The Canada-UK Deep Submillimeter Survey VI: The 3-Hour Field

Abstract

We present the complete submillimeter data for the Canada-UK Deep Submillimeter Survey (CUDSS) 3-hour field. The obeservations were taken with the Submillimeter Common-User Bolometer Array (SCUBA) on the James Clerk Maxwell Telescope on Mauna Kea. The 3-hour field is one of two main fields in our survey and covers 60 square arcminutes to a 3-sigma depth of 3 mJy. In this field we have detected 27 sources above 3-sigma and 15 above 3.5-sigma. We assume the source counts follow the form $N(S) {\propto} S^{-\alpha}$ and measure $\alpha$ = 3.3$^{+1.4}_{-1.0}$. This is in good agreement with previous studies and further supports our claim (Eales et al., 2000) that SCUBA sources brighter than 3 mJy produce ~20% of the 850$\mu$m background energy. Using preliminary ISO 15 $\mu$m maps and VLA 1.4 GHz data we have identified counterparts for six objects and have marginal detections at 450$\mu$m for two additional sources. With this information we estimate a median redshift for the sample of 2.0$\pm$0.5, with $\sim$10% lying at $z<$ 1. We have measured the angular clustering of S850 > 3 mJy sources using the source catalogues from the CUDSS two main fields, the 3-hour and 14-hour fields, and find a marginal detection of clustering, primarily from the 14-hour field, of $\omega(\theta)=4.4\pm2.9 \theta^{-0.8}$. This is consistent with clustering at least as strong as that seen for the Lyman-break galaxy population and the Extremely Red Objects. Since SCUBA sources are selected over a broader range in redshifts than these two populations the strength of the true spatial clustering is expected to be correspondingly stronger.

Full text

Page 1

arXiv:astro-ph/0009154v1 11 Sep 2000
The Canada-UK Deep Submillimeter Survey: IV. The Survey of the 14-Hour
Field
Stephen Eales1, Simon Lilly2, Tracy Webb2, Loretta Dunne1, Walter Gear1,David Clements1and
Min Yun3
ABSTRACT
We have used SCUBA to survey an area of ≃ 50 arcmin2, detecting 19 sources down
to a 3σ sensitivity limit of ∼3.5 mJy at 850µm. Monte-Carlo simulations have shown
that the fluxes of sources in this and similar SCUBA surveys are biased upwards by
the effects of source confusion and noise, leading to an overestimate by a factor of ∼1.4
in the fraction of the 850µm background that has been resolved by SCUBA. Once a
correction is made for this effect, about 20% of the background has been resolved. The
simulations have also been used to quantify the effects of confusion on source positions.
Of the 19 SCUBA sources, five are µJy radio sources and two are ISO 15µm sources.
The radio/submillmetre flux ratios imply that the dust in these galaxies is being heated
by young stars rather than AGN. The upper limit to the average 450µm/850µm flux
ratio implies either that the SCUBA galaxies are at z >> 2 or, if they are at lower
redshifts, that the dust is generally colder than in ULIRGs.
We have used simple evolution models to address the major questions about the
SCUBA sources: (1) what fraction of the star formation at high redshift is hidden by
dust? (2) Does the submillimetre luminosity density reach a maximum at some redshift?
(3) If the SCUBA sources are proto-ellipticals, when exactly did ellipticals form? We
show, however, that the observations are not yet good enough to answer these questions.
There are, for example, acceptable models in which 10 times as much high-redshift star
formation is hidden by dust as is seen at optical wavelengths, but also acceptable ones
in which the amount of hidden star formation is less than that seen optically. There
are also acceptable models in which very little star formation occurred before a redshift
of three (as might be expected in models of hierarchical galaxy formation), but also
ones in which 30% of the stars have formed by this redshift. The key to answering
these questions are measurements of the dust temperatures and redshifts of the SCUBA
sources.
Subject headings: galaxies: evolution — galaxies: formation — infrared: galaxies —
cosmology: observations — surveys
1Department of Physics and Astronomy, Cardiff University, P.O. Box 913, Cardiff CF2 3YB, UK
2Department of Astronomy, University of Toronto, 60 St. George Street, Toronto, Ontario M5S 1A1, Canada
3National Radio Astronomy Observatory, P.O. Box 0, 1003 Lopezville Road, Socorro, NM 87801

