MNRAS 000, 1–5 (2022) Preprint 21 September 2022 Compiled using MNRAS LATEX style file v3.0
Ammonia in the interstellar medium of a starbursting disc at z = 2.6
M. J. Doherty,1★ J. E. Geach,1 R. J. Ivison,2 K. M. Menten,3 A. M. Jacob,4 J. Forbrich,1 & S. Dye5
1Department of Physics, Astronomy & Mathematics, University of Hertfordshire, College Lane, Hatfield, AL10 9AB
2European Southern Observatory, Karl-Schwarzschild-Straße 2, D-85748 Garching, Germany
3Max Planck Institute for Radio Astronomy, Auf dem Hügel 69, 53121 Bonn, Germany
4William H. Miller III Department of Physics & Astronomy, Johns Hopkins University, Baltimore, MD 21218, USA
5School of Physics and Astronomy, University of Nottingham, University Park, Nottingham NG7 2RD, UK
21 September 2022
ABSTRACT
We report the detection of the ground state rotational emission of ammonia, ortho-NH3 (𝐽𝐾 = 10 → 00) in a gravitationally
lensed, intrinsically hyperluminous, star-bursting galaxy at 𝑧 = 2.6. The integrated line profile is consistent with other molecular
and atomic emission lines which have resolved kinematics well-modelled by a 5 kpc-diameter rotating disc. This implies that
the gas responsible for NH3 emission is broadly tracing the global molecular reservoir, but likely distributed in pockets of high
density (𝑛 & 5 × 104 cm−3). With a luminosity of 2.8 × 106 𝐿, the NH3 emission represents 2.5 × 10−7 of the total infrared
luminosity of the galaxy, comparable to the ratio observed in the Kleinmann-Low nebula in Orion and consistent with sites of
massive star formation in the Milky Way. If 𝐿NH3/𝐿IR serves as a proxy for the ‘mode’ of star formation, this hints that the
nature of star formation in extreme starbursts in the early Universe is similar to that of Galactic star-forming regions, with a large
fraction of the cold interstellar medium in this state, plausibly driven by a storm of violent disc instabilities in the gas-dominated
disc. This supports the ‘full of Orions’ picture of star formation in the most extreme galaxies seen close to the peak epoch of
stellar mass assembly.
Key words: gravitational lensing: strong – galaxies: starburst, high redshift – submillimetre: galaxies, ISM
1 INTRODUCTION
The progenitors of the most massive galaxies today are most likely
the population of intense star-bursting galaxies seen at 𝑧 & 2 (Smail
et al. 1997; Hughes et al. 1998). These starbursts have large gas
reservoirs (Bothwell et al. 2013) representing a significant fraction
of baryonic mass (Wiklind et al. 2019) fuelling very high rates of
star formation, possibly up to three orders of magnitude greater than
the Milky Way (Chapman et al. 2004; Barger et al. 2014).
Generally, these gas- and dust-rich systems are obscured in the op-
tical, but radiate strongly in the submillimetre andmillimetre through
their thermal dust emission. Indeed, the emission from key molecu-
lar and atomic tracers of the cool and cold dense interstellar medium
(ISM) responsible for fuelling star formation, and its immediate en-
vironment is also observed at these wavelengths.
Two key factors have improved our understanding of the nature of
these distant, dusty, prodigiously star-forming galaxies over the past
decade. One is the identification of large samples of strongly gravi-
tationally lensed systems (e.g. Danielson et al. 2013; Spilker et al.
2014; Rybak et al. 2015; Geach et al. 2015; Rybak et al. 2020). Lens-
ing amplifies flux, revealing emission features otherwise too faint to
detect, and magnifies images of galaxies to provide access to spatial
scales not achievable by any other means (Rybak et al. 2015; Geach
et al. 2018; Rybak et al. 2020). The second is the advent of sensitive,
wide bandwidth interferometry across the submillimetre–millimetre
★ E-mail: m.doherty2@herts.ac.uk
using large interferometric arrays, in particular the Atacama Large
Millimetre/submillimetre Array (ALMA). The follow-up of lensed
galaxies with ALMA has accelerated the establishment of a clearer,
albeit still incomplete, picture of the nature of galaxies undergoing
intense starbursts in the early Universe.
