Tracing the evolution of dust-obscured activity using sub-millimetre galaxy populations from STUDIES and AS2UDS

TL;DR: In this paper, the physical properties of 121 SNR and 5 sub-millimetre galaxies were analyzed using MAGPHYS+photo-z and compared to similar modelling of 850μm-selected SMG sample from AS2UDS to understand the fundamental physical differences between the two populations at the observed depths.

Abstract: We analyse the physical properties of 121 SNR ≥ 5 sub-millimetre galaxies (SMGs) from the STUDIES 450 μm survey. We model their UV-to-radio spectral energy distributions using MAGPHYS+photo-z and compare the results to similar modelling of 850 μm-selected SMG sample from AS2UDS, to understand the fundamental physical differences between the two populations at the observed depths. The redshift distribution of the 450-μm sample has a median of z = 1.85 ± 0.12 and can be described by strong evolution of the far-infrared luminosity function. The fainter 450-μm sample has ∼14 times higher space density than the brighter 850-μm sample at z ≲ 2, and a comparable space density at z = 2–3, before rapidly declining, suggesting LIRGs are the main obscured population at z ∼ 1–2, while ULIRGs dominate at higher redshifts. We construct rest-frame ∼180-μm-selected and dust-mass-matched samples at z = 1–2 and z = 3–4 from the 450 and 850-μm samples, respectively, to probe the evolution of a uniform sample of galaxies spanning the cosmic noon era. Using far-infrared luminosity, dust masses, and an optically thick dust model, we suggest that higher redshift sources have higher dust densities due to inferred dust continuum sizes which are roughly half of those for the lower redshift population at a given dust mass, leading to higher dust attenuation. We track the evolution in the cosmic dust mass density and suggest that the dust content of galaxies is governed by a combination of both the variation of gas content and dust destruction time-scale.

Summary

Introduction

• Key words: galaxies: evolution – galaxies: starburst – infrared: galaxies.
• These surveys confirmed the cosmological significance of far-infrared-luminous galaxies, in particular their potentially significant contribution to the starformation rate density at high redshifts (see Madau & Dickinson 2014).
• While covering huge areas, these surveys are limited in sensitivity due to the large beam size and resulting bright confusion limit,1 which makes it challenging to detect all but the brightest sources at z 1 (Symeonidis, Page & Seymour 2011, although see Shu et al.
• The identification of differences in the physical properties of SMGs selected in the different sub-millimetre wavebands from such studies have been limited by their modest sample sizes and also their biases towards brighter sources due to the sensitivity limits at 450μm.

2.1 Photometric coverage

• STUDIES is a SCUBA-2 450-μm imaging survey within the CANDELS region in the COSMOS field.
• Briefly, the data from STUDIES, combined with archival data taken by Geach et al. (2013) and Casey et al. (2013), yields the deepest single-dish map currently available at 450μm, reaching a 1-σ noise level of 0.65 mJy.
• The 850-μm flux densities of the 450-μm-selected STUDIES sources were obtained from the 850-μm map at the 450- μm positions.
• The authors provide a brief description of the counterpart identification and multi-wavelength photometric data available for the sample from UV to radio wavelengths, which is then used to model the SEDs of the sources.

2.1.1 Counterpart identification

• The identification of optical counterparts for the STUDIES 450- μm sources is described in Lim et al. (2020).
• Briefly, the 450-μm sources were matched with the VLA-COSMOS 3-GHz catalogue (Smolčić et al. 2017) using a 4 arcsec search radius (set by the JCMT 450-μm beam) yielding ∼1 per cent false positive rate (based on the probability of false matches using the number densities of both catalogues).
• In total 109 of the counterparts have IRAC detections.

2.1.2 Far-infrared to radio observations

• To constrain the SED of each galaxy at radio wavelengths the authors utilize 1.4 and 3-GHz data from the Very Large Array (VLA)-COSMOS Large Project (Schinnerer et al. 2010; Smolčić et al. 2017).
• Hence, to better constrain the shape of the far-infrared SEDs for the galaxies in their sample, and so improve the constraints on the far-infrared luminosities, the authors include observations with the Spectral and Photometric Imaging Receiver (SPIRE: Griffin et al. 2010) and the Photodetector Array Camera and Spectrometer (PACS: Poglitsch et al. 2010) on the Herschel Space Observatory.
• The deblending of the SPIRE maps used positional priors for sources based on the 3-GHz and 24-μm (see below) catalogues, as well as machinelearning identified SMG counterparts from An et al. (2019) (see Section 2.1.1).
• The uncertainties on the flux densities (and limits) are found by attempting to recover model sources injected into the maps (see Swinbank et al. 2014 for details), and yield typical 3-σ detection limits of 7.0 and 8.0 mJy at 250 and 350μm, respectively.

2.1.3 Optical to near-/mid-infrared observations

• To model the stellar SEDs of the counterparts to their 450-μm sources, the authors require the photometry in the optical/infrared bands.
• For the u∗Bgrizy bands the authors adopt the photometry from COSMOS2015 (Laigle et al. 2016) catalogue.
• In an equivalent manner to Simpson et al. (2020), the authors measure 2 arcsec diameter aperture photometry at the positions of each SMG in each band.
• For each filter, the correction is determined by convolving the filter response with the scaled extinction curve.

2.2 SED fitting model

• To derive the physical properties of the STUDIES SMGs, the authors employ MAGPHYS+photo-z (da Cunha et al. 2015; Battisti et al. 2019) to model the SEDs from optical to radio wavelengths, using the available photometry in 24 bands.
• This approach allows for a simple comparison with similar fits to other 450, 850-μm, and ALMA samples.
• The authors stress that it is likely that the dust emission from the sources in their sample is not optically thin in the far-infrared and T MBBd and Md (through strong dependence on T MBBd ) are affected by the dust opacity assumptions.
• The authors then re-run MAGPHYS+photo-z with the adjusted photometry to obtain the bestfitting SEDs (all SEDs are shown in Fig. 1b), redshift and physical properties for the 450-μm sources.
• The authors note that, for consistency, they use the MAGPHYS+photo-z derived photometric redshifts for all sources in the analysis in this paper.