Page 2

– 2 –
1.Introduction
The extragalactic background contains all the energy ever emitted by galaxies, with the radi-
ation from galaxies at a redshift z being weighted by a factor (1 + z)−1. Thus determing the level
of the background in different wavebands and resolving the background into individual sources
is of great importance for determing the history of the energy output of galaxies. Recent mea-
surements suggest that (setting aside the cosmic microwave-background radiation from the very
early universe) the extragalactic background is dominated by emission in two spectral regions, the
optical/near-infrared and the submillimetre wavebands (100µm < λ < 1mm), with roughly equal
integrated emission in the two wavebands (Dwek et al. 1998). The spectral shape of the back-
ground emission in the submillimetre waveband is characteristic of dust emission spread over a
range of redshift (Puget et al. 1996; Lagache, Puget and Gispert 1999). Therefore, although there
is some uncertainty from the (1 + z)−1weighting factor, this approximate equality suggests that
half the energy ever directly emitted by stars and active galactic nuclei (AGN) has been absorbed
by dust and then re-radiated at long wavelengths. Since young stars are generally more heavily
obscured than old stars, the fraction the of the universe’s star formation that is hidden by dust
may well be significantly greater than half. The only caveat to these arguments is if there is some
more exotic population of objects dominating the submillimetre background (Bond, Carr & Hogan
1986, 1991).
If the extragalactic background is dominated by stars, the history of the energy output of
galaxies is closely linked to the history of star formation in the universe. The background radiation
and the local properties of galaxies give complementary information about this history. Consider
the problem of the origin of elliptical galaxies and spiral bulges. The stars in nearby ellipticals and
spiral bulges are often extremely old (Bower, Lucey & Ellis 1992), implying that much of the star
formation in these objects must have occurred at very early times. Ellipticals and spiral bulges
in this initial star-forming phase should be very luminous, yet optical surveys have failed to find
convincing evidence of such “proto-spheroids” (De Propris et al. 1993). However, these objects
must make a major contribution to the extragalactic background radiation. The relation between
the integrated background emission produced by a population at a redshift z and the average cosmic
density of processed material, < ρ(Z + ∆Y ) >, produced by this population is:
?∞
0
Iνdν =0.007 < ρ(Z + ∆Y ) > c3
4π(1 + z)
,(1)
(Pagel 1997). Given the quantity of metals associated with nearby ellipticals (Edmunds and Phillipps 1997),
this relation implies that ellipticals in their rapid star-forming phase should be responsible for about
half the background radiation. Thus resolving the extragalactic background into individual sources
must solve this problem.
The measurement of the integrated optical background recently made by Bernstein and col-