What astrophysics is responsible for the existence of extreme star-
bursts at high-z? Compared to the merger-dominated ultra luminous
infrared galaxies (ULIRGs) in the local universe with intense cir-
cumnuclear star formation, high-z starbursts of equivalent luminos-
ity appear to be sustaining star formation across much larger scales
Rujopakarn et al. (2011).
Mergers undoubtedly play a role in the triggering high-z star-
bursts (Tacconi et al. 2010; Engel et al. 2010), but there is some
observational evidence that some of the most intensely star-forming
galaxies are simply consistent with large gas-dominated rotationally
supported discs (Hodge et al. 2016; Jiménez-Andrade et al. 2018;
Geach et al. 2018; Gullberg et al. 2018). In these systems a signifi-
cant fraction of the ISM appears to have been driven to high density
(Oteo et al. 2017; Geach et al. 2018; Doherty et al. 2020), possibly
via violent disc instabilities (VDIs, e.g. Toomre 1964; Dekel et al.
2009b,a), and in this case the extreme star formation rates measured
can naturally be explained by the sheer quantity of gas available for
active participation in star formation (Geach & Papadopoulos 2012;
Papadopoulos & Geach 2012). We can learn more about the actual
conditions of the star-forming ISM, and the conditions of star for-
mation in general by exploiting gravitational lensing to study the
astrochemistry of these systems (Danielson et al. 2011; Spilker et al.
© 2022 The Authors
ar
X
iv
:2
20
9.
09
26
8v
1 
 [a
str
o-
ph
.G
A]
  1
9 S
ep
 20
22
2 M. J. Doherty et al.
Figure 1.Maps and spectra for the NH3 and CO(5–4) emission in 9io9. (left) Maps show the continuum-subtracted image plane emission averaged over the line
with contours shown for the fainter NH3 starting at 3𝜎. Synthesised beams are indicated top right. (right) Spectra show the source-integrated emission, with the
much brighter CO line acting as a reference profile that can be directly compared with the weaker NH3, where we overlay a scaled version of the CO(5–4) line
to demonstrate the strong similarity between the line profiles, despite NH3 generally tracing much denser gas (see Section 4 for a discussion). We bin the NH3
spectrum to ∼50 km s−1 channels for visualisation purposes. The ‘dip’ in NH3 emission at ∼100 km s−1 is not a significant feature.
2014; Zhang et al. 2018; Dye et al. 2022). If the density distribution
of the cold molecular reservoir is important for driving globally high
star formation rates, then observations of species that trace the dens-
est environments are required, particularly heavy rotor molecules
(Oteo et al. 2017; Béthermin et al. 2018).
In this work we report the detection of the ground state emission
line of ammonia (NH3) in a (now well-studied) strongly lensed star-
burst galaxy at 𝑧 = 2.6. A tracer of the dense molecular ISM and
intimately linked to the sites of star formation, NH3 was the first
polyatomic molecule detected in the ISM (Cheung et al. 1968) and
is amongst the most studied species in the local universe, primarily
through its radio inversion lines (Ott et al. 2011; Schmidt et al. 2016;
Fehér et al. 2022). The ground state NH3 (𝐽𝐾 = 10 → 00) ortho line
emits at 572.498GHz in the rest frame, and therefore any ground-
based studies of this feature at 𝑧 = 0 are hampered by the near-zero
transmission of the atmosphere at this frequency. In Section 2 we
present the observations and data reduction, in Section 3 we present
our analysis and results, and in Sections 4 and 5 we provide our inter-
pretation and conclusions. Throughout we assume a ‘Planck 2015’
cosmology where 𝐻0 = 68 km s−1Mpc−1 and Ωm = 0.31 (Planck
Collaboration et al. 2016).