3 A NA LY SIS AND RESULTS

• The authors investigate the broad photometric properties and the derived physical properties of the 450-μm sample based on their MAGPHYS+photo-z analysis of their SEDs.
• The authors compare the results of the 450-μm-selected sample to an 850-μm-selected sample, AS2UDS (Stach et al. 2019), which has been analysed in a consistent manner by fitting MAGPHYS+photo-z to the available photometry in 22 bands (D20).
• AS2UDS is a follow-up survey of sources detected in the SCUBA2 Cosmology Legacy Survey (S2CLS; Geach et al. 2017) 850-μm map of the ∼0.9 deg2 UKIDSS UDS field and provides a large homogeneously selected sample of ALMA-identified SMGs.
• Throughout the paper, the authors will refer to this as the 850-μm sample (they note that this ALMA selection formally corresponds to 870μm, which is the wavelength used in the analysis).

3.1 Photometric properties of 450-μm sources

• Before the authors discuss the physical properties of the STUDIES 450- μm SMGs in detail, they first investigate the observed and rest-frame optical and infrared colour properties of the sample.
• The authors overlay the track of the composite SED of the 450-μm sample (further discussed in Section 3.3) as a function of redshift, which demonstrates that IRAC colours indicative of AGN are degenerate with those expected for dusty star-forming galaxies at z 2–3, where many members of this population lie.
• The authors indicate the median for each sample as a large circle in the respective colour, with the 16–84th percentile range shown as the black error bar.
• The authors note that the composite SED of the 450-μm sample has bluer IRAC colours at z 2 compared to the 850-μm composite (see also Section 3.3.2).

3.2 Redshift distribution

• The redshift distribution is a fundamental quantity providing constraints on formation models for the given population and is also essential for reliable derivation of their intrinsic properties and evolutionary trends.
• To compare the 450 and 850μm selections in more detail, the authors take advantage of their well-defined and almost effectively complete redshift distribution to investigate the space density of the 450-μmselected population.
• Thus, the authors estimate the survey area for each of the sources by calculating the area within the SCUBA-2 map within which each SMG would be detectable at SNR 5, given their 450-μm flux density, using the 450-μm RMS map from Lim et al. (2020).
• The authors test whether the two distributions are significantly different by using a χ2 test to compare the space density values at each bin including the errors.

3.3.1 Far-infrared properties

• As the majority of the emission from these dusty systems is coming from the far-infrared, the authors begin by investigating the dust properties of the SMGs by deriving their far-infrared luminosities.
• The selection trends in Fig. 4(a), together with the evolution of the far-infrared luminosity function explains the lower redshift distribution of the 450-μm sources in comparison to 850-μm selected sources (e.g. Béthermin et al. 2015).
• In agreement with the photometric properties of the samples (see Section 3.1 and Fig 2), their dust mass results in Fig. 4(b), suggest that 450-μm selection is sensitive to lower dust mass sources at lower redshifts.
• The 450-μm sample has a systematically higher characteristic dust temperature than the 850-μm sample.

3.3.2 Optical/near-infrared properties

• The rest-frame UV/optical/near-infrared features in the SED are dominated by the stellar emission, thus physical properties such as stellar mass and dust attenuation can be inferred.
• The median dust attenuation is significantly lower than the AS2UDS value of AV = 2.89 ± 0.04 mag, as also suggested from the comparison of the rest-frame UV slopes in Fig. 5(a).
• Again, the authors highlight the reliability of the sections of the SED for each sample with lines of variable thickness.

4 D ISCUSSION

• So far, the authors have investigated the physical properties of the full SNR ≥ 5 450-μm-selected sample and compared these to those selected at 850μm.
• As seen in Fig. 4, selection at different wavelengths (in populations whose space density peaks at different redshifts, Fig. 3b) leads to a range of potential selection effects.
• The authors note that the median rest-frame wavelength for the samples at z = 1–2 and z = 3–4 differs by ∼5 per cent, but they confirm that precisely matching the redshift distributions to achieve perfect agreement in their median wavelengths does not change their results.
• With both samples selected at the same rest-frame wavelength, λrest ∼ 180μm, and occupying the same parameter space in dust mass (roughly equating to sub-millimetre flux limit), the authors examine whether there are any physical differences between identical far-infrared-selected galaxies as the age of the Universe doubled between z ∼ 3.5 and z ∼ 1.5.

4.1 Comparing rest-frame-selected populations

• First, the authors look at the overall properties of their z ∼ 1.5 and z ∼ 3.5 samples by investigating their composite SEDs in Fig. 5(b).
• The uncertainty is then estimated by taking the 16th and 84th percentile values at each wavelength.
• Overall, the authors conclude that the z ∼ 1.5 sources have properties lying between those of the local templates of Arp 220 and M82, while the z ∼ 3.5 sources are more extreme than Arp 220 in terms of their low rest-frame optical to far-infrared luminosity ratios.
• Thus, the brighter optical SED of the z ∼ 1.5 sample arises from the combination of both slightly higher stellar mass and lower dust attenuation.

4.1.1 Gas fraction and star-formation efficiency

• The authors analysis suggests that at z ∼ 1.5 far-infrared selected galaxies are different to those at z ∼ 3.5, in both the far-infrared and optical regimes, even when selected at the same rest-frame wavelength and the same dust mass limit.
• The authors explore two approaches to determine the appropriate value for δgdr.
• The results indicate that the star-formation is slower at z ∼ 1.5, while at z ∼ 3.5 the more gas-rich galaxies are forming stars more rapidly, and so consuming the larger gas reservoirs in a comparable amount of time.
• D20 showed that the 850-μm SMGs are broadly consistent with this homologous and homogeneous population model of centrally illuminated dust clouds, with the dust continuum size of SMGs broadly following the expected trend with far-infrared luminosity-to-gas mass ratio.
• The authors note that the median dust mass for their z ∼ 3.5 sample is, on average, two times higher than for the z ∼ 1.5 population.

