy 2

galaxy 3
For the cross-correlation analysis, we also require a compar-
ison population in the same field. For this we adopt the ∼ 50,000 0.004
galaxies detected in the Spitzer IRAC/MUSYC Public Legacy Sur-
vey in the Extended CDF-South (Damen et al. 2011). We use an

f(χ) (h Mpc-1)
IRAC selected sample to ensure that each galaxy has photom- 0.003

SMG redshift
etry in a sufficient number of bands, and over a wide enough
wavelength range, to allow robust estimates of photometric red-
0.002
shift. Photo-zs are calculated using template fits to the optical and
IRAC photometry in an identical method to that used for the SMGs
(see Wardlow et al. 2011). The fits are performed with HYPER - Z 0.001
(Bolzonella, Miralles & Pell´ 2000) and the resulting redshift dis-
o
tribution, compared to that for the SMGs, is shown in Figure 2.
0.000
The photometric analysis uses chi-squared minimisation, which al-
3000 3500 4000 4500 5000
lows the calculation of confidence intervals for the best-fit redshift.
Comoving distance χ (h-1 Mpc)
These can be presented as a probability distribution function (PDF)
for the redshift, or equivalently, the comoving line-of-sight distance
χ (calculated for our assumed cosmology). We define the PDF for Figure 3. Example probability distribution functions for three IRAC galax-
ies and an SMG. We mark the “best” (peak) comoving distance for each
each galaxy as f (χ), where f (χ)dχ = 1. Examples of the PDFs
galaxy. Note that for each galaxy in this example, the line-of-sight distance
for the galaxies are shown in Figure 3. between the “peak” redshift of the galaxy and the SMG redshift is far too
Finally, in order to calculate the correlation functions, we large for them to be physically associated. However, because of the uncer-
first create random catalogues of “galaxies” at random positions tainty in the galaxy redshifts (shown by the PDFs), there is a non-negligible
within the actual spatial coverage of our survey. Like many fields, probability that the galaxies lie close to the line-of-sight distance of the
the ECDFS contains several bright stars with large haloes, around SMG.
which few galaxies are detected. Therefore, we use the background
map produced by SE XTRACTOR (Bertin & Arnouts 1996) from
in the redshift distributions of the galaxies and SMGs. Our cluster-
the combined IRAC image during the source extraction proce-
ing analysis is identical in most respects to the QSO-galaxy cross-
dure to create a mask. This mask is applied to the random cat-
correlation study presented in Hickox et al. (2011, hereafter H11).
alogues, the SMGs and the IRAC galaxies, so that the positions
Because the method is somewhat involved, we present only the key
of the random galaxies are unbiased with respect to the SMG
details here and refer the reader to H11 for a full discussion.
and IRAC galaxy samples, and thus the mask does not affect the
cross-correlation measurement. As discussed in Biggs et al. (2011)
and Wardlow et al. (2011), some of the SMG identifications were 3.1 Cross-correlation method
performed manually by examining the regions around the SMGs.
These additional sources are excluded from the clustering analysis The two-point correlation function ξ(r) is defined as the probability
so as not to bias the results. The sky positions of the SMGs and above Poisson of finding a galaxy in a volume element dV at a
galaxies that are outside the masked regions are shown in Figure 1. physical separation r from another randomly chosen galaxy, such
that
dP = n[1 + ξ(r)]dV, (1)
3 CORRELATION ANALYSIS where n is the mean space density of the galaxies in the sample.
The projected correlation function wp (R) is defined as the integral
To measure the spatial clustering of SMGs, we can in principle
of ξ(r) along the line of sight,
derive the autocorrelation of the SMGs themselves. However, as
πmax
we have discussed, current SMG samples are too limited in size and
wp (R) = 2 ξ(R, π)dπ, (2)
available redshift information to make this feasible. Alternatively, 0
we can measure the cross-correlation of a population with a sample
where R and π are the projected comoving separations between
of other sources (for example, less-active galaxies) which populate
galaxies in the directions perpendicular and parallel, respectively,
the same volume (e.g., Gawiser et al. 2001; Adelberger & Steidel
to the mean line of sight from the observer to the two galaxies.
2005; Blake et al. 2006; Coil et al. 2007; Hickox et al. 2009). The
By integrating along the line of sight, we eliminate redshift-space
much larger number of galaxies in the ECDFS (∼ 1000 × more
distortions owing to the peculiar motions of galaxies, which dis-
than the SMGs in a comparable redshift range) allows far greater
tort the line-of-sight distances measured from redshifts. wp (R) has
statistical accuracy in the measurement of clustering.
been used to measure correlations in a number of surveys (e.g.,
To calculate the real-space projected cross-correlation func-
Zehavi et al. 2005; Li et al. 2006; Gilli et al. 2007; Coil et al. 2007,
tion wp (R) between SMGs and galaxies we employ a method de-
2008; Wake et al. 2008a; Myers, White & Ball 2009; Hickox et al.
rived by Myers, White & Ball (2009). This method enables us to
2009; Coil et al. 2009; Gilli et al. 2009; Krumpe, Miyaji & Coil
take advantage of the full photo-z PDF for each galaxy, by weight-
2010; Donoso et al. 2010; Hickox et al. 2011; Starikova et al. 2011;
ing pairs of SMGs and galaxies based on the probability of their
Allevato et al. 2011).
overlap in redshift space. This method allows us to calculate the
In the range of separations 0.3 r 50 h−1 Mpc, ξ(r) for
SMG-galaxy cross-correlation using the full sample of z ≈ 50, 000
galaxies and QSOs is roughly observed to be a power-law,
IRAC galaxies, while the derive the clustering of the galaxies them-
selves using a smaller sample that is selected to match the overlap ξ(r) = (r/r0 )−γ , (3)

c 2011 RAS, MNRAS 000, 1–13

with γ typically ≈1.8 (e.g., Zehavi et al. 2005; Coil et al. 2008, evolution of large scale structure, and because the use of a flux-
2007; Ross et al. 2009). For sufficiently large πmax such that we limited sample means we select more luminous galaxies at higher
average over all line-of-sight peculiar velocities, wp (R) can be di- z. This will affect the measurements of relative bias between SMGs
rectly related to ξ(r) (for a power law parameterisation) by and galaxies, since the redshift distribution of the SMGs peaks at
γ higher z than that for the galaxies and so relatively higher-z galax-
r0 Γ(1/2)Γ[(γ − 1)/2]
wp (R) = R . (4) ies dominate the cross-correlation signal. To account for this in our
R Γ(γ/2) measurement of galaxy autocorrelation, we randomly select galax-
To calculate wp (R) for the cross-correlation between SMGs ies based on the overlap of the PDFs with the SMGs in comoving
and galaxies, we use the method of M09, which accounts for the distance (in the formalism of § 3.1 this is fi,j for each galaxy, av-
photometric redshift probability distribution for each galaxy indi- eraged all SMGs). We select the galaxies so their distribution in
vidually. Following M09, the projected cross-correlation function redshift is equivalent to the weighted distribution for all galaxies
can be calculated using: (weighted by fi,j ). The redshift distribution of this galaxy sam-
ple is shown in Figure 2. We use this smaller galaxy sample to
DS DG (R) calculate the angular autocorrelation of IRAC galaxies.
wp (R) = NR NS ci,j − ci,j (5)
DS RG (R)
i,j i,j