2 OBSERVATIONS AND DATA REDUCTION
9io9 (J2000, 02h09m41.s3, 00◦15′58.′′5, 𝑧 = 2.5543) was observed
with the ALMA 12m array during project 2019.1.01365.S. The
C43-3 configuration was used, employing 48 antennas with base-
line separations of 15–784m. We executed a spectral scan in Band
4 across 𝜈obs = 152.5–162.9GHz with a total on-source time of ap-
proximately 195minutes over five executions. The precipitable water
vapour column was 1.9–3.5mm and the average system temperature
was 𝑇sys = 69–91K over the five executions. Atmosphere, bandpass,
phase and pointing calibrators included the sources J0423−0120
and J0217+0144. We use the pipeline-restored calibrated measure-
ment set for imaging. We image and CLEANed the data using casa
(v.5.1.0-74.el7) tclean with multiscale cleaning at scales of 0′′,
0.5′′, and 1.25′′. First we produce dirty cubes to establish the r.m.s.
(1𝜎) noise per channel, and then cleaned down to a stopping thresh-
old of 3𝜎. With natural weighting, and setting a common beam to
the whole datacube, the synthesized beam has a full width at half
maximum of 1.3′′ × 1.0′′ (position angle 72◦). The r.m.s. noise per
10MHz (20 km s−1) channel is 0.3mJy beam−1.
We adopt the same lens model as Geach et al. (2018). Briefly, the
lens model includes the gravitational potential of both the primary
lensing galaxy (𝑧 ≈ 0.2) and its smaller northern companion (as-
sumed to be at the same redshift). The model uses the semi-linear
inversion method of Warren & Dye (2003) to reconstruct the source
plane surface brightness that best matches the observed Einstein ring
for a given lens model. This process is iterated, varying the lens
model and creating a source reconstruction at each iteration, until
the global best fit lens model is found (Geach et al. 2018). The best
fit model is used to produce source-plane cubes. In the following,
we use the source-plane cube to extract the integrated spectrum, ac-
counting for magnification of the sources of line emission. However,
due to the relatively coarse resolution and signal to noise of the data,
we present maps of 9io9 in the image plane.
3 RESULTS
We use Splatalogue (Remijan et al. 2007) to identify emission lines
in the total spectrum. CO 𝐽 = 5 → 4 at 𝜈obs = 162.133GHz is
detected at high significance as expected, and exhibits the charac-
teristic double horned profile as other molecular and atomic lines
(Geach et al. 2015, 2018; Harrington et al. 2019; Doherty et al.
2020), and well-modelled in the source-plane reconstruction by a
rotating disc (Geach et al. 2018). We also detect a fainter, but sig-
MNRAS 000, 1–5 (2022)
Ammonia at 𝑧 = 2.6 3
nificant, emission feature at 𝜈obs = 161.072GHz that is consistent
with the redshifted ground state ortho-NH3 𝐽𝐾 = 10 → 00 rota-
tional line at 𝜈rest = 572.498GHz (Cazzoli et al. 2009, hereafter we
refer to the line as NH3). Image plane integrated spectra of the CO
𝐽 = 5 → 4 (hereafter CO(5–4)) and NH3 lines and maps are pre-
sented in Figure 1. To our knowledge, the previous highest redshift
detection of this transition of ammonia was in absorption in a spiral
galaxy at 𝑧 = 0.89, where the galaxy is acting as a lens, magnify-
ing the strong (sub)millimeter continuum emission of the famous
background quasar PKS 1830−211 at 𝑧 = 2.51 (Menten et al. 2008;
Muller et al. 2014).