4.2 Dust properties of far-infrared-selected galaxies

• Dust, while a small component of the overall baryonic mass of a galaxy, is a useful tracer of the ISM.
• (a) Gas fraction as a function of redshift.
• The binned median values are shown as large circles, where the authors split the larger z ∼ 3.5 sample into three independent bins of dust mass, and the errors are derived by a bootstrap method.

4.2.1 Dust mass function

• The authors calculate the dust mass function for the z ∼ 1.5 sources using an accessible volume method: φ(Md) Md = (1/Vi), where φ(M) M is the number density of sources with dust masses between M and M+ M and Vi is the co-moving volume within which the i-th source would be detected in a given dust mass bin.
• The difference in the shape of their two dust mass functions indicates that the characteristic dust mass of the high-redshift sources is higher than that of the low-redshift sample.

4.2.2 Dust mass density

• Given the apparently different shapes of the dust mass function in their two λrest ∼ 180-μm-matched samples, the authors opt to assess the evolutionary differences in the dust properties of galaxy populations using integrated dust mass density as the most robust measurement available.
• It does not follow the observational results from other studies at z 1.
• To investigate how the physical properties of infrared luminous galaxies evolve with redshift, the authors also select z = 1–2 450-μm sources and z = 3–4 850-μm sources both with Md ≥ 2 × 108 M , to construct rest-frame wavelength (λrest ∼ 180μm) matched subsets.
• UD acknowledges the support of Science and Technology Facilities Council (STFC) studentship (ST/R504725/1).

Full text

Durham Research Online
Deposited in DRO:
17 November 2020
Version of attached le:
Published Version
Peer-review status of attached le:
Peer-reviewed
Citation for published item:
Dudzevicit, U and Smail, Ian and Swinbank, A M and Lim, C-F and Wang, W-H and Simpson, J M and Ao,
Y and Chapman, S C and Chen, C-C and Clements, D and Dannerbauer, H and Ho, L C and Hwang, H S and
Koprowski, M and Lee, C-H and Scott, D and Shim, H and Shirley, R and Toba, Y (2021) 'Tracing the
evolution of dust-obscured activity using sub-millimetre galaxy populations from STUDIES and AS2UDS.',
Monthly notices of the Royal Astronomical Society., 500 (1). pp. 942-961.
Further information on publisher's website:
https://doi.org/10.1093/mnras/staa3285
Publisher's copyright statement:
This article has been accepted for publication in Monthly notices of the Royal Astronomical Society.
c

: 2020 The
Author(s). Published by Oxford University Press on behalf of the Royal Astronomical Society. All rights reserved.
Additional information:
Use policy
The full-text may be used and/or reproduced, and given to third parties in any format or medium, without prior permission or charge, for
personal research or study, educational, or not-for-prot purposes provided that:
•
a full bibliographic reference is made to the original source
•
a link is made to the metadata record in DRO
•
the full-text is not changed in any way
The full-text must not be sold in any format or medium without the formal permission of the copyright holders.
Please consult the full DRO policy for further details.
Durham University Library, Stockton Road, Durham DH1 3LY, United Kingdom
Tel : +44 (0)191 334 3042 | Fax : +44 (0)191 334 2971
https://dro.dur.ac.uk

MNRAS 500, 942–961 (2021) doi:10.1093/mnras/staa3285
Advance Access publication 2020 October 24
Tracing the evolution of dust-obscured activity using sub-millimetre
galaxy populations from STUDIES and AS2UDS
U. Dudzevi
ˇ
ci
¯
ut
˙
e ,
1‹
Ian Smail ,
1
A. M. Swinbank ,
1
C.-F. Lim,
2,3
W.-H. Wang,
3
J. M. Simpson,
1
Y. Ao,
4
S. C. Chapman,
5,6,7
C.-C. Chen ,
3,8
D. Clements ,
9
H. Dannerbauer,
10,11
L. C. Ho,
12,13
H. S. Hwang ,
14
M. Koprowski,
15
C.-H. Lee,
16
D. Scott,
5
H. Shim,
17
R. Shirley
10,11
and Y. Toba
3,18,19
1
Centre for Extragalactic Astronomy, Department of Physics, Durham University, South Road, Durham DH1 3LE, UK
2
Graduate Institute of Astrophysics, National Taiwan University, Taipei 10617, Taiwan
3
Academia Sinica Institute of Astronomy and Astrophysics (ASIAA), No. 1, Section 4, Roosevelt Rd., Taipei 10617, Taiwan
4
Purple Mountain Observatory, Chinese Academy of Sciences, Nanjing 210033, People’s Republic of China
5
Department of Physics and Astronomy, University of British Columbia, 6225 Agricultural Road, Vancouver, BC V6T 1Z1, Canada
6
National Research Council, Herzberg Astronomy and Astrophysics, 5071 West Saanich Road, Victoria, BC V9E 2E7, Canada
7
Department of Physics and Atmospheric Science, Dalhousie University, Halifax, NS B3H 4R2, Canada
8
European Southern Observatory, Karl Schwarzschild Strasse 2, D-85748 Garching, Germany
9
Blackett Lab, Imperial College, London, Prince Consort Road, London SW7 2AZ, UK
10
Instituto de Astrof
´
ısica de Canarias, E-38205 La Laguna, Tenerife, Spain
11
Universidad de La Laguna, Departamento de Astrof
´
ısica, E-38206 La Laguna, Tenerife, Spain
12
Kavli Institute for Astronomy and Astrophysics, Peking University, Beijing 100871, People’s Republic of China
13
Department of Astronomy, School of Physics, Peking University, Beijing 100871, People’s Republic of China
14
Korea Astronomy and Space Science Institute, 776 Daedeokdae-ro, Yuseong-gu, Daejeon 34055, Republic of Korea
15
Institute of Astronomy, Faculty of Physics, Astronomy and Informatics, Nicolaus Copernicus University, Grudziadzka 5, PL-87-100 Torun, Poland
16
NSF’s National Optical-Infrared Astronomy Research Laboratory, 950 North Cherry Avenue, Tucson, AZ 85719, USA
17
Department of Earth Science Education, Kyungpook National University, Daegu 41566, Republic of Korea
18
Department of Astronomy, Kyoto University, Kitashirakawa-Oiwake-cho, Sakyo-ku, Kyoto 606-8502, Japan
19
Research Center for Space and Cosmic Evolution, Ehime University, 2-5 Bunkyo-cho, Matsuyama, Ehime 790-8577, Japan
Accepted 2020 October 12. Received 2020 October 12; in original form 2020 July 31
ABSTRACT
We analyse the physical properties of 121 SNR ≥ 5 sub-millimetre galaxies (SMGs) from the STUDIES 450 μm survey. We
model their UV-to-radio spectral energy distributions using
MAGPHYS+photo-z and compare the results to similar modelling of
850 μm-selected SMG sample from AS2UDS, to understand the fundamental physical differences between the two populations
at the observed depths. The redshift distribution of the 450-μm sample has a median of z = 1.85 ± 0.12 and can be described by
strong evolution of the far-infrared luminosity function. The fainter 450-μm sample has ∼14 times higher space density than the
brighter 850-μm sample at z  2, and a comparable space density at z = 2–3, before rapidly declining, suggesting LIRGs are the
main obscured population at z ∼ 1–2, while ULIRGs dominate at higher redshifts. We construct rest-frame ∼180-μm-selected
and dust-mass-matched samples at z = 1–2 and z = 3–4 from the 450 and 850-μm samples, respectively, to probe the evolution
of a uniform sample of galaxies spanning the cosmic noon era. Using far-infrared luminosity, dust masses, and an optically thick
dust model, we suggest that higher redshift sources have higher dust densities due to inferred dust continuum sizes which are
roughly half of those for the lower redshift population at a given dust mass, leading to higher dust attenuation. We track the
evolution in the cosmic dust mass density and suggest that the dust content of galaxies is governed by a combination of both the
variation of gas content and dust destruction time-scale.
Key words: galaxies: evolution – galaxies: starburst – infrared: galaxies.
1 INTRODUCTION
The discovery of a significant energy density in the extragalactic
background of the Universe at wavelengths ≥150 μm (Puget et al.