where 3.3 Uncertainties and model fits

ci,j = fi,j / 2
fi,j . (6) We estimate uncertainties on the clustering directly from the data
i,j
using bootstrap resampling. Following H11, we divide the field into
a small number of sub-areas (we choose Nsub = 8), and for each
Here R is the projected comoving distance from each SMG, for bootstrap sample we randomly draw a total of 3Nsub sub-areas
a given angular separation θ and radial comoving distance to the (with replacement), which has been shown to best approximate the
SMG of χ∗ , such that R = χ∗ θ. DS DG and DS RG are the intrinsic uncertainties in the clustering amplitude (Norberg et al.
number of SMG–galaxy and SMG–random pairs in each bin of 2009). To account for shot noise owing to the relatively small size
R, and NS and NR are the total numbers of SMGs and random of the SMG sample, we take the sets of 3Nsub bootstrap sub-
galaxies, respectively. fi,j is defined as the average value of the ra- areas and randomly draw from them (with replacement) a sample
dial PDF f (χ) for each galaxy i, in a window of size ∆χ around of sources (SMGs or galaxies) equal in size to the parent sample;
the comoving distance to each spectroscopic source j. We use only pairs including these sources are used in the resulting cross-
∆χ = 100 h−1 Mpc to effectively eliminate redshift space dis- correlation calculation. We use the bootstrap results to derive the
tortions, although the results are insensitive to the details of this covariance between different bins of R, calculating the covariance
choice. We refer the reader to M09 and H11 for a detailed deriva- matrix using Equation 12 of H11.
tion and discussion of these equations. In this calculation as well We fit the observed wp (R) with two models: a power law and
as in the galaxy autocorrelation, we account for the integral con- a simple bias model (described in § 3.4). We compute model param-
straint as described in H11. This correction increases the observed eters by minimising χ2 (taking into account the covariance matrix
clustering amplitude by ≈15%. as in Equation 13 of H11) and derive 1σ errors in each parame-
ter by the range for which ∆χ2 = 1. We use the same formal-
ism for computing fits to the angular correlation functions, where
3.2 Galaxy autocorrelation ω(θ) = Aθ−δ . We convert A and δ to real-space clustering param-
To estimate DM halo masses for the SMGs, we calculate the rel- eters r0 and γ following the procedure described in § 4.6 of H11.
ative bias between SMGs and galaxies, from which we derive the
absolute bias of the SMGs relative to DM. As discussed below, 3.4 Absolute bias and dark matter halo mass
calculation of absolute bias (and thus halo mass) requires a mea-
surement of the autocorrelation function of the IRAC galaxies. The The masses of the DM haloes in which galaxies and SMGs reside
large size of the galaxy sample enables us to derive the clustering are reflected in their absolute clustering bias babs relative to the DM
2
of the galaxies accurately from the angular autocorrelation function distribution. The linear bias babs is given by the ratio of the autocor-
ω(θ) alone. Although we expect the photometric redshifts for the relation function of the galaxies (or SMGs) to that of the DM. We
IRAC galaxies to be reasonably well-constrained (as discussed in determine babs following the method outlined in § 4.7 of H11, sim-
§ 2), by using the angular correlation function we minimize any un- ilar to the approach used previously by a number of studies (e.g.,
certainties relating to individual galaxy photo-zs for this part of the Myers et al. 2006, 2007; Coil et al. 2007, 2008, 2009; Hickox et al.
analysis. The resulting clustering measured for the galaxies has sig- 2009); in what follows we briefly describe this procedure.
nificantly smaller uncertainties than that for the SMG-galaxy cross- We first calculate the two-point autocorrelation of DM as
correlation. a function of redshift. We use the HALOFIT code of Smith et al.
We calculate the angular autocorrelation function ω(θ) using (2003) assuming our standard cosmology, and the slope of the ini-
the Landy & Szalay (1993) estimator: tial fluctuation power spectrum, Γ = Ωm h = 0.21, to derive
the DM power spectrum, and thus its projected correlation func-
1 DM
tion wp (R), averaged over the redshift distribution for which the
ω(θ) = (DD − 2DR + RR), (7)
RR SMGs and galaxies overlap. We then fit the observed wp (R) of the
where DD, DR, and RR are the number of data-data, data- SMG-galaxy cross-correlation, on scales 0.3–15 h−1 Mpc, with a
DM
random, and random-random galaxy pairs, respectively, at a sep- model comprising a simple linear scaling of wp (R). The best-fit
aration θ, where each term is scaled according to the total numbers linear scaling of the DM correlation function corresponds to bS bG ,
of SMGs, galaxies, and randoms. the product of the linear biases for the SMGs and galaxies, respec-
The galaxy autocorrelation varies with redshift, owing to the tively. This simple model produces a goodness-of-fit comparable

c 2011 RAS, MNRAS 000, 1–13


105
SMG-galaxy cross-correlation (1<z<3) galaxy autocorrelation
104 10 -1

103
wp(R)/R

10-2

ω (θ)
102
dark
mat
101 -3
ter
dar
km 10
atte
r
100

10-1 10-4
0.1 1.0 10.0 0.1 1.0 10.0
R (h-1 Mpc) θ (arcmin)

Figure 4. The projected SMG-galaxy cross-correlation function (derived Figure 5. The angular autocorrelation function of IRAC galaxies, selected
using Equation 5). Uncertainties are estimated from bootstrap resampling. to match the overlap of the SMGs and galaxies in redshift space. Uncer-
A power-law fit to wp (R) is shown by the solid line, and the projected tainties are estimated from bootstrap resampling. The angular correlation
correlation function for DM is shown by the dotted line. Fits are performed function for DM, evaluated for the redshift distributions of the galaxies, is
over the range in separation of R = 0.3–15 h−1 Mpc. Both the power law shown by the dotted gray line. The power law fit was performed on scales
model with γ = 1.8 and a linear scaling of the DM correlation function 0.3′ –10′ and is shown as the solid line. Both the power law model with
provide satisfactory fits to the observed wp (R). Together with the observed δ = 0.8 and a linear scaling of the DM correlation function provide sat-
galaxy autocorrelation, this measurement yields the clustering amplitude isfactory fits to the observed ω(θ). The observed amplitude of the galaxy
and DM halo mass for the SMGs, as described in § 4. autocorrelation yields the absolute bias of the galaxies, which we use to
obtain the absolute bias and DM halo mass of the SMGs.