To measure the line properties, we subtract continuum emission
on a pixel-by-pixel basis, using a simple linear fit to the spectrum
in line-free regions around NH3. We then model the integrated line
emission using an empirical template based on the high SNR CO(5–
4) emission line. By simply shifting the position of the CO(5–4)
emission in frequency space and scaling its amplitude, we can min-
imise the 𝜒2 difference between the scaled CO(5–4) and the NH3
line. Figure 1 shows how the scaled CO(5–4) line provides an ex-
cellent fit to the NH3 emission. We discuss the implications of this
later. The total integrated line flux is evaluated by summing the flux
within an aperture defined by the 3𝜎 contour of the averaged band
4 data cube. We measure 𝜇𝐿NH3 = (3.3 ± 0.2) × 107𝐿 , where 𝜇
is the lensing magnification. To estimate uncertainties in this pro-
cedure, we simply add Gaussian noise to each channel, with a 𝜎
determined from the r.m.s. in off-source regions of the data cube
and then repeat the fit 1000 times. Applying this same procedure
to the source plane reconstructions, we obtain a source plane line
luminosity of 𝐿NH3 = (2.8 ± 0.2) × 106𝐿 . If instead of using the
scaled CO(5–4) as a model of the emission, we just integrate over
the range Δ𝑉 = ±500 km s−1, we obtain a source plane luminosity
of 𝐿NH3 = (3.1 ± 0.3) × 106𝐿 .
4 INTERPRETATION
As can be seen from Figure 1, the scaled CO(5–4) emission is an
excellent description of the NH3 line emission. In turn, the line
profile of the integrated, projected CO emission is well modelled by
a nearly edge-on rotating disc (when modelled in the source plane
and projected into the image plane) and this profile is shared by the
majority of detected lines within this system covering a wide range
of conditions, from the relatively low density molecular reservoir
traced by C i (1–0) to the warmer, dense, ionised gas traced by N+
(Su et al. 2017; Geach et al. 2018; Harrington et al. 2019; Doherty
et al. 2020). The striking similarity in the observed line profiles imply
that the observed NH3 emission is broadly co-located with the CO-
emitting gas, likely emanating from discrete sites of on-going star
formation scattered throughout the gas-rich disc.
Ho&Townes (1983) note thatNH3 is a rather ubiquitousmolecule,
tracing a broad range of interstellar environments containing molec-
ular gas. However, for the rotational transitions in the millimetre,
NH3 is expected to trace dense gas. In the optically thin limit, the
critical density of NH3 is 𝑛crit & 107 cm3 for kinetic temperatures
𝑇𝑘 < 100K (Shirley 2015). In realistic scenarios, the NH3 emission
will be optically thick, and therefore subject to radiative trapping
(Draine 2011). This serves to lower the effective critical density, but
even so, the effective densities for optically thick NH3 emission are
still probing dense gas, with 𝑛eff & 5 × 104 cm3 for 𝑇𝑘 < 100K and
assuming a column density commensuratewith dense gas clumps and
cores in the Milky Way (log10 (𝑁ref/cm−2) = 14.3, Shirley (2015)).
However, the effective density will further scale down with increas-
Table 1. A comparison of the properties of 9io9 and Galactic sources where
NH3 is detected in emission.
Source 𝐿IR 𝐿NH3 𝐿NH3/𝐿IR
𝐿 𝐿 ×10−7
9io9 (this work) 1.1 × 1013 2.8 × 106 2.5
Orion-KL 8 × 104 0.01 1.3
W31 C ∼106 >0.042 >0.42
W49 N ∼107 >0.048 >0.48
ing column density as 𝑁ref/𝑁 . Another caveat is the presence of
significant far-infrared background fields, which will be dominated
by the ambient radiation field of the galaxy itself due to dust emis-
sion, with the most intense emission likely co-located with the dense
star-forming gas. This background could lead to significant radiative
pumping of NH3 molecule and therefore a non-collisional route to
rotational emission; indeed Schmidt et al. (2016) discuss the poten-
tial role of pumping of the NH3 rotational ground state emission as a
solution to the discrepancy between the abundances derived via the
NH3 ground state and its radio inversion lines in some local systems.
While radiative pumping would further serve to lower the effective
density of the gas responsible for NH3 emission, we can be reason-
ably confident that the observed NH3 is tracing some of the densest
molecular gas in 9io9, and therefore the actual sites of star formation.