E-mail: ugne.dudzeviciute2@durham.ac.uk
1996; Fixsen et al. 1998), suggested the existence of a population of
dust-enshrouded galaxies that is much more significant than their
local analogues (Low & Kleinmann 1968;Rieke&Low1972;
Neugebauer et al. 1984; see Hauser & Dwek 2001 and Lagache,
Puget & Dole 2005 for reviews). In such far-infrared luminous
galaxies, the radiation from young massive stars is absorbed by dust
grains in their interstellar medium (ISM) and re-radiated as thermal
C

2020 The Author(s)
Published by Oxford University Press on behalf of the Royal Astronomical Society
Downloaded from https://academic.oup.com/mnras/article/500/1/942/5937488 by University of Durham user on 17 November 2020

Tracing the dust evolution using FIR galaxies 943
continuum emission at far-infrared wavelengths (for reviews, see
Casey, Narayanan & Cooray 2014;Lutz2014). In the mid-1990s, the
first surveys at sub-millimetre wavelengths (450 and 850 μm) using
the Sub-millimeter Common User Bolometric Array (SCUBA) on
James Clerk Maxwell Telescope (JCMT) began to resolve this far-
infrared/sub-millimetre background into its constituent galaxies and
identified the first statistical samples of high-redshift, sub-millimetre
bright galaxies (SMGs – Smail, Ivison & Blain 1997;Bargeretal.
1998; Hughes et al. 1998; Eales et al. 1999). These surveys confirmed
the cosmological significance of far-infrared-luminous galaxies,
in particular their potentially significant contribution to the star-
formation rate density at high redshifts (see Madau & Dickinson
2014).
Due to atmospheric transmission, large-scale surveys of the high-
redshift SMG populations are undertaken primarily in wavebands
around 850 μm and 1.2 mm (e.g. Coppin et al. 2006; Scott et al.
2008;Weißetal.2009; Hatsukade et al. 2011; Mocanu et al. 2013;
Umehata et al. 2014; Geach et al. 2017; Miettinen et al. 2017;Cowie
et al. 2018; Stach et al. 2019). These wavebands typically select
galaxies based on their luminosity at rest-frame wavelengths around
300 μm and are thus sensitive to the cool dust mass of the galaxies
(Dudzevi
ˇ
ci
¯
ut
˙
eetal.2020, hereafter D20). Subsequent studies of
this population have suggested that these systems are strongly dust
obscured systems with high far-infrared luminosities and lying at
high redshifts, with a number density peaking at z ∼ 2–3 (Chapman
et al. 2005; da Cunha et al. 2015;Koprowskietal.2016; Brisbin et al.
2017; Danielson et al. 2017; D20). SMGs have huge gas reservoirs
of the order of 10
10–11
M

and star-formation rates ranging over
100–1000 M

yr
−1
, meaning that they have the capacity to rapidly
increase their already high stellar masses (∼10
11
M