to that of the power-law model in which the slope γ is allowed to
float. law model, and show the correlation function of the DM calcu-
To determine bS we therefore need to estimate bG . We obtain lated as in § 3.4, which we fit to the data through a linear scal-
bG for the galaxies from their angular autocorrelation in a similar ing. The power-law and linear bias fit parameters are presented
manner to that applied to the SMG–galaxy cross-correlation. Again in in Table 1. For SMGs the observed real-space projected cross-
we calculate the autocorrelation for the DM ωDM (θ), by integrat- correlation is well-detected on all scales from 0.1–15 h−1 Mpc, and
ing the power spectrum from HALOFIT using Equation (A6) of the power-law fits return γ ∼ 1.8, similar to many previous corre-
Myers et al. (2007). We fit the observed ω(θ) with a linear scaling lation function measurements for galaxies (e.g., Zehavi et al. 2005;
of ωDM (θ) on scales 0.3′ –10′ (corresponding to 0.3–10 h−1 Mpc Coil et al. 2008) and QSOs (e.g., Coil et al. 2007; Ross et al. 2009).
at z = 2). This linear scaling corresponds to b2 and thus (combined
G The best-fit parameters for the SMG-galaxy cross-correlation are
with the cross-correlation measurement) yields the SMG bias bS . r0,SG = 5.3 ± 0.8 h−1 Mpc, γ = 1.7 ± 0.2. If we fix the value
Finally, we convert bG and bS to Mhalo using the prescription of of γ to 1.8, we obtain r0,SG = 5.1 ± 0.6 h−1 Mpc, corresponding
Sheth, Mo & Tormen (2001), as described in H11. This character- to a clustering signal that is significant at the > 4σ level, the most
istic Mhalo corresponds to the top-hat virial mass (see e.g., Peebles significant measurement of SMG clustering to date. From the fit of
1993, and references therein), in the simplified case in which all the DM model, we obtain bS bG = 5.83 ± 1.36.
objects in a given sample reside in haloes of the same mass. This We next compute the autocorrelation of IRAC galaxies for the
assumption is justified by the fact (as discussed below in § 4.4) that sample described in § 3.2. The observed ω(θ) is shown in Fig. 5,
SMGs have a very small number density compared to the popu- along with the corresponding power-law fit and scaled correlation
lation of similarly-clustered DM haloes, such that it is reasonable function for DM, calculated as discussed in §3.4. Fit parameters
that SMGs may occupy haloes in a relatively narrow range in mass. are given in Table 1. The power-law model fits well on the chosen
We note that this method differs from some prescriptions in the lit- scales of 0.3′ –10′ . The best-fit power law parameters are r0,GG =
erature which assume that sources occupy all haloes above some 3.3 ± 0.3 and γ = 1.8 ± 0.2, and the best-fit scaled DM model
minimum mass; this is particularly relevant for populations with yields b2 = 2.99 ± 0.40 or bG = 1.73 ± 0.12.
G
high number densities that could exceed the numbers of available This accurate value for bG yields bS = 3.37 ± 0.82 for the
DM haloes over a limited mass range. Given the halo mass func- SMGs. Converting this to DM halo mass using the prescription
tion at z ∼ 2 (e.g., Tinker et al. 2008) the derived minimum mass of Sheth, Mo & Tormen (2001) as described in §3.4, we arrive at
is typically a factor of ∼2 lower, for the same clustering amplitude, log (Mhalo [h−1 M⊙ ]) = 12.8+0.3 . The corresponding halo mass
−0.5
than the “average” mass quoted here. for the galaxies is log (Mhalo [h−1 M⊙ ]) = 11.5 ± 0.2.
For comparison with other studies that attempted to directly
measure the autocorrelation function of SMG, it is useful to present
the SMG clustering in terms of effective power-law parameters for
4 RESULTS AND DISCUSSION
their autocorrelation. Assuming linear bias, the SMG autocorrela-
2
The projected cross-correlation function of the SMG sample with tion can be inferred from the cross-correlation by ξSS = ξSG /ξGG
the IRAC galaxies is shown in Figure 4. We plot the best-fit power- (e.g., Coil et al. 2009). Adopting a fixed γ = 1.8 for the SMG-

c 2011 RAS, MNRAS 000, 1–13


Table 1. Correlation results

Power law fitc Bias model fitd Halo masse
Subset Nsrc a z b r0 (h−1 Mpc) γ χ2
ν bS bG (b2 )
G bS (bG ) χ2
ν (log h−1 M⊙ )

SMGs 50 2.02 7.7+1.8
−2.3 1.8 ± 0.2 0.8 5.83 ± 1.36 3.37 ± 0.82 0.7 12.8+0.3
−0.5
galaxies 11,241 2.13 3.3 ± 0.3 1.8 ± 0.2 1.8 2.99 ± 0.40 1.73 ± 0.12 1.8 11.5 ± 0.2
a Number of objects in the SMG sample and in the galaxy sample used for the galaxy autocorrelation.
b Median redshift for the SMG sample and for the galaxy sample used for the galaxy autocorrelation.
c Power law model parameters are for the autocorrelation of SMGs (derived from SMG-galaxy projected spatial cross-

correlation, along with the galaxy angular autocorrelation) and galaxies (derived from their angular autocorrelation).
d Parameters derived from the observed linear fit of the DM model to the observed correlation function, in order to obtain the

the absolute bias for the SMGs and galaxies (denoted bS and bG , respectively). The linear scaling from the fit corresponds to
bS bG for the SMG-galaxy cross-correlation, and b2 for the galaxy autocorrelation, which in turn yield bG and bS .
G
e DM halo mass derived from the absolute bias, using the method described in § 3.4.

galaxy cross-correlation, we thus obtain r0,SS = 7.7+1.8 h−1 Mpc
−2.3 submm flux limits are shown in Figure 6a. Our measurement is sig-
for the autocorrelation of the SMGs. nificantly more accurate than previous measurements, owing to the
inclusion of redshift information and the improved statistics in the
cross-correlation. The uncertainties are comparable to those quoted
4.1 Effects of SMG photo-z errors by Blain et al. (2004) who estimated r0 using counts of close pairs
in redshift space from spectroscopic surveys. However, these au-
One uncertainty in our estimate of wp (R) for the SMG-galaxy thors did not account for significant additional sources of error, as
cross-correlation is due to the lack of accurate (that is, spectro- discussed by Adelberger (2005). Uncertainties in the redshift se-
scopic) redshifts for roughly half of the SMG population. As de- lection function for spectroscopic objects, along with the presence
scribed in § 3, in calculating wp (R) for the cross-correlation, we of redshift spikes and angular clustering of sources, can strongly
simply assume that the SMGs lie exactly at the best redshifts from impact the number of expected pair counts for an unclustered dis-
the photo-z analysis of Wardlow et al. (2011). Any uncertainties tribution, and therefore significantly affect the results for the clus-
in the SMGs photo-zs could therefore affect the resulting cluster- tering amplitude (Adelberger 2005). In Figure 6a the large error
ing measurement. (Note that photo-z uncertainties in the galaxies bars for the Blain et al. (2004) point represent the increase in the
are accounted for implicitly in the correlation analysis, as we uti- uncertainty by 60% due to angular clustering of sources and red-
lize the full galaxy photo-z PDFs.) To examine the effects of SMG shift spikes (as estimated by Adelberger 2005), but does not in-
photo-z errors, we follow the procedure outlined in § 6.3 of H11. clude the additional uncertainty on the redshift selection function.
We take advantage of the 44% of SMGs that do have spectroscopic Nonetheless, our measurement of r0 is consistent with most previ-
redshifts, and determine how errors in those redshifts affect the ob- ous angular clustering estimates as well as the Blain et al. (2004)
served correlation amplitude. result, and represents a significant improvement in precision.
Specifically, we shift the redshifts of the spectroscopic SMGs
As discussed in § 3.4, we convert the observed clustering
by offsets ∆z/(1+z) selected from a Gaussian random distribution
amplitude to Mhalo by assuming that SMGs obey simple linear
with dispersion σz /(1 + z). To ensure that this step does not arti-
bias relative to the dark matter and reside in haloes of similar
ficially smear out the redshift distribution beyond the range probed
mass. Motivated by the presence of a large overdensity of SMGs
by the galaxies, we require that the random redshifts lie between
and powerful star-forming galaxies in one redshift survey field,
1 < z < 3; any random redshift that lies outside this range is dis-
Chapman et al. (2009) proposed that SMGs obey “complex bias”
carded and a new redshift is selected from the random distribution.
that depends on large-scale environment and merger history, and
Using these new redshifts we recalculate wp (R), using the full for-
that they may reside in somewhat smaller haloes than would be in-
malism described in § 3. We perform the calculation 10 times for
ferred from a linear bias model. Future studies using significantly
each of several values of σz /(1 + z) from 0.05 up to 0.3 (corre-
larger SMG samples may be able to confirm the existence of more
sponding to the range of photo-z uncertainties). For each trial we
complex clustering, but for the present analysis we adopt the sim-
obtain the relative bias by calculating the mean ratio of wp (R), on
plest scenario and derive Mhalo assuming linear bias.
scales 1–10 h−1 Mpc, relative to the wp (R) for the best estimates
of redshift. We then average the ten trials at each σz , and find that at The characteristic halo mass we measure for SMGs is simi-
most the photo-z errors cause the clustering amplitude to decrease lar to that measured for bright far-IR sources (with fluxes > 30
by ∼ 10%. The precise magnitude of this effect is unclear given mJy at 250 µm) detected by the Herschel Space Observatory us-
the range of uncertainties in the SMG photo-z estimates, but it is is ing an angular clustering analysis (Cooray et al. 2010). While it re-
significantly smaller than the statistical uncertainties. We therefore mains uncertain to what extent bright 250 µm sources and 850 µm-
neglect this effect in our final error estimates. selected SMGs represent a common population, both samples com-
prise the luminous end of the star-forming galaxy population de-
tected at those wavelengths and so may represent physically similar
systems. In contrast, our observed SMG clustering is significantly
4.2 Comparison with previous results
stronger than that reported by Amblard et al. (2011) for “submil-
Here we compare our results to other measurements of SMG clus- limetre galaxies” based on a power-spectrum analysis of Herschel
tering in the literature. The observed clustering may depend on the 350 µm maps, which yields a minimum Mhalo of ∼ 3 × 1011 M⊙ .
flux limit of the submm sample, as discussed by Williams et al. The differences in clustering amplitude compared to SMGs result
(2011); measurements of r0 that use SMG samples with similar from the fact that the power spectrum analysis includes unresolved