Page 3

– 3 –
laborators (Bernstein 1999) is only about twice that obtained by simply adding up the emission in
deep galaxy counts (Williams et al. 1996) and indeed Bernstein argues that the integrated optical
background can be explained by known populations of objects. Thus the optical background may
already have been completely resolved and no obvious proto-spheroids have been found. For exam-
ple, the galaxies at z ∼ 3 found by the Lyman-break technique (Steidel et al. 1999) contribute ∼2%
of the background in the I-band and thus, by the metallicity argument above, are unlikely to be be
the missing proto-spheroids. This apparent failure to find proto-spheroids at optical wavelengths,
together with the metallicity argument, suggests that proto-spheroids must dominate the other im-
portant component of the background at submillimetre wavelengths. A physical explanation of this
would be if the initial formation of stars in the spheroids leads to the rapid creation of dust, which
then absorbs the optical/UV radiation (Eales and Edmunds 1996; Eales and Edmunds 1997).
Resolving the submillimetre background radiation into individual sources is thus of great in-
terest. It has recently become possible to do this, at least partially, with the ISOPHOT instrument
on the Infrared Space Observatory at short wavelengths and with the SCUBA submillimeter array
(Holland et al. 1999) on the James Clerk Maxwell Telescope (JCMT) at long wavelengths. This
paper is the fourth of a series of papers describing the results of a deep submillimetre survey with
SCUBA (Eales et al. 1999; Lilly et al. 1999—henceforth Papers I and II). Before describing the
scope of the present paper, we will briefly summarize the results of our and other submillimetre
surveys.
The ISOPHOT surveys are important because the submillimetre background radiation peaks
at short wavelengths, and thus, in energy terms, it is more important to resolve the background
here than at the longer wavelengths sampled by SCUBA. Surveys with ISOPHOT have found
sources that contribute about 10% of the submillimetre background at 175 µm (Kawara et al. 1998;
Puget et al. 1999). However, the large size of the ISOPHOT beam (1.9 arcmin, full-width half-
maximum) means that it is remarkably difficult to identify the galaxies responsible for the submil-
limetre emission (Scott et al. 2000).
The main operating wavelength of SCUBA is 850µm, where the submillimetre background is
much lower than at its peak (νIνlower by a factor of ∼30), but SCUBA has the great advantage over
ISOPHOT that its much smaller beamsize (14 arcsec, full-width half-maximum) means that it is
possible (although still difficult) to identify the galaxies responsible for the submillimetre emission.
SCUBA has been used to investigate the high-redshift universe through three different methods.
The first is to obtain submillimetre images in the directions of rich clusters, thus using the lensing
effect of the clusters both to amplify the submillimetre fluxes of high-redshift submillimetre sources
and to reduce the effect of source confusion (Smail, Ivison & Blain 1997; Smail et al. 1998; Blain
et al. 1999b). The disadvantages of this are the need to separate background from cluster sources
and for an accurate lensing model for each cluster. The second method is to carry out truly ‘blind-
field’ submillimetre surveys, which do not have the disadvantages of the cluster method but at the
expense of greater problems with confusion and much longer integration times to get to the same
effective sensitivity. The deepest blind survey that has so far been carried out is a survey of the

Page 4

– 4 –
Hubble Deep Field (Hughes et al. 1998). Our survey, which is not so deep but which covers a wider
area, is also of fields which have been extensively surveyed at other wavelengths: the Canada-France
Redshift Survey fields (CFRS; Lilly et al. 1995). The Hawaii group has been carrying out a survey
with a similar depth to our own (Barger et al. 1998; Barger, Cowie & Sanders 1999). The third
of the methods is to carry out SCUBA observations of known high-redshift objects. This will be
discussed below but we will first summarize the scientific results of the surveys.
It has been claimed (Blain et al. 1999b,c) that the SCUBA surveys have resolved close to
100% of the background at 850µm, although we will argue in this paper that this is an overesti-
mate. Despite controversy over the details of which galaxies are responsible for the submillimetre
emission, caused by the large positional errors of SCUBA, all the teams carrying out SCUBA surveys
agree that these galaxies are similar to the ultra-luminous infrared galaxies (ULIRG’s; Sanders &
Mirabel 1996) found in the local universe (Hughes et al. 1998; Barger et al. 1998; Lilly et al. 1999;
Smail et al. 1998). From these two results one can draw an extremely important conclusion (Paper
I). If ULIRG’s make up >50% of the submillimetre background, they must constitute >≃25% of
the total extragalactic background radiation; and thus, as long as the dust in ULIRG’s is heated
by stars rather than AGN, roughly one quarter of all the stars that have ever formed must have
formed in extreme systems like this rather than in normal galaxies like our own. This fact, to-
gether with the extreme bolometric luminosities of ULIRG’s, which imply star-formation rates of
102−103M⊙year−1(Rieke et al. 1985), make it difficult to avoid the conclusion that these systems
are proto-spheroids.
Hughes et al. (1998) and Barger, Cowie and Richards (2000) have also argued that the results
of the SCUBA surveys and the strength of the submillimetre background imply the amount of star
formation hidden by dust in the early universe is an order of magnitude greater than that seen
directly in optical surveys—in effect that the universe had a ‘Dark Age’ in which the fraction of
star formation hidden by dust was much greater than it is today. In our earlier work, we found
little evidence for this, concluding that the cosmic evolution seen in the submillimetre waveband is
similar to that seen at optical wavelengths and thus that the fraction of young stars hidden by dust
is the same at all cosmic epochs (Paper I). Moreover, the fraction of the SCUBA sources identified
with galaxies at z < 1 suggested that the submillimetre luminosity-density declines at z > 2 (Paper
II), which is where the other studies claim the majority of the dust emission is occurring. We will
revisit this question in this paper.
There are two fundamental uncertainties in the conclusions above. The first of these is simply
the sensitivity of these conclusions to dust temperature, which for the SCUBA sources is largely
unknown. Up to z ∼ 3, the main SCUBA operating wavelength of 850µm falls, in the rest frame, on
the Rayleigh-Jeans side of the typical dust spectral energy distribution, and thus dust temperature
is critical for calculating the total dust luminosity; for a dust source with a dust emissivity index of
two, the total dust emission is proportional to the dust temperature to the sixth power, and thus an
uncertainty of a factor of two in dust temperature leads to an uncertainty of a factor of 64 in total
luminosity. This uncertainty holds for both individual sources and for the population of SCUBA

Page 5

– 5 –
sources as a whole. Even the conclusion that a SCUBA galaxy is essentially a ULIRG is affected.
Fig. 1 shows the bolometric luminosity calculated for a SCUBA source with S850µm= 4mJy as a
function of redshift and for a number of assumed dust temperatures. Also shown is the bolometric
luminosity of the archetypical ULIRG Arp 220. If instead of assuming a dust temperature typical of
a ULIRG, for which there is rarely any observational justification, one assumes a dust temperature
typical of a normal spiral galaxy, one obtains a bolometric luminosity an order-of-magnitude lower
than that of Arp 220. Thus, the argument that the estimated bolometric luminosities imply SCUBA
galaxies are ULIRGs is circular: if one assumes the dust temperature typical of a ULIRG, one
necessarily obtains the bolometric luminosity of a ULIRG. We will explore the importance of this
uncertainty for the population as a whole later in this paper.
The second of the uncertainties is whether the dust in a SCUBA galaxy is actually being heated
by young stars or whether it is being heated by an obscured active galactic nucleus (AGN); if it is
the latter, of course, then the conclusion that the SCUBA galaxies are proto-ellipticals is completely
wrong. Two results suggest that the AGN hypothesis may be the correct one. First, the recent
discovery that there are black holes in the centres of most nearby galaxies (Magorrian et al. 1998)
suggests that most galaxies have passed through a phase in which they harboured an active nucleus,
implying in turn that AGN might contribute a significant fraction of the extragalactic background
radiation. Estimating the exact fraction of the total extragalactic background emission that is
produced by AGN is difficult, mainly because one does not know how efficiently mass was turned
into radiation as the black hole formed (Haehnelt, Natarajan and Rees 1998), and even simple
estimates range from 10% (Eales et al. 1999) to 60% (Lilly, unpublished). Second, it has been
known for a long time (Setti & Woltjer 1989) that one way of explaining the spectral dependence
of the X-ray background is by a new population of highly obscured AGN, and the submillimetre
waveband would be the natural place for the absorbed emission to reappear.
Settling this issue is unfortunately difficult. Although the evidence from mid-infrared line
ratios (Genzel et al. 1998; Lutz et al. 1998) and from VLBI radio observations (Smith et al. 1998)
is that stars rather than AGN are the main energy source in low-redshift ULIRGs, it is not possible
to use these techniques for the SCUBA galaxies. One result that does suggest that this is also
the case for the SCUBA galaxies is that they are often detected in very deep radio surveys with
ratios of dust to radio emission similar to those seen in samples of nearby star-forming galaxies
(Lilly et al. 1999; Barger, Cowie & Richards 2000). Barger et al., for example, found that five of
the seven sources detected in a SCUBA survey of the Hubble Flanking Fields were detected in the
radio survey of the same region. One way of addressing the possibility that the SCUBA galaxies
are the obscured AGN needed to explain the X-ray background would be deep X-ray surveys of
fields surveyed with SCUBA, something now possible with Chandra and Newton. The first of these
surveys (Fabian et al. 2000; Hornschemeier et al. 2000) suggest that it is young stars rather than
AGN heating the dust.
The third of the methods used to investigate the high-redshift universe with SCUBA has been
to use SCUBA to observe known classes of object. Observations of samples of high-redshift radio