) on a short
time-scale (Ivison et al. 2011;Bothwelletal.2013, D20). Although
these studies have provided insight into the physical properties of
the SMGs, observations at such long wavelengths do not sample the
peak of the far-infrared emission and, as noted, are more sensitive
to sources with larger masses of cool dust, as well as those at higher
redshifts (Blain & Longair 1993).
At high redshift (z>2), selection at shorter sub-millimetre and
far-infrared wavelengths (e.g. ≤ 500 μm) samples the spectral energy
distribution (SED) of the dust continuum emission closer to the peak
of the far-infrared emission, ratherthan the Rayleigh–Jeans tail which
is traced by the 850 μm and 1.2 mm surveys. Therefore, surveys at
shorter wavelengths are more sensitive to far-infrared luminosity,
than the cold dust mass. This can potentially lead to surveys detecting
physically different sources when selected at differentwavebands and
redshifts.
Surveys at far-infrared wavelengths, specifically 250, 350, and
500 μm, using the SPIRE instrument on Herschel have mapped
hundreds of square degrees of sky (Eales et al. 2010; Oliver et al.
2012;Wangetal.2014; Valiante et al. 2016). While covering huge
areas, these surveys are limited in sensitivity due to the large beam
size and resulting bright confusion limit,
1
which makes it challenging
to detect all but the brightest (unlensed) sources at z  1 (Symeonidis,
Page & Seymour 2011, although see Shu et al. 2016; Jin et al. 2018;
Liu et al. 2018). Moreover, the large beam size makes it difficult
to reliably locate counterparts needed to understand their properties.
However, higher resolution imaging can be obtained from single-
dish telescopes on the ground through the atmospheric windows
at 350 μm (Khan et al. 2007; Coppin et al. 2008) and 450 μm
1
Defined as the sensitivity limit arising from unresolved sources which cannot
be improved by increasing the integration time.
(Blain et al. 1999;Chenetal.2013). Unfortunately the atmospheric
transmission at 450 μm is around half of that at 850 μm, and
obtaining deep, large-area surveys with ground-based observations
is therefore challenging. Although Atacama Large Millimeter Array
(ALMA) could, in principle, produce deep, high-resolution imaging
at 450 μm, large surveys would be observationally expensive due to
the very limited field of view of ∼0.02 arcmin
2
at this wavelength. In
contrast, the SCUBA-2 camera (Holland et al. 2013) on the JCMT,
with a field of view of 45 arcmin
2
and a beam size of 7.9 arcsec at
450 μm (yielding an approximately 20 times lower confusion limit
than SPIRE at 500 μm), provides the sensitivity, mapping speed, and
angular resolution necessary to identify 450-μm sources and their
counterparts over fields of 100s arcmin
2
.
Studies of sources selected at 450 μm, closer to the peak of the
far-infrared emission in systems at the epoch of peak star formation
(z ∼ 2–3), have suggested that the 450-μm-selected population at
mJy flux limits lies at lower redshift than those selected at 850 μm,
with a distribution that peaks at z ∼ 1.5–2.0 (Casey et al. 2013; Geach
et al. 2013; Roseboom et al. 2013; Zavala, Aretxaga & Hughes 2014;
Bourne et al. 2017;Zavalaetal.2017; Lim et al. 2020). 450-μm-
selected sources are also suggested to have higher characteristic dust
temperatures than 850-μm-selected SMGs by T
d
 10 K (Casey
et al. 2013; Roseboom et al. 2013), although this may be partly due
to selection effects. However, the identification of differences in the
physical properties of SMGs selected in the different sub-millimetre
wavebands from such studies have been limited by their modest
sample sizes and also their biases towards brighter sources due to the
sensitivity limits at 450 μm. Comparison to the 850-μm population
is further complicated by uncertain or incomplete identifications in
the longer wavelength samples (e.g. Hodge et al. 2013), as well as the
use of different photometric redshift and SED modelling methods on
different samples.
This study aims to better understand the physical properties of
SMGs and in particular the relationship between samples selected
at 450 and 850 μm, by exploiting a very deep 450-μm imaging
survey: the SCUBA-2 Ultra Deep Imaging EAO Survey (STUDIES;
Wang et al. 2017; Lim et al. 2020) in the Cosmic Evolution Survey
(COSMOS; Scoville et al. 2007) field. STUDIES is a multi-year
JCMT survey within the CANDELS region (∼300 arcmin
2
), which
obtained the deepest single-dish map at 450 μm currently available,
with a 1-σ depth of 0.65 mJy. The source catalogue and physical
properties of the 450-μm sample are presented in Lim et al. (2020),
while the structural parameters and morphological properties have
been analysed by Chang et al. (2018).
In this paper, we compare the properties of galaxies selected
from deep 450-μm observations to those selected from typical 850-
μm surveys. SCUBA-2 simultaneously maps at 450 and 850 μm,
however, the STUDIES 850-μm map is confusion limited at an 850-
μm flux limit of ∼2 mJy and it cannot be used to reliably identify faint
850-μm sources. Therefore, for our 850-μm comparison sample we
utilize the largest available ALMA-identified 850-μm-selected SMG
sample, from the ALMA/SCUBA-2 Ultra Deep Survey (AS2UDS;
Stach et al. 2019, D20).
We revisit the modelling of the UV-to-radio SEDs of the 450-
μm galaxies from STUDIES using
MAGPHYS+photo-z (da Cunha,
Charlot & Elbaz 2008; da Cunha et al. 2015; Battisti et al. 2019),
which was employed by D20 on the AS2UDS 850-μm sample, thus
ensuring that the comparison of the physical properties between the
two samples is free from systematic differences due to the modelling.
With these two large, consistently analysed samples we empirically
compare the physical properties of 450-μm-detected galaxies with
the 850-μm population. At the observed depths, the two surveys
MNRAS 500, 942–961 (2021)
Downloaded from https://academic.oup.com/mnras/article/500/1/942/5937488 by University of Durham user on 17 November 2020

944 U. Dudzevi
ˇ
ci
¯
ut
˙
eetal.
sample down to the ULIRG/LIRG limit (10
11–12
L

)atz = 1–3 and
although there is some overlap between these flux-limited samples,
as discussed in Lim et al. (2020), the combination and comparison of
450 and 850-μm surveys provide a more complete view of luminous
far-infrared activity in the Universe over a wider redshift range than
possible with either individual sample. In particular, we exploit these
large samples to construct subsets that are matched in rest-frame
wavelength to allow us to quantify the physical differences between
an identically selected sample of dusty galaxies at z ∼ 1.5 and z ∼
3.5.
The paper is structured as follows. In Section 2, we give details on
the multi-wavelength data which we use to construct the SEDs for
our sources and describe the SED fitting procedure. In Section 3, we
present our results, including a comparison of the STUDIES 450-μm
selected galaxies to the 850-μm selected galaxies from the AS2UDS
survey. We discuss the implications of our results in Section 4 and
present our conclusions in Section 5. We adopt a CDM cosmology
with with H
0
= 70 km s
−1
Mpc
−1
, 