c 2011 RAS, MNRAS 000, 1–13

faint sources corresponding to far fainter far-IR luminosities, char-
acteristic of typical z ∼ 2 star-forming galaxies rather than the 20
powerful, luminous starbursts that are conventionally referred to as (a)
SMGs in the literature.
15
Weiss et al. (2009)

r0 (h-1 Mpc)
4.3 Progenitors and descendants of SMGs Webb et al. (2003)

Our improved clustering measurement allows us to place SMGs 10
Williams et al. (2011)
in the context of the cosmological history of star formation and
This work
growth of DM structures. Because the clustering amplitude of dark
matter haloes and their evolution with redshift are directly predicted 5
by simulations and analytic theory, we can use the observed clus-
tering to connect the SMG populations to their descendants and Blain et al. (2004)

progenitors, estimate lifetimes, and constrain starburst triggering 0
mechanisms. 3 4 5 6
We first compare the clustering amplitude of SMGs with other 850 µm flux limit (mJy)
galaxy populations over a range of redshifts1 . Figure 6b shows the
approximate ranges of measurements of r0 for a variety of galaxy
and AGN populations. We also show the evolution of r0 with red- 15 clusters

shift for DM haloes of different masses, determined by fitting a (b)
power law with γ = 1.8 to the DM correlation function output by 1014 h-1 MO
•

HALOFIT . Finally, we show the observed r0 for the current SMG 1013 h-1 MO
•

sample, along with the expected evolution in r0 for haloes that 10
r0 (h-1 Mpc)
~2-3 L*
have the observed Mhalo for SMGs at z = 2, calculated using ellipticals
LRGs
SMGs QSOs
the median growth rate of haloes as a function of Mhalo and z
(Fakhouri, Ma & Boylan-Kolchin 2010)2 . red galaxies 1012 h-1 MO
•
Figure 6b shows that while the DM halo mass for the SMGs 5 QSOs
will increase with time from z ∼ 2 to z = 0, the observed r0
stays essentially constant, meaning that the progenitors and descen- blue galaxies MIPS SFGs LBGs
1011 h-1 MO
•

dants of SMGs will be populations with similar clustering ampli-
tudes. Our measurement of r0 shows that the clustering of SMGs 0
is consistent with optically-selected QSOs (e.g., Croom et al. 0.0 0.5 1.0 1.5 2.0 2.5 3.0
ˆ
2005; Myers et al. 2006; da Angela et al. 2008; Ross et al. 2009). z
SMGs are more strongly clustered than the typical star-forming
galaxy populations at all redshifts (e.g. Adelberger et al. 2005; Figure 6. (a) Our new measurement of the autocorrelation length r0 for
Gilli et al. 2007; Hickox et al. 2009; Zehavi et al. 2011), and are SMGs, compared to previous results using samples with similar ∼850 µm
clustered similarly or weaker than massive, passive systems (e.g., flux limits. The two sets of error bars on the Webb et al. (2003) measure-
Quadri et al. 2007, 2008; Wake et al. 2008b; Blanc et al. 2008; ment indicate statistical (±3 h−1 Mpc) and systematic (±3 h−1 Mpc) un-
Kim et al. 2011; Zehavi et al. 2011). The clustering results indicate certainties separately. On the Blain et al. (2004) measurement, the smaller
that SMGs will likely evolve into the most massive, luminous early errors represent the uncertainties quoted by the authors, while the larger
type galaxies at low redshift. We note that the descendants of typi- errors account for angular clustering and redshift spikes as estimated by
Adelberger (2005). Our results are consistent with previous measurements
cal SMGs are not likely to reside in massive clusters at z = 0, but
and represent a significant improvement in precision. (b) Our measure-
into moderate- to high-mass groups of ∼ a few ×1013 h−1 M⊙ .
ment of the autocorrelation length r0 of SMGs, compared to the approxi-
Although some SMGs could evolve into massive cluster galaxies, mate r0 (with associated measurement uncertainties) for a variety of galaxy
the observed clustering suggests that most will end up in less mas- and AGN populations: optically-selected SDSS QSOs at 0 < z < 3
sive systems. (Myers et al. 2006; Ross et al. 2009), Lyman-break galaxies (LBGs) at
A schematic picture of the evolution of SMGs is 1.5 z 3.5 (Adelberger et al. 2005), MIPS 24 µm-selected star-forming
shown in Figure 7, which shows evolution in the mass of galaxies at 0 < z < 1.4 (Gilli et al. 2007), typical red and blue galax-
haloes with redshift as traced by their median growth rate ies at 0.25 z 1 from the AGES (Hickox et al. 2009) and DEEP2
(Fakhouri, Ma & Boylan-Kolchin 2010). The typical progenitors (Coil et al. 2008) spectroscopic surveys, luminous red galaxies (LRGs) at
of SMGs would have Mhalo ∼ 1012 h−1 M⊙ at z ∼ 5, which cor- 0 < z < 0.7 (Wake et al. 2008b), and optically-selected galaxy clusters at
responds to the host haloes of bright LBGs at those redshifts (e.g., 0.1 < z < 0.3 (Estrada, Sefusatti & Frieman 2009). In addition, we show
the full range of r0 for low-redshift galaxies with r-band luminosities in the
Hamana et al. 2004; Lee et al. 2006). At low redshift, the SMG
range 1.5 to 3.5 L∗ , derived from the luminosity dependence of clustering
descendants will have Mhalo = (0.6–5) × 1013 h−1 M⊙ . Halo presented by Zehavi et al. (2011); these luminous galaxies are primarily el-
occupation distribution fits to galaxy clustering suggest that these lipticals, as discussed in § 4.3. Dotted lines show r0 versus redshift for DM
haloes host galaxies with luminosities L ∼ 2–3L∗ (Zehavi et al. haloes of different masses. The thick solid line shows the expected evolu-
tion in r0 , accounting for the increase in mass of the halo, for a halo with
mass corresponding to the best-fit estimate for SMGs at z = 2. The results
1 Myers et al. (2006) and Ross et al. (2009) determine r from QSOs as- indicate that SMGs are clustered similarly to QSOs at z ∼ 2 and can be
0
suming a power law correlation function with γ = 2. To estimate r0 for expected to evolve into luminous elliptical galaxies in the local Universe.
γ = 1.8, we multiply the quoted values by 0.8, appropriate for fits over the
range 1 R 100 h−1 Mpc.