Can we relate the properties of 9io9 to local star formation? Do-
herty et al. (2020) show that the average electron density (𝑛e ≈
300 cm−3) associated with warm ionised gas as traced by N+ fine-
structure emission is consistent with the typical density of Galactic
star-forming regions. The conclusion is that the conditions are not
‘extreme’ compared to sites of active star formation in the Milky
Way, but clearly a larger fraction of the ISM is participating in star
formation in 9io9 and galaxies like it compared to the Milky Way.
What of the efficiency, or ‘mode’ of star formation?A crude approach
is to compare proxies for the star formation rate and dense gas that is
fuelling it; more efficient star formation is characterised by a higher
rate per unit dense gas mass, with a theoretical upper limit set by the
Eddington limit (Murray et al. 2005). With the integrated infrared
luminosity as a proxy for the total star formation rate (for galaxies
dominated by dust) and NH3 as a tracer of the dense molecular gas
actively participating in star formation, we can use 𝐿IR/𝐿NH3 as an
empirical tracer of the star formation efficiency. In 9io9 we measure
𝐿NH3 = 2.8×106𝐿 and luminosity of 𝐿IR = 1.1×1013𝐿 , yielding
𝐿NH3/𝐿IR ≈ 3 × 10−7.
There are relatively few regions where we have robust NH3 (10 −
00) and integrated infrared luminosities. One such region, the Klein-
mann–Low nebula in Orion (Orion-KL) – a dense, hot molecular
cloud core close to the Trapezium cluster, which excites the Orion
Nebula – is frequently used as a local benchmark in many studies, not
the least because of its proximity at ∼400 pc. With 𝐿IR ∼ 8×104 𝐿
(Gezari et al. 1998) and 𝐿NH3 ≈ 0.01𝐿 (Olofsson et al. 2007;
Persson et al. 2007), Orion-KL has 𝐿NH3/𝐿IR ≈ 1.3 × 10−7, within
a factor of a few of the ‘global’ 9io9 ratio, despite eight orders of
magnitude separating the infrared luminosities.
While the Orion molecular clouds have been extensively stud-
ied (e.g. Genzel & Stutzki 1989), we note that it has been argued
that the energetics of Orion-KL are not dominated by high mass
star formation, but rather by an explosion (Zapata et al. 2011). Al-
though one would expect a correspondingly high density of super-
novae in 9io9, Orion-KL is arguably not a typical region in which
MNRAS 000, 1–5 (2022)
4 M. J. Doherty et al.
20 0 20 40 60
LSR [km s 1]
1.0
1.5
2.0
2.5
3.0
T A
[K
]
o NH3
572 GHz
W31 C
50 0 50
LSR [km s 1]
1.5
2.0
2.5
T A
[K
]
o NH3
572 GHz
W49 N
Figure 2. The ortho-NH3(10–00) line towards W31 C (top) and W49 N
(bottom) observed using Herschel/HIFI. Vertical dashed grey lines mark the
systemic velocities of the two sources.
presently high mass stars are forming. In contrast W31C (G10.6−4)
(𝐿IR ∼ 106𝐿 ) and the ‘mini-starburst’ W49 N (𝐿IR ∼ 107𝐿
Wright et al. 1977) are luminous Galactic high-mass star-forming
regions located at distances of 4.8 kpc and 11.2 kpc, for which ob-
servations of the NH3 (10 − 00) line have been published (Persson
et al. 2010, 2012). Toward both sources, the spectra were taken as
part of the Herschel key guaranteed time project PRISMAS1 us-
ing the Heterodyne Instrument for the Far-Infrared (HIFI). Unlike
the corresponding para-NH3 lines, which show almost exclusively
absorption, toward both W49 N andW31 C, the spectra of the ortho-
NH3 (10−00) line is far more complex, displaying strong emission at
the velocities of the background sources with self-absorption features
slightly offset from the systemic velocities (Figure 2). We model the
emission by fitting Gaussian profiles centred at the systemic velocity
of each source with the line widths optimised to fit the emission
wings. Subsequently, the self-absorption features modelled using
narrower Gaussian profiles centred at 0 km s−1 and 13 km s−1 for
W31 C and W49 N, respectively, are removed. The resulting fits are
then used to derive integrated line intensities of 6.33 K km s−1 and
1.33 K km s−1 for W31 C and W49 N, respectively. Using a conver-
sion factor of 482 Jy/K this yields line luminosities 𝐿NH3 = 0.042𝐿
and 𝐿NH3 = 0.048𝐿 , and 𝐿NH3/𝐿IR of 4.2 × 10−8 and 4.8 × 10−8
toward W31 C and W49 N, respectively. The NH3 line luminosities
should be considered lower limits due to uncertainties in the line
intensities estimated and the nature of the observed self-absorption.