= 0.7, 
m
= 0.3. When
quoting magnitudes, we use the AB photometric system.
2 OBSERVATIONS AND SED FITTING
2.1 Photometric coverage
STUDIES is a SCUBA-2 450-μm imaging survey within the CAN-
DELS region in the COSMOS field. A detailed description of the
SCUBA-2 observations and data reduction can be found in Wang
et al. (2017) and Lim et al. (2020). Briefly, the data from STUDIES,
combined with archival data taken by Geach et al. (2013)andCasey
et al. (2013), yields the deepest single-dish map currently available at
450 μm, reaching a 1-σ noise level of 0.65 mJy. This survey detects
256 sources with a signal-to-noise ratio (SNR) of SNR ≥ 4 (of which
126 have SNR ≥ 5) in an area of 300 arcmin
2
. The confusion-limited
850-μm map reaches an instrumental noise level of 0.1 mJy in the
deepest regions and has an estimated confusion noise of 0.7 mJy
(Lim et al. 2020). The 850-μm flux densities of the 450-μm-selected
STUDIES sources were obtained from the 850-μm map at the 450-
μm positions. The source is classed as detected at 850 μmiftheflux
density has SNR ≥ 5, otherwise it is treated as a limit.
In this section, we provide a brief description of the counterpart
identification and multi-wavelength photometric data available for
the sample from UV to radio wavelengths, which is then used
to model the SEDs of the sources. For a full description of the
photometric data and counterpart identification for the STUDIES
450-μm sample, please refer to Lim et al. (2020).
2.1.1 Counterpart identification
The identification of optical counterparts for the STUDIES 450-
μm sources is described in Lim et al. (2020). Briefly, the 450-μm
sources were matched with the VLA-COSMOS 3-GHz catalogue
(Smol
ˇ
ci
´
cetal.2017) using a 4 arcsec search radius (set by the
JCMT 450-μm beam) yielding ∼1 per cent false positive rate (based
on the probability of false matches using the number densities
of both catalogues). For the 450-μm sources above SNR≥ 5this
yielded 89/126 counterparts (and 134/256 for SNR ≥ 4). These radio
counterparts were cross-matched with the Spitzer IRAC catalogue
(Sanders et al. 2007) using a 1 arcsec search radius with a ∼3
per cent false positive rate. For the 450-μm sources that did not
have 3-GHz radio counterparts, these were cross-matched with the
Spitzer MIPS 24-μm catalogue (Sanders et al. 2007) with a search
radius of 4 arcsec resulting in 27/37 matches for 450-μm sources
with SNR ≥ 5 (and 76/122 for SNR ≥ 4). These MIPS counterparts
were in turn used to find IRAC counterparts within 2 arcsec with a
∼2 per cent false positive rate. The identification rates in different
ancillary bands are presented in Lim et al. (2020). The remaining ten
SNR ≥ 5 450-μm sources with no radio or MIPS counterparts (and
the 46 with SNR ≥ 4) were matched using a 1 arcsec matching radius
to the catalogue of colour/radio-selected candidate sub-millimetre
counterparts from An et al. (2019). This catalogue was constructed
using a radio+machine-learning method applied to a training set
comprising ALMA-identified 870-μmSMGsintheCOSMOSand
UDS fields with the goal of identifying multi-wavelength counter-
parts of S2COSMOS 850-μm single-dish detected sub-millimetre
sources. This produced five identifications for the SNR ≥ 5 sources
(and 15 at SNR ≥ 4).
For the SNR ≥ 5 450-μm sample this process yields reliable
counterparts for 121/126 (96 per cent) of the sources, which declines
to 207/256 (81 per cent) for those with SNR ≥ 4. As we wish to have a
highly complete and hence unbiased sample, in this paper we analyse
the SNR ≥ 5 450-μm sources, equivalent to S
450
≥ 3.25 mJy, which
have almost complete identifications. For this sample of 126 sources:
89 are located through radio counterparts, 27 are identified through
MIPS counterparts and from the remaining ten sources, five have
counterparts derived from the machine-learning method. In total 109
of the counterparts have IRAC detections. Although some studies
have shown that mis-identifications are possible due to the large
beam sizes (∼15–20 arcsec full width at half-maximum, FWHM) of
single-dish telescopes at long wavelengths (e.g. Hodge et al. 2013),
this is much less of an issue for the 450-μm beam (7.9 arcsec FWHM)
and is further reduced by the SNR ≥ 5 cut as the centroid position is
more precise ( 1 arcsec).
For our sample, we find that, on average, the sources are detected in
19 bands (16–84th percentile range of N
det
= 13–22). The detection
rate is 70/105 in B-band, 87/87 in z-band, 115/116 in H-band,
107/109 at 4.5 μm, 91/121 at 250 μm, and 43/121 at 850 μm.
2.1.2 Far-infrared to radio observations
To constrain the SED of each galaxy at radio wavelengths we utilize
1.4 and 3-GHz data from the Very Large Array (VLA)-COSMOS
Large Project (Schinnerer et al. 2010; Smol
ˇ
ci
´
cetal.2017). The 3-
GHz survey has a noise of 2.3 μJy beam
−1
and an angular resolution
of 0.7 arcsec. The 1.4 GHz data is compiled in the COSMOS2015
catalogue (Laigle et al. 2016) and covers the entire COSMOS
field with σ = 12 μJy beam
−1
with an angular resolution of 2.5
arcsec.
At z  2, the far-infrared SED of a source with a characteristic
temperature of T
d
∼30 K is expected to peak at an observed
wavelength of ∼300 μm. Hence, to better constrain the shape of the
far-infrared SEDs for the galaxies in our sample, and so improve the
constraints on the far-infrared luminosities, we include observations
with the Spectral and Photometric Imaging Receiver (SPIRE: Griffin
et al. 2010) and the Photodetector Array Camera and Spectrometer
(PACS: Poglitsch et al. 2010)ontheHerschel Space Observatory.
We specifically make use of the 100 and 160 μm PACS (Lutz et al.
2011), and 250 and 350 μm SPIRE observations taken as part of the
Herschel Multi-tiered Extragalactic Survey (HerMES; Oliver et al.
2012). We adopt the PACS 100- and 160-μm flux densities from
Lutz et al. (2011) (as listed in the the COSMOS2015 catalogue),
who presented the observations of the 2 deg
2
COSMOS field which
reach a 3-σ depth of 10.2 mJy at 160 μm.
MNRAS 500, 942–961 (2021)
Downloaded from https://academic.oup.com/mnras/article/500/1/942/5937488 by University of Durham user on 17 November 2020