c 2011 RAS, MNRAS 000, 1–13

2011), which has been shown to contain a somewhat smaller den-
sity of SMGs compared to other surveys (Weiß et al. 2009).
14.0 The ratio of these space densities yields a duty cycle (the frac-
tion of haloes that host an SMG at any given time) of ∼ 10%. We
13.5 ~2-3 L* assume the SMGs occupy the redshift range 1.5 < z < 2.5, which
ellipticals
log(Mhalo [h-1 MO])

includes roughly half of the SMGs in the Wardlow et al. (2011)
•

SMGs
sample and corresponds to ∆t = 1.6 Gyr. We thus obtain a life-
13.0 time for SMGs of tSMG = 110+280 Myr. Clearly, even our im-
−80
proved measurement of SMG clustering yields only a weak con-
12.5 straint on the lifetime, but this is consistent with lifetimes esti-
QSOs mated from gas consumption times and star-formation timescales
12.0
(e.g., Greve et al. 2005; Tacconi et al. 2006; Hainline et al. 2011)
bright and theoretical models of SMG fueling through mergers (e.g.,
LBGs
Mihos & Hernquist 1994; Springel, Di Matteo & Hernquist 2005;
11.5
Narayanan et al. 2010).
0 1 2 3 4 5
z Constraints on SMG descendants from clustering can also
yield insights into their their formation histories. Measurements
of the stellar plus molecular gas masses of SMGs from SED fit-
Figure 7. Broad schematic for the evolution of halo mass versus redshift
ting and dynamical studies are in the range ∼ (1–5) × 1011 M⊙
for SMGs, showing the approximate halo masses corresponding to likely
progenitors and descendants of SMGs. Lines indicate the median growth (Swinbank et al. 2006; Wardlow et al. 2011; Hainline et al. 2011;
rates of haloes with redshift (Fakhouri, Ma & Boylan-Kolchin 2010). SMG Ivison et al. 2011; Michałowski et al. 2011). While these estimates
host haloes are similar to those those of QSOs at z ∼ 2, and correspond to can be uncertain by factors of a few, they are in a similar range to
bright LBGs at z ∼ 5 (Hamana et al. 2004; Lee et al. 2006) and ∼ 2–3L∗ the stellar masses of SMG descendants as indicated by their clus-
ellipticals at z = 0 (Zehavi et al. 2011; Stott et al. 2011). tering, as discussed above. This correspondence suggests that if a
significant fraction of the molecular gas is converted to stars dur-
ing the SMG phase, then these galaxies will subsequently experi-
2011), a population dominated by ellipticals with predominantly ence relatively little growth in mass from z ∼ 2 to the present.
slow-rotating kinematics (e.g., Tempel et al. 2011; Cappellari et al. This in turn puts limits on the star formation history. Star-forming
2011). Assuming typical mass-to-light ratios for massive galaxies galaxies at z ∼ 2 typically exhibit specific star formation rates of
(e.g., Baldry, Glazebrook & Driver 2008), these luminosities ˙
M⋆ /M⋆ ∼ 2 Gyr−1 (Elbaz et al. 2011), at which the SMGs would
correspond to stellar masses ∼ (1.5–2.5) × 1011 M⊙ , in close only need to form stars for 500 Myr in order to double in mass.
agreement with direct measurements of the relationship between We may therefore conclude, from the clustering and stellar masses
halo mass and central galaxy stellar mass for X-ray selected groups alone, that the SMGs evolve from star-forming to passive states
and clusters, for which log M⋆ ≈ 0.27 log Mhalo + 7.6 (Stott et al. relatively quickly (within a Gyr or so) after the starburst phase,
2011). and that the descendants spend most of their remaining time as
relatively passive systems. This scenario is consistent with mea-
surements of the stellar populations in ∼ 2–3 L∗ ellipticals, which
4.4 SMG lifetime and star formation history have typical ages of ∼10 Gyr and show little evidence for younger
components (e.g., Nelan et al. 2005; Allanson et al. 2009), imply-
We next estimate the SMG lifetime, making the simple assumption ing that the vast majority of stars were formed above z ∼ 2 with
that every dark matter halo of similar mass passes through an SMG little additional star formation at lower redshifts.
phase3 , so that The halo masses of SMGs may also provide insight into
nSMG the processes that prevent their descendants from forming new
tSMG = ∆t , (8)
nhalo stars. Star formation can be shut off rapidly at the end of the
where ∆t is the time interval over the redshift range covered by the SMG phase, either by exhaustion of the gas supply, or by energy
SMG sample, and nSMG and nhalo are the space densities of SMGs input from a QSO (e.g., Di Matteo, Springel & Hernquist 2005;
and DM haloes, respectively. Springel, Di Matteo & Hernquist 2005). Powerful winds are ob-
Using the halo mass function of Tinker et al. (2008), the served in luminous AGN (e.g., Feruglio et al. 2010; Fischer et al.
space density of haloes with log (Mhalo [h−1 M⊙ ]) = 12.8+0.3 2010; Sturm et al. 2011; Greene et al. 2011) and have also been
−0.5
is dnhalo /d ln M = (2.1+7.3 ) × 10−4 Mpc−3 . We adopt a space seen in some SMGs (e.g., Alexander et al. 2010, Harrison et al. in
−1.5
−5 −3
density of SMGs at z ∼ 2 of ∼ 2 × 10 Mpc , correspond- preparation), although for the SMGs is unclear whether the winds
ing to results from previous surveys (e.g., Chapman et al. 2005; are driven by the starburst or AGN. Even if the formation of stars
Coppin et al. 2006; Schael et al. in preparation). This density is is rapidly quenched, over longer timescales the galaxy would be
∼ 50% higher than that observed in the LESS field (Wardlow et al. expected to accrete further gas from the surrounding halo, result-
ing in significant additional star formation (e.g., Bower et al. 2006;
Croton et al. 2006). Recent work suggests that energy from ac-
2 creting supermassive black holes, primarily in the form of radio-
Note that here we use the median growth rate of haloes, which for haloes
of ∼ 1013 h−1 M⊙ is ≈35% lower than the mean growth rate, owing to
bright relativistic jets, can couple to the hot gas in the surround-
the long high-mass tail in the halo mass distribution. ing halo, producing a feedback cycle that prevents rapid cooling
3 If the average halo experiences more or fewer SMG phases in the given (e.g., Rafferty, McNamara & Nulsen 2008). This mechanical black
time interval, the lifetime of each episode will be correspondingly shorter hole feedback is an key ingredient of successful models for the pas-
or longer, respectively. sive galaxy population (e.g., Croton et al. 2006; Bower et al. 2006;

c 2011 RAS, MNRAS 000, 1–13

Springel, Di Matteo & Hernquist 2005; Hopkins et al. 2006), sec-
1.2 ular instabilities (e.g., Mo, Mao & White 1998; Bower et al. 2006;
halo mass crossing rate
x molecular gas fraction Genzel et al. 2008) or accretion of recycled cold gas from evolved
normalized number counts; fmol