If there is significant self-absorption of the NH3 emission when aver-
aged over galaxy scales in 9io9, then our measured luminosity could
also be considered a lower limit. In Table 1 and Figure 3 we compare
the luminosity ratios we derive for the Galactic sources with 9io9.
The ratios are broadly consistent within a factor of a few, despite the
fact that the 9io9 measurement is galaxy-integrated across a system
1 PRobing InterStellarMoleculeswithAbsorption line Studies (PI:M.Gerin)
Figure 3.A comparison of the 𝐿NH3/𝐿IR ratio versus 𝐿IR for 9io9 at 𝑧 = 2.6
and a small sample of Galactic sources where NH3 (10 − 00) is detected.
The x-axis spans ten orders of magnitude in infrared luminosity, whereas the
luminosity ratios are consistent within a factor of 10.
with an overall rate of star formation several orders of magnitude
greater than the Milky Way.
5 CONCLUSIONS
The actual structure of star-forming regions in gas-dominated, high-
redshift discs such as 9io9 remains unclear. While some studies have
argued for the presence of ‘giant clumps’ with properties similar to
the cores of local Giant Molecular Clouds, but scaled up to sizes of
order 100 pc (e.g. Rybak et al. 2015; Hatsukade et al. 2015), others
have pointed out that the reality of such features is questionable, and
that star formation may well be smoother, or structured on smaller
scales than can be reliably imaged interferometrically, even with the
assistance of lensing (Ivison et al. 2020). Regardless, it is evident
that in order to drive globally elevated star formation, a large fraction
of the cold ISM must be driven to high densities.
The introduction of supersonic turbulence is a key mechanism to
achieve high gas density fractions (Geach & Papadopoulos 2012),
with the dispersion of the log-normal distribution describing the
molecular gas density sensitive to the 1-dimensional average Mach
number: M = 𝜎𝑣/𝑐𝑠 , with 𝜎𝑣 the gas velocity dispersion and 𝑐𝑠
the speed of sound in the medium (Padoan & Nordlund 2002). In
the local Universe, mergers drive upM (e.g. Narayanan et al. 2011),
and is the primary mechanism for ultraluminous emission in galaxies
(Solomon & Vanden Bout 2005). 9io9 – like many other high red-
shift starbursts – does not appear to be undergoing a major merger
(although see Liu et al. 2022), but VDIs (Dekel et al. 2009b; Inoue
et al. 2016) are a viable alternative mechanism for locally driving up
M resulting in pockets of high-density gas, and therefore star for-
mation, across the gas-dominated disc. Confirming this in practice
will require reliable high-resolution imaging (noting the caveat refer-
enced above for interferometric data) that could map out the relative
distribution of dense molecular gas compared to the bulk reservoir. It
is important to note that minor mergers and interactions can catalyse
VDIs (e.g. Swinbank et al. 2011; Saha & Cortesi 2018).
That a large fraction of the molecular ISM in 9io9 resembles envi-
ronments like Orion-KL and other Galactic environments, with the
broad kinematics of NH3 consistent with ordered disc rotation across
the full range of molecular gas densities, we can picture an ensemble
MNRAS 000, 1–5 (2022)
Ammonia at 𝑧 = 2.6 5
of millions of ‘Orion-KLs’ embedded throughout the compact disc
of 9io9; perhaps individually unremarkable, but en masse driving
globally high star formation. This echos the evocative picture Rybak
et al. (2020) present of another dusty, star-forming lensed galaxy,
SDP.81 (𝑧 ≈ 3): they describe the system as ‘full of Orions’, based
on the similarity of the ISM conditions on sub-kpc scales in SDP.81
compared to Orion. Our results appear to support this picture, and
highlight the utility of the fainter, heavy rotor tracers in revealing the
structure of the high-density ISM that is physically proximate with
active star formation in young massive galaxies.