Tracing the dust evolution using FIR galaxies 945
Due to the coarse resolution of the SPIRE maps (∼18 and
∼25 arcsec FWHM at 250 and 350 μm, respectively), we use the
method described in Swinbank et al. (2014) to deblend these maps
and obtain reliable flux densities for our catalogue. The deblending
of the SPIRE maps used positional priors for sources based on the
3-GHz and 24-μm (see below) catalogues, as well as machine-
learning identified SMG counterparts from An et al. (2019)(see
Section 2.1.1). The observed flux density distribution is fitted with
beam-sized components at the position of a given source in the prior
catalogue using a Monte Carlo algorithm. The method is first applied
to the 250-μm data, then to avoid ‘over-blending’ only sources that
are detected at > 2-σ at 250 μm are propagated to the prior list for
the 350-μm deblending. The uncertainties on the flux densities (and
limits) are found by attempting to recover model sources injected into
the maps (see Swinbank et al. 2014 for details), and yield typical 3-σ
detection limits of 7.0 and 8.0 mJy at 250 and 350 μm, respectively.
2
2.1.3 Optical to near-/mid-infrared observations
To model the stellar SEDs of the counterparts to our 450-μm
sources, we require the photometry in the optical/infrared bands. For
the u
∗
Bgrizy bands we adopt the photometry from COSMOS2015
(Laigle et al. 2016) catalogue. The u
∗
-band data are from the Canada–
France–Hawaii Telescope (CFHT/MegaCam; Boulade et al. 2003)
and covers the entire COSMOS field with a 5-σ depth of u
∗
 26.5
(Ilbert et al. 2009). The B-band imaging was taken with Subaru
Suprime-Cam as part of COSMOS-20 survey (Taniguchi et al. 2007)
and has a 5-σ depth of B = 27.2 in a 2 arcsec diameter aperture.
Images in the grizy-bands are taken from the second data release
(DR2) of the Hyper-SuprimeCam (HSC) Subaru Strategic Program
(SSP; Aihara et al. 2019). The nominal 5-σ depths are g = 27.3, r =
26.9, i = 26.7, z = 26.3, and y = 25.3 in 2 arcsec diameter apertures.
In addition we employ YJHK
s
imaging from the fourth data release
(DR4) of the UltraVISTA survey (McCracken et al. 2012). In an
equivalent manner to Simpson et al. (2020), we measure 2 arcsec
diameter aperture photometry at the positions of each SMG in each
band. The uncertainty on the derived flux densities is estimated
in a 1 × 1arcmin
2
region centred on the position of each SMG.
Finally, we convert the derived flux densities to a total flux density
by applying an aperture correction of a factor of 1.80, 1.74, 1.52, and
1.46 for the YJHK
s
bands, respectively. This is done by comparing
the DR4 photometry to the UltraVISTA DR2 photometry from
COSMOS2015, for those SMGs with a counterpart in the catalogue.
For the near-infrared photometry, we employ the Spitzer IRAC
data from Sanders et al. (2007). IRAC 3.6-, 4.5-, 5.8-, and 8.0-μm
imaging was obtained as part of the S-COSMOS survey, which covers
the entire COSMOS field and has an angular resolution of 1.7 arcsec
at 3.6 μm. The 24-μm catalogue was generated by Lim et al. (2020),
whousedtheS-COSMOS24μm image (Sanders et al. 2007). The
catalogue has a 3.5-σ limit of 57 μJy.
We correct the u
∗
-band to IRAC 8.0-μm photometry of each
source for Galactic extinction based on its sky position, the extinction
maps of Schlafly & Finkbeiner (2011), and the extinction curve of
Fitzpatrick (1999), assuming a reddening law with R
V
= 3.1. For each
filter, the correction is determined by convolving the filter response
with the scaled extinction curve.
2
Comparison of our measurements to the deblended Herschel sources from
Jin et al. (2018) showed agreement within the quoted errors with abs(S–
S
Jin + 2018
)/S
err
of 1.25 and 1.15 at 250 and 350 μm, respectively.
2.2 SED fitting model
To derive the physical properties of the STUDIES SMGs, we employ
MAGPHYS+photo-z (da Cunha et al. 2015; Battisti et al. 2019)to
model the SEDs from optical to radio wavelengths, using the avail-
able photometry in 24 bands. The attenuation of the stellar emission
by dust in the UV/optical and near-infrared and the consequent
re-radiation in the far-infrared is coupled via an energy balance
technique. This model allows us to constrain the physical parameters
of the galaxies, as well as providing a consistent methodology to that
applied to the large ALMA-identified 850-μm SMG sample from
the AS2UDS survey by D20. Thus, the physical properties of the
two samples can be investigated for any differences arising from the
different wavelength selection.
MAGPHYS+photo-z uses stellar population models from Bruzual &
Charlot (2003) and a Chabrier (2003) IMF. Star-formation histories
(SFH) are modelled as continuous delayed exponential functions
(Lee et al. 2010) with superimposed bursts. Dust attenuation is
modelled by a two-component model of Charlot & Fall (2000),
combining the effective attenuation from dust in stellar birth clouds
and diffuse interstellar medium, parametrised by the reddening in the
V-band. The star-formation rate is calculated using the best-fitting
model star-formation history, after accounting for dust attenuation
and is defined as the average of the SFH over the last 10 Myr. The
far-infrared emission from dust in
MAGPHYS+photo-z is determined
self-consistently from the dust-attenuated stellar emission. The far-
infrared luminosity is measured by integrating the SED in the rest-
frame between 8 and 1000 μm and is calculated through the sum of
the birth cloud and ISM luminosities, including contributions from
the polycyclic aromatic hydrocarbons, and mid-infrared continuum
from hot, warm, and cold dust in thermal equilibrium. The dust
mass is calculated fitting a two-component modified blackbody with
emissivity index, β, fixed at 1.5 for the warm components and 2.0 for
the cold components. We note that different assumptions in model
emissivity index, dust opacity, and dust mass absorption coefficient
will impact the dust mass measurements, therefore care must be taken
when interpreting and comparing results (see Casey et al. 2014).
MAGPHYS+photo-z estimates a characteristic dust temperature using
five free parameters that combine the contribution from the warm
(birth clouds) and cold (diffuse ISM) components. A full description
of
MAGPHYS+photo-z code and parameter derivation can be found
in da Cunha et al. (2015) and Battisti et al. (2019).
To fit the observed multi-wavelength photometry of each galaxy,
for each star-formation history
MAGPHYS+photo-z creates a library
of SEDs at random redshifts, resulting in ∼300 000 templates. The
best-fitting SED is selected using a χ
2
test returning the photometric
redshift, best-fitting parameters, and their probability distributions.
The uncertainties for each parameter are taken as 16–84th percentile
of the probability distribution. For our analysis, we used the updated
MAGPHYS+photo-z code from da Cunha et al. (2015) and Battisti
et al. (2019), which is optimised for high redshift star-forming
galaxies with extended prior distributions for dust optical depths and
star-formation histories. The code also includes a parametrization to
reproduce the intergalactic medium absorption of UV photons.
We note that the far-infrared SEDs of our SMGs are covered by
at most six photometric bands (see Fig. 1a), with weaker constraints
near the peak of the SEDs, hence to provide a robust estimate of
dust temperature, we adopt a conservative approach and fit a single
modified blackbody to the available Herschel 100-, 160-, 250-, and
350-μm photometry and the SCUBA-2 flux densities at 450 and
850 μm. This approach allows for a simple comparison with similar
fits to other 450, 850-μm, and ALMA samples. We estimate the
MNRAS 500, 942–961 (2021)
Downloaded from https://academic.oup.com/mnras/article/500/1/942/5937488 by University of Durham user on 17 November 2020