1.0 stars (Ciotti & Ostriker 2007; Ciotti, Ostriker & Proga 2010), and
is similar to the mass at which galaxy populations transition from
0.8 star-forming to passive (e.g., Coil et al. 2008; Brown et al. 2008;
Conroy & Wechsler 2009; Tinker & Wetzel 2010). The observed
0.6 fmol clustering of SMGs at z ∼ 2 from the present work is con-
sistent with that for QSOs, as well as highly active obscured
objects including powerful obscured AGN (H11; Allevato et al.
0.4
2011) and dust-obscured galaxies (Brodwin et al. 2008). Thus
these may indeed represent different phases in the same evolu-
0.2 tionary sequence, and energy input from the QSO may be re-
sponsible for the rapid quenching of star formation at the end
0.0
of the SMG phase (e.g., Di Matteo, Springel & Hernquist 2005;
0 2 4 6
Springel, Di Matteo & Hernquist 2005) as discussed in § 4.4.
z
A connection with QSOs may imply that triggering of SMGs
is also related (at least indirectly) to the mass of the parent DM
Figure 8. Redshift distribution of LESS SMGs (Wardlow et al. 2011), com- halo. In this case, the evolution of large-scale structure may broadly
pared to the simple models for SMG triggering based on the rate at which
explain why the SMG population peaks at z ∼ 2.5 and falls at
haloes cross a threshold mass Mthresh = 6 × 1012 h−1 M⊙ (see § 4.5).
The uncertainties in the number counts are an approximation of Poisson
higher and lower redshifts. In the simplest possible such scenario,
counting statistics (Gehrels 1986). The black dotted line shows the (arbi- SMG activity is triggered when the halo reaches a certain mass
trarily normalized) number of haloes crossing this threshold in each red- Mhalo = Mthresh (see Figure 16 of Hickox et al. 2009 for a
shift interval (Equation 9) while the dashed red line shows this distribution schematic illustration of this picture). In a given volume, the num-
multiplied by the evolution in the molecular gas fraction (Equation 10), ber of haloes crossing this mass threshold as a function of redshift
where fmol is taken from the model predictions of Lagos et al. (2011) and is:
is shown by the gray dot-dashed line. The remarkable agreement between dNthresh dV
the second model and the observed number counts suggests that the evolu- ˙
∝ nhalo (Mthresh , z)Mhalo (Mthresh , z)tSMG , (9)
tion of the SMG population can be described simply in terms of two quan-
dz dz
tities: the growth of DM structures and the variation with redshift of the ˙
where nhalo and Mhalo are the number density (e.g., Tinker et al.
molecular gas fraction in galaxies. 2008) and typical growth rate (Fakhouri, Ma & Boylan-Kolchin
2010), respectively, of haloes of mass Mthresh at redshift z, tSMG
is the SMG lifetime, and dV /dz is the differential comoving vol-
Bower, McCarthy & Benson 2008; Somerville et al. 2008). Inter- ume over the survey area. If an SMG is triggered every time a halo
estingly, the clustering of radio galaxies at z 0.8 indicates that reaches Mthresh , then the observed number density of SMGs will
they reside in haloes of mass 1013 h−1 M⊙ (e.g., Wake et al. be proportional to dNthresh /dz. However, the huge star forma-
2008a; Hickox et al. 2009; Mandelbaum et al. 2009; Donoso et al. tion rates of SMGs require a large reservoir of molecular gas (e.g.,
2010; Fine et al. 2011), precisely the environments that will host Greve et al. 2005; Tacconi et al. 2006, 2008), and the molecular gas
the descendants of SMGs. Thus the strong observed clustering for fraction increases strongly with redshift (e.g., Tacconi et al. 2010;
SMGs can relate them directly to the radio-bright active galactic Geach et al. 2011; Lagos et al. 2011). This evolution may explain
nucleus population that may regulate their subsequent star forma- why the most powerful starbursts at low redshift (ULIRGs) have
tion. lower typical SFRs than z ∼ 2 SMGs (e.g., Le Floc’h et al. 2005;
Rodighiero et al. 2010). Therefore it may be reasonable to assume
that the number counts of SMGs also depend on fmol , with the
4.5 Evolutionary links with QSOs and the SMG redshift simplest possible prescription being:
distribution dNSMG dNthresh
∝ fmol (z). (10)
Finally, the observed clustering of SMGs provides insights into dz dz
the processes that trigger and (possibly) shut off their rapid star In Figure 8 we show the observed redshift distribution of
formation activity. As discussed in § 1, powerful local starbursts LESS SMGs (Wardlow et al. 2011), compared to the distributions
(i.e. ULIRGs) are predominantly associated with major mergers predicted by Equations (9) and (10), assuming Mthresh = 6 × 1012
and appear to be associated with the fueling of luminous QSOs as h−1 M⊙ . For simplicity, the evolution in fmol is taken from pre-
part of an evolutionary sequence (e.g., Sanders et al. 1988). How- dictions of the GALFORM model of Lagos et al. (2011), which
ever it is unclear if a similar connection exists between SMGs and agrees broadly with observations (see Figure 2 of Geach et al.
high-z QSOs. One robust prediction of any evolutionary picture is 2011) and so provides a simple parameterisation of the current em-
that SMGs and QSOs must display comparable large-scale clus- pirical limits on the molecular gas fraction in galaxies. It is clear
tering, since the evolutionary timescales are signiﬁcantly smaller from Figure 8 that there is remarkable correspondence between our
than those for the growth of DM haloes. At all redshifts, QSOs extremely simple prescription and the observed redshifts of SMGs.
are found in haloes of similar mass ∼ a few ×1012 h−1 M⊙ Of course this “model” does not account for a wide range of pos-
ˆ
(e.g., Croom et al. 2005; Myers et al. 2006; da Angela et al. 2008; sible complications and the normalisations of the distributions are
Ross et al. 2009; Figure 6). The characteristic Mhalo provides arbitrary. However, this exercise clearly demonstrates that if SMGs,
a strong constraint on models of QSO fueling by the major like QSOs, are found in haloes of a characteristic mass, then their
mergers of gas-rich galaxies (e.g., Kauffmann & Haehnelt 2000; observed redshift distribution may be explained simply by two ef-