ACKNOWLEDGEMENTS
We are grateful to the anonymous referee for their constructive com-
ments.We also thankMatus Rybak for useful discussions.M.J.D. and
J.E.G. acknowledge support from theRoyal Society. S.D. is supported
by an STFC Rutherford Fellowship. This paper makes use of the fol-
lowing ALMA data: ADS/JAO.ALMA#2019.1.01365.S. ALMA is
a partnership of ESO (representing its member states), NSF (USA)
and NINS (Japan), together with NRC (Canada), MOST and ASIAA
(Taiwan), and KASI (Republic of Korea), in cooperation with the Re-
public of Chile. The Joint ALMA Observatory is operated by ESO,
AUI/NRAO and NAOJ. Funded by the Deutsche Forschungsgemein-
schaft (DFG,GermanResearch Foundation) under Germany’s Excel-
lence Strategy — EXC-2094 — 390783311. This research has made
use of the University of Hertfordshire high-performance computing
facility (http://stri-cluster.herts.ac.uk).
DATA AVAILABILITY
Data will be shared on reasonable request to the corresponding au-
thor.
REFERENCES
Barger A. J., et al., 2014, The Astrophysical Journal, 784, 9
Bothwell M. S., et al., 2013, Monthly Notices of the Royal Astronomical
Society, 429, 3047
Béthermin M., et al., 2018, Astronomy & Astrophysics, 620, A115
Cazzoli G., Dore L., Puzzarini C., 2009, Astronomy & Astrophysics, 507,
1707
Chapman S. C., Smail I., Windhorst R., Muxlow T., Ivison R. J., 2004, ApJ,
611, 732
Cheung A. C., Rank D.M., Townes C. H., Thornton D. D., WelchW. J., 1968,
Physical Review Letters, 21, 1701
Danielson A. L. R., et al., 2011, Monthly Notices of the Royal Astronomical
Society, 410, 1687
Danielson A. L. R., et al., 2013, Monthly Notices of the Royal Astronomical
Society, 436, 2793
Dekel A., et al., 2009a, Nature, 457, 451
Dekel A., Sari R., Ceverino D., 2009b, The Astrophysical Journal, 703, 785
Doherty M. J., Geach J. E., Ivison R. J., Dye S., 2020, The Astrophysical
Journal, 905, 152
Draine B. T., 2011, Physics of the Interstellar and Intergalactic Medium.
Princeton University Press
Dye S., et al., 2022, Monthly Notices of the Royal Astronomical Society, 510,
3734
Engel H., et al., 2010, The Astrophysical Journal, 724, 233
Fehér O., et al., 2022, The Astrophysical Journal Supplement Series, 258, 17
Geach J. E., Papadopoulos P. P., 2012, The Astrophysical Journal, 757, 156
Geach J. E., et al., 2015, Monthly Notices of the Royal Astronomical Society,
452, 502
Geach J. E., Ivison R. J., Dye S., Oteo I., 2018, The Astrophysical Journal,
866, L12
Genzel R., Stutzki J., 1989, ARA&A, 27, 41
Gezari D.Y., BackmanD. E.,WernerM.W., 1998, TheAstrophysical Journal,
509, 283
Gullberg B., et al., 2018, The Astrophysical Journal, 859, 12
Harrington K. C., et al., 2019, Monthly Notices of the Royal Astronomical
Society, 488, 1489
Hatsukade B., Tamura Y., Iono D., Matsuda Y., Hayashi M., Oguri M., 2015,
Publications of the Astronomical Society of Japan, 67, 93
Ho P. T. P., Townes C. H., 1983, Annual Review of Astronomy and Astro-
physics, 21, 239
Hodge J. A., et al., 2016, The Astrophysical Journal, 833, 103
