Publications of the Astronomical Society of Australia (PASA)

c Astronomical Society of Australia 2017; published by Cambridge University Press.

doi: 10.1017/pas.2017.xxx.

Tracing HI Beyond the Local Universe

Martin Meyer1 , Aaron Robotham1 , Danail Obreschkow1, Tobias Westmeier1 , Alan Duffy2 ,
arXiv:1705.04210v1 [astro-ph.CO] 11 May 2017

Lister Staveley-Smith1
1 InternationalCentre for Radio Astronomy Research (ICRAR), The University of Western Australia, 35 Stirling Highway, Crawley,
WA 6009, Australia
2 Centre for Astrophysics and Supercomputing, Swinburne University of Technology, PO Box 218, Hawthorn, VIC 3122, Australia

Abstract
The SKA and its pathfinders will enable studies of HI emission at higher redshifts than ever before. In
moving beyond the local Universe, this will require the use of cosmologically appropriate formulae that have
traditionally been simplified to their low-redshift approximations. In this paper, we summarise some of the
most important relations for tracing HI emission in the SKA era, and present an online calculator to assist
in the planning and analysis of observations (http://hifi.icrar.org).

Keywords: radio lines: galaxies galaxies: distances and redshifts cosmology

1 Introduction 2 Redshift
Neutral atomic hydrogen (HI) is one of the most im- For observational HI studies, the redshift of a source is
portant tracers for studying the assembly of mass, simply defined as:
angular momentum, and structure in the Universe
hi obs
(Staveley-Smith & Oosterloo 2015; Blyth et al. 2015; zobs = 1= 1, (1)
Kim et al. 2015, 2016; Obreschkow et al. 2016). How- obs hi
ever, the difficulty of observing emission from this ma- where hi and hi are the emitted frequency and wave-
terial has meant that most studies have been restricted length of the HI line, and obs and obs are the corre-
to the local Universe, with only a small fraction of de- sponding frequency and wavelength at which it is ob-
tections occurring beyond redshift z0.1 (Jaffe et al. served. This redshift is the combined result of a number
2013; Giovanelli & Haynes 2016). As powerful new of potentially contributing effects, including the cos-
radio telescopes come on-line (ASKAP, MeerKAT, mological redshift due to the expansion of the Uni-
WRST-Apertif, FAST, SKA), such observations will be- verse, local motions in either the source or observer rest
come commonplace (Duffy et al. 2012; Holwerda et al. frames, and gravitational redshifts caused by (evolv-
2012; Staveley-Smith & Oosterloo 2015), necessitating ing) potential wells along the line of sight. Considering
the use of formulae that allow for an evolving Uni- just the cosmological redshift (zcos ) and the peculiar
verse. A basic summary of the cosmologically appro- source observer
motions of the source (zpec ) and observer (zpec )
priate formulae for a variety of standard HI quantities (Davis & Scrimgeour 2014):
is given in the following sections, with the aim of pro- source observer
viding a reference document for upcoming pathfinder 1 + zobs = (1 + zcos )(1 + zpec )(1 + zpec ) . (2)
studies. In the final section, we introduce an accompa- Different standards of rest or flow models are often
nying online calculator based on the provided formu- used to minimise peculiar velocity effects, particularly
lae. Useful references for work underpinning the ma- in the nearby Universe. At higher redshifts (e.g. z &
terial here include Hogg (1999), Abdalla & Rawlings 0.03, Baldry et al. 2012), the cosmic microwave back-
(2005), Obreschkow et al. (2009), Ned Wrights online ground offers a reference frame that can be used to re-
tutorial1 , and CosmoCalc.2 move local peculiar velocity effects, reducing Equation 2
to:
source
1 + zobs = (1 + zcos )(1 + zpec ). (3)
martin.meyer@uwa.edu.au
1 www.astro.ucla.edu/
~ wright/cosmo_01.htm
In this paper, unless otherwise specified, we make
2 cosmocalc.icrar.org the overall simplifying assumption that the observed
1
2 Meyer et al.

redshift is equal to the cosmological redshift (zobs = ing distance (DM ), the luminosity distance to a galaxy
observer source
zcos = z, i.e. zpec = zpec = 0), but be aware that, (DL ), and the angular diameter distance (DA ). Follow-
particularly in the local Universe, there can be a non- ing Hogg (1999), the comoving distance DC is given by:
negligible difference between these redshifts and you Z z
may need to adjust for this assumption. c
DC (z) = E 1 (z )dz , (8)
H0 0

3 Parametrizations of Cosmic Expansion where E(z) is as expressed in Equation 5. As before, this

can also be modified to include a model for a varying
The relationship between proper (physical) distance
vacuum energy density if desired. From DC we can