c 2011 RAS, MNRAS 000, 1–13

fects: the cosmological growth of structure combined with the evo- Adelberger K. L., Steidel C. C., Pettini M., Shapley A. E., Reddy
lution of the molecular gas fraction. Thus SMGs likely represent a N. A., Erb D. K., 2005, ApJ, 619, 697
short-lived but universal phase in massive galaxy evolution, associ- Alexander D. M., Bauer F. E., Chapman S. C., Smail I., Blain
ated with the transition between cold gas-rich, star-forming galax- A. W., Brandt W. N., Ivison R. J., 2005, ApJ, 632, 736
ies and passively evolving systems. Alexander D. M. et al., 2008, AJ, 135, 1968
Alexander D. M., Swinbank A. M., Smail I., McDermid R., Nes-
vadba N. P. H., 2010, MNRAS, 402, 2211
5 CONCLUSIONS Allanson S. P., Hudson M. J., Smith R. J., Lucey J. R., 2009, ApJ,
702, 1275
In this paper we measure the cross-correlation between SMGs and
Allevato V. et al., 2011, ApJ, 736, 99
galaxies in the LESS survey of ECDFS, and observe signiﬁcant
Almeida C., Baugh C. M., Lacey C. G., 2011, MNRAS, 417, 2057
clustering at the > 4σ level. We obtain an autocorrelation length for
Amblard A. et al., 2011, Nature, 470, 510
the SMGs of r0 = 7.7+1.8 h−1 Mpc, assuming γ = 1.8. This clus-
−2.3
Assef R. J. et al., 2011, ApJ, 728, 56
tering amplitude corresponds to a characteristic DM halo mass of
Baldry I. K., Glazebrook K., Driver S. P., 2008, MNRAS, 388,
log (Mhalo [h−1 M⊙ ]) = 12.8+0.3 . Using this estimate of Mhalo
−0.5
945
and the space density of SMGs, we obtain a typical SMG lifetime
+280 Bardeen J. M., Bond J. R., Kaiser N., Szalay A. S., 1986, ApJ,
of tSMG = 110−80 Myr.
304, 15
The observed clustering indicates that the low-redshift descen-
dants of typical SMGs are massive (∼ 2–3 L∗ ) elliptical galax- Barger A. J., Cowie L. L., Sanders D. B., Fulton E., Taniguchi Y.,
ies at the centers of moderate- to high-mass groups. This predic- Sato Y., Kawara K., Okuda H., 1998, Nature, 394, 248
tion is consistent with previous suggestions based on the dynami- Bennert V. N., Treu T., Woo J.-H., Malkan M. A., Le Bris A.,
cal (Swinbank et al. 2006) and stellar masses (e.g., Hainline et al. Auger M. W., Gallagher S., Blandford R. D., 2010, ApJ, 708,
2011) of SMGs, and is also consistent with observations of local 1507
massive ellipticals, which indicate that they formed the bulk of their Bertin E., Arnouts S., 1996, A&AS, 117, 393
stars at z > 2 and have been largely passive since. The clustering Biggs A. D. et al., 2011, MNRAS, 413, 2314
of SMGs is very similar to that observed for QSOs at the same Blain A. W., Chapman S. C., Smail I., Ivison R., 2004, ApJ, 611,
redshifts, consistent with evolutionary scenarios in which SMGs 725
and QSOs are triggered by a common mechanism. Assuming that Blain A. W., Smail I., Ivison R. J., Kneib J.-P., Frayer D. T., 2002,
SMGs, like QSOs, are transient phenomena that are observed in Phys. Rep., 369, 111
haloes of similar mass at all redshifts, the redshift distribution of Blake C., Pope A., Scott D., Mobasher B., 2006, MNRAS, 368,
SMGs can be explained remarkably well by the combination of the 732
cosmological growth of structure and the evolution of the molecu- Blanc G. A. et al., 2008, ApJ, 681, 1099
lar gas fraction in galaxies. Bolzonella M., Miralles J.-M., Pell´ R., 2000, A&A, 363, 476
o
This accurate clustering measurement thus provides a valuable Borys C., Chapman S., Halpern M., Scott D., 2003, MNRAS, 344,
observational constraint on the role of SMGs in the cosmic evolu- 385
tion of galaxies and large-scale structures. We conclude that SMGs Bothwell M. S. et al., 2010, MNRAS, 405, 219
likely represent a short-lived but universal phase in massive galaxy Bower R. G., Benson A. J., Malbon R., Helly J. C., Frenk C. S.,
evolution that is associated with the rapid growth of black holes Baugh C. M., Cole S., Lacey C. G., 2006, MNRAS, 370, 645
as luminous QSOs, and corresponds to the transition between cold Bower R. G., McCarthy I. G., Benson A. J., 2008, MNRAS, 390,
gas-rich, star-forming galaxies and passively evolving systems. 1399
Brodwin M. et al., 2008, ApJ, 687, L65
Brown M. J. I. et al., 2008, ApJ, 682, 937
ACKNOWLEDGMENTS Cappellari M. et al., 2011, MNRAS, 413, 813
Carrera F. J., Page M. J., Stevens J. A., Ivison R. J., Dwelly T.,
We thank the anonymous referee for helpful comments. RCH ac- Ebrero J., Falocco S., 2011, MNRAS, 413, 2791
knowledges support through an STFC Postdoctoral Fellowship and Chapman S. C., Blain A., Ibata R., Ivison R. J., Smail I., Morrison
AMS from an STFC Advanced Fellowship. IRS, DMA, ALRD, and G., 2009, ApJ, 691, 560
JPS acknowledge support from STFC. IRS acknowledges support Chapman S. C., Blain A. W., Ivison R. J., Smail I. R., 2003, Na-
through a Leverhulme Research Fellowship. DMA is grateful to the ture, 422, 695
Royal Society and the Leverhulme Trust for their generous support.
Chapman S. C., Blain A. W., Smail I., Ivison R. J., 2005, ApJ,
ADM was generously funded by the NASA ADAP program under
622, 772
grant NNX08AJ28G. JSD acknowledges the support of the Euro-
Ciotti L., Ostriker J. P., 2007, ApJ, 665, 1038
pean Research Council through the award of an Advanced Grant,
Ciotti L., Ostriker J. P., Proga D., 2010, ApJ, 717, 708
and the support of the Royal Society via a Wolfson Research Merit
Coil A. L. et al., 2009, ApJ, 701, 1484
award. This study is based on observations made with ESO tele-
Coil A. L., Hennawi J. F., Newman J. A., Cooper M. C., Davis
scopes at the Paranal and Atacama Observatories under programme
M., 2007, ApJ, 654, 115
numbers: 171.A-3045, 168.A-0485, 082.A-0890 and 183.A-0666.
Coil A. L. et al., 2008, ApJ, 672, 153
Conroy C., Wechsler R. H., 2009, ApJ, 696, 620
Cooray A. et al., 2010, A&A, 518, L22+
REFERENCES
Coppin K. et al., 2006, MNRAS, 372, 1621
Adelberger K. L., 2005, ApJ, 621, 574 Coppin K. E. K. et al., 2008, MNRAS, 389, 45
Adelberger K. L., Steidel C. C., 2005, ApJ, 630, 50 Croom S. M. et al., 2005, MNRAS, 356, 415