Hughes D. H., et al., 1998, Nature, 394, 241
Inoue S., Dekel A., Mandelker N., Ceverino D., Bournaud F., Primack J.,
2016, Monthly Notices of the Royal Astronomical Society, 456, 2052
Ivison R. J., Richard J., Biggs A. D., Zwaan M. A., Falgarone E., Arumugam
V., van der Werf P. P., Rujopakarn W., 2020, Monthly Notices of the
Royal Astronomical Society: Letters, 495, L1
Jiménez-Andrade E. F., et al., 2018, Astronomy & Astrophysics, 615, A25
Liu B., et al., 2022, The Astrophysical Journal, 929, 41
Menten K. M., Guesten R., Leurini S., Thorwirth S., Henkel C., Klein B.,
Carilli C. L., Reid M. J., 2008, Astronomy & Astrophysics, 492, 725
Muller S., et al., 2014, A&A, 566, A112
Murray N., Quataert E., Thompson T. A., 2005, ApJ, 618, 569
Narayanan D., Krumholz M., Ostriker E. C., Hernquist L., 2011, MNRAS,
418, 664
Olofsson A. O. H., et al., 2007, Astronomy & Astrophysics, 476, 791
Oteo I., et al., 2017, The Astrophysical Journal, 850, 170
Ott J., Henkel C., Braatz J. A., Weiß A., 2011, The Astrophysical Journal,
742, 95
Padoan P., Nordlund A., 2002, The Astrophysical Journal, 576, 870
Papadopoulos P. P., Geach J. E., 2012, The Astrophysical Journal, 757, 157
Persson C. M., et al., 2007, A&A, 476, 807
Persson C. M., et al., 2010, A&A, 521, L45
Persson C. M., et al., 2012, A&A, 543, A145
Planck Collaboration et al., 2016, Astronomy and Astrophysics, 594, A13
Remijan A. J., Markwick-Kemper A., ALMA Working Group on Spec-
tral Line Frequencies 2007. https://ui.adsabs.harvard.edu/abs/
2007AAS...21113211R
Rujopakarn W., Rieke G. H., Eisenstein D. J., Juneau S., 2011, The Astro-
physical Journal, 726, 93
Rybak M., Vegetti S., McKean J. P., Andreani P., White S. D. M., 2015,
Monthly Notices of the Royal Astronomical Society: Letters, 453, L26
Rybak M., Hodge J. A., Vegetti S., van der Werf P., Andreani P., Graziani L.,
McKean J. P., 2020, Monthly Notices of the Royal Astronomical Society,
494, 5542
Saha K., Cortesi A., 2018, ApJ, 862, L12
Schmidt M. R., et al., 2016, Astronomy & Astrophysics, 592, A131
Shirley Y. L., 2015, Publications of the Astronomical Society of the Pacific,
127, 299
Smail I., Ivison R. J., Blain A. W., 1997, The Astrophysical Journal Letters,
490, L5
Solomon P., Vanden Bout P., 2005, Annual Review of Astronomy and Astro-
physics, 43, 677
Spilker J. S., et al., 2014, The Astrophysical Journal, 785, 149
Su T., et al., 2017, Monthly Notices of the Royal Astronomical Society, 464,
968
Swinbank A. M., et al., 2011, The Astrophysical Journal, 742, 11
Tacconi L. J., et al., 2010, Nature, 463, 781
Toomre A., 1964, The Astrophysical Journal, 139, 1217
Warren S. J., Dye S., 2003, The Astrophysical Journal, 590, 673
Wiklind T., et al., 2019, The Astrophysical Journal, 878, 83
Wright E. L., Fazio G. G., Low F. J., 1977, ApJ, 217, 724
Zapata L. A., Schmid-Burgk J., Menten K. M., 2011, A&A, 529, A24
Zhang Z.-Y., Romano D., Ivison R. J., Papadopoulos P. P., Matteucci F., 2018,
Nature, 558, 260
This paper has been typeset from a TEX/LATEX file prepared by the author.
MNRAS 000, 1–5 (2022)