Frequently asked questions

Q1. What are the contributions mentioned in the paper "Tracing the evolution of dust-obscured activity using sub-millimetre galaxy populations from studies and as2uds" ?

In this paper, the physical properties of 121 SNR ≥ 5 sub-millimetre galaxies ( SMGs ) from the STUDIES 450μm survey were analyzed using MAGPHYS+photo-z and compared to similar modelling of 850μm-selected SMG sample from AS2UDS. 

Q2. How do the authors estimate the gas-to-dust ratios for the z ?

In addition, as gas-to-dust mass ratio is expected to vary with stellar mass and redshift, the authors can also estimate the expected dust-to-gas ratios for the median stellar mass at the median redshift of each sample. 

Q3. How much is the systematic offset between the dust masses retrieved using these two different methods?

The systematic offset between the dust masses retrieved using these two different methods is 10 per cent, which is within the uncertainty of the MAGPHYS+photo-z dust mass values. 

Q4. What is the flux density distribution of the sources in the prior catalogue?

The observed flux density distribution is fitted with beam-sized components at the position of a given source in the prior catalogue using a Monte Carlo algorithm. 

Q5. How many uncertainties were calculated on the dust mass of the two samples?

The uncertainties on the dust functions of both samples were calculated by resampling the dust mass and redshift probability distributions to construct multiple dust mass functions. 

Q6. What is the effect of the dust opacity assumptions?

the authors stress that it is likely that the dust emission from the sources in their sample is not optically thin in the far-infrared and T MBBd and Md (through strong dependence on T MBBd ) are affected by the dust opacity assumptions. 

Q7. How do the authors determine the median gas mass fraction of fgas?

The authors derive a median gas mass fraction of fgas = 0.19 ± 0.06 with a 68th percentile range of fgas = 0.10–0.58, assuming a gas-to-dust ratio of 100. 

Q8. How do the authors calculate the dust mass density for the rest 180m sample?

The authors calculate the total dust mass density for the λrest ∼ 180- μm-matched samples at z ∼ 1.5 and z ∼ 3.5 by combining the dust mass density estimates extrapolated down to Md ∼104 M from the best-fitting Schechter function for their respective dust mass functions. 

Q9. What is the typical offset for any given physical property?

The authors find that the typical systematic offset for any given physical property (SFR, LFIR, M∗, Md, AV) is ∼10 per cent, which is within the typical uncertainties. 

Q10. What is the flux density of the AS2UDS SMGs?

As seen in Fig. 3(b), the space density for the S850 ≥ 3.6 mJy subset of the AS2UDS SMGs is significantly lower than the 450- μm population, however, this is primarily due to the different flux density and luminosity limits of the two studies. 

Q11. What is the median stellar mass of the 850-m sample?

The median stellar mass of the 850-μm sample is M∗ = (1.26 ± 0.05) × 1011 M , similar to the 450-μm population (but typically seen at an earlier epoch). 

Q12. How do the authors derive the dust mass density for a sample with Md 2.0?

The authors derive a dust mass density of ρ = (2.6 ± 0.5 ) × 104 M Mpc−3 at z ∼ 1.5 and ρ = (2.41 ± 0.13 ) × 104 M Mpc−3 at z ∼ 3.5, for a sample with Md ≥ 2.0× 108 M .