c 2011 RAS, MNRAS 000, 1–13

Croom S. M., Schade D., Boyle B. J., Shanks T., Miller L., Smith Lagos C. D. P., Baugh C. M., Lacey C. G., Benson A. J., Kim
R. J., 2004, ApJ, 606, 126 H.-S., Power C., 2011, MNRAS, 1776
Croton D. J. et al., 2006, MNRAS, 365, 11 Landy S. D., Szalay A. S., 1993, ApJ, 412, 64
ˆ
da Angela J. et al., 2008, MNRAS, 383, 565 Lauer T. R., Tremaine S., Richstone D., Faber S. M., 2007, ApJ,
Damen M. et al., 2011, ApJ, 727, 1 670, 249
Dav´ R., Finlator K., Oppenheimer B. D., Fardal M., Katz N.,
e Le Floc’h E. et al., 2005, ApJ, 632, 169
Kereˇ D., Weinberg D. H., 2010, MNRAS, 404, 1355
s Lee K.-S., Giavalisco M., Gnedin O. Y., Somerville R. S., Fergu-
Decarli R., Falomo R., Treves A., Labita M., Kotilainen J. K., son H. C., Dickinson M., Ouchi M., 2006, ApJ, 642, 63
Scarpa R., 2010, MNRAS, 402, 2453 Li C., Kauffmann G., Wang L., White S. D. M., Heckman T. M.,
Di Matteo T., Springel V., Hernquist L., 2005, Nature, 433, 604 Jing Y. P., 2006, MNRAS, 373, 457
Donoso E., Li C., Kauffmann G., Best P. N., Heckman T. M., Lindner R. R. et al., 2011, ApJ, 737, 83
2010, MNRAS, 407, 1078 Mandelbaum R., Li C., Kauffmann G., White S. D. M., 2009, MN-
Dunlop J. S., 2011, ASP Conf. Series in press (arXiv:1108.5679) RAS, 393, 377
Eales S., Lilly S., Gear W., Dunne L., Bond J. R., Hammer F., Le Marconi A., Axon D. J., Maiolino R., Nagao T., Pastorini G.,
F` vre O., Crampton D., 1999, ApJ, 515, 518
e Pietrini P., Robinson A., Torricelli G., 2008, ApJ, 678, 693
Elbaz D. et al., 2011, A&A, 533, A119 Merloni A. et al., 2010, ApJ, 708, 137
Engel H. et al., 2010, ApJ, 724, 233 Michałowski M. J., Dunlop J. S., Cirasuolo M., Hjorth
Estrada J., Sefusatti E., Frieman J. A., 2009, ApJ, 692, 265 J., Hayward C. C., Watson D., 2011, A&A submitted
Fakhouri O., Ma C.-P., Boylan-Kolchin M., 2010, MNRAS, 406, (arXiv:1108.6058)
2267 Mihos J. C., Hernquist L., 1994, ApJ, 431, L9
Feruglio C., Maiolino R., Piconcelli E., Menci N., Aussel H., Mo H. J., Mao S., White S. D. M., 1998, MNRAS, 295, 319
Lamastra A., Fiore F., 2010, A&A, 518, L155+ Myers A. D., Brunner R. J., Nichol R. C., Richards G. T., Schnei-
Fine S., Croom S. M., Bland-Hawthorn J., Pimbblet K. A., Ross der D. P., Bahcall N. A., 2007, ApJ, 658, 85
N. P., Schneider D. P., Shanks T., 2010, MNRAS, 409, 591 Myers A. D. et al., 2006, ApJ, 638, 622
Fine S., Shanks T., Nikoloudakis N., Sawangwit U., 2011, MN- Myers A. D., White M., Ball N. M., 2009, MNRAS, 399, 2279
RAS in press (arXiv:1107.5666) Narayanan D., Hayward C. C., Cox T. J., Hernquist L., Jonsson
Fischer J. et al., 2010, A&A, 518, L41 P., Younger J. D., Groves B., 2010, MNRAS, 401, 1613
Gawiser E., Wolfe A. M., Prochaska J. X., Lanzetta K. M., Yahata Nelan J. E., Smith R. J., Hudson M. J., Wegner G. A., Lucey J. R.,
N., Quirrenbach A., 2001, ApJ, 562, 628 Moore S. A. W., Quinney S. J., Suntzeff N. B., 2005, ApJ, 632,
Geach J. E., Smail I., Moran S. M., MacArthur L. A., Lagos 137
C. d. P., Edge A. C., 2011, ApJ, 730, L19+ Netzer H., Marziani P., 2010, ApJ, 724, 318
Gehrels N., 1986, ApJ, 303, 336 Norberg P., Baugh C. M., Gazta˜ aga E., Croton D. J., 2009, MN-
n
Genzel R. et al., 2008, ApJ, 687, 59 RAS, 396, 19
Gilli R. et al., 2007, A&A, 475, 83 Overzier R. A., R¨ ttgering H. J. A., Rengelink R. B., Wilman
o
—, 2009, A&A, 494, 33 R. J., 2003, A&A, 405, 53
Greene J. E., Zakamska N. L., Ho L. C., Barth A. J., 2011, ApJ, Page M. J., Stevens J. A., Ivison R. J., Carrera F. J., 2004, ApJ,
732, 9 611, L85
Greve T. R. et al., 2005, MNRAS, 359, 1165 Peebles P. J. E., 1993, Principles of physical cosmology. Princeton
˚
G¨ sten R., Nyman L. A., Schilke P., Menten K., Cesarsky C.,
u University Press, Princeton, NJ
Booth R., 2006, A&A, 454, L13 Peterson B. M. et al., 2004, ApJ, 613, 682
Hainline L. J., Blain A. W., Smail I., Alexander D. M., Armus L., Pope A. et al., 2008, ApJ, 675, 1171
Chapman S. C., Ivison R. J., 2011, ApJ, 740, 96 Quadri R. et al., 2007, ApJ, 654, 138
Hamana T., Ouchi M., Shimasaku K., Kayo I., Suto Y., 2004, MN- Quadri R. F., Williams R. J., Lee K., Franx M., van Dokkum P.,
RAS, 347, 813 Brammer G. B., 2008, ApJ, 685, L1
Hickox R. C. et al., 2009, ApJ, 696, 891 Rafferty D. A., McNamara B. R., Nulsen P. E. J., 2008, ApJ, 687,
—, 2011, ApJ, 731, 117 899
Hopkins P. F., Somerville R. S., Hernquist L., Cox T. J., Robertson Richards G. T. et al., 2006, AJ, 131, 2766
B., Li Y., 2006, ApJ, 652, 864 Rodighiero G. et al., 2010, A&A, 515, A8+
Hughes D. H. et al., 1998, Nature, 394, 241 Ross N. P. et al., 2009, ApJ, 697, 1634
Ivison R. J., Papadopoulos P. P., Smail I., Greve T. R., Thomson Sanders D. B., Soifer B. T., Elias J. H., Madore B. F., Matthews
A. P., Xilouris E. M., Chapman S. C., 2011, MNRAS, 412, 1913 K., Neugebauer G., Scoville N. Z., 1988, ApJ, 325, 74
Kaiser N., 1984, ApJ, 284, L9 Scott S. E. et al., 2002, MNRAS, 331, 817
Kauffmann G., Haehnelt M., 2000, MNRAS, 311, 576 Shen Y., Greene J. E., Strauss M. A., Richards G. T., Schneider
Kelly B. C., Vestergaard M., Fan X., Hopkins P., Hernquist L., D. P., 2008, ApJ, 680, 169
Siemiginowska A., 2010, ApJ, 719, 1315 Shen Y., Kelly B. C., 2010, ApJ, 713, 41
Kim J., Edge A. C., Wake D. A., Stott J. P., 2011, MNRAS, 410, Sheth R. K., Mo H. J., Tormen G., 2001, MNRAS, 323, 1
241 Siringo G. et al., 2009, A&A, 497, 945
Kollmeier J. A. et al., 2006, ApJ, 648, 128 Smail I., Ivison R. J., Blain A. W., 1997, ApJ, 490, L5
Kotilainen J. K., Falomo R., Decarli R., Treves A., Uslenghi M., Smith R. E. et al., 2003, MNRAS, 341, 1311
Scarpa R., 2009, ApJ, 703, 1663 Somerville R. S., Hopkins P. F., Cox T. J., Robertson B. E., Hern-
Krumpe M., Miyaji T., Coil A. L., 2010, ApJ, 713, 558 quist L., 2008, MNRAS, 391, 481
Lacey C., Cole S., 1993, MNRAS, 262, 627 Springel V., Di Matteo T., Hernquist L., 2005, MNRAS, 361, 776

c 2011 RAS, MNRAS 000, 1–13

Starikova S. et al., 2011, ApJ, 741, 15
Stevens J. A., Page M. J., Ivison R. J., Carrera F. J., Mittaz J. P. D.,
Smail I., McHardy I. M., 2005, MNRAS, 360, 610
Stott J. P., Kim J., Edge A. C., Wake D. A. a., 2011, MNRAS
submitted, 410, 241
Sturm E. et al., 2011, ApJ, 733, L16
Swinbank A. M., Chapman S. C., Smail I., Lindner C., Borys C.,
Blain A. W., Ivison R. J., Lewis G. F., 2006, MNRAS, 371, 465
Tacconi L. J. et al., 2010, Nature, 463, 781
—, 2008, ApJ, 680, 246
—, 2006, ApJ, 640, 228
Tempel E., Saar E., Liivam¨ gi L. J., Tamm A., Einasto J., Einasto
a
M., M¨ ller V., 2011, A&A, 529, A53+
u
Tinker J., Kravtsov A. V., Klypin A., Abazajian K., Warren M.,
Yepes G., Gottl¨ ber S., Holz D. E., 2008, ApJ, 688, 709
o
Tinker J. L., Wetzel A. R., 2010, ApJ, 719, 88
Vestergaard M., 2002, ApJ, 571, 733
Vestergaard M., Peterson B. M., 2006, ApJ, 641, 689
Wake D. A., Croom S. M., Sadler E. M., Johnston H. M., 2008a,
MNRAS, 391, 1674
Wake D. A. et al., 2008b, MNRAS, 387, 1045
Wardlow J. L. et al., 2011, MNRAS, 415, 1479
Webb T. M. et al., 2003, ApJ, 582, 6
Weiß A. et al., 2009, ApJ, 707, 1201
Williams C. C. et al., 2011, ApJ, 733, 92
Xia J.-Q., Negrello M., Lapi A., De Zotti G., Danese L., Viel M.,
2011, arXiv:1111.4212
Zehavi I. et al., 2011, ApJ, 736, 59
—, 2005, ApJ, 630, 1

c 2011 RAS, MNRAS 000, 1–13

Laboca survey of_extended_chandra_deep_field_south

More Related Content

What's hot (20)

Viewers also liked (18)

Similar to Laboca survey of_extended_chandra_deep_field_south (20)

More from Sérgio Sacani (20)

Recently uploaded (20)

Laboca survey of_extended_chandra_deep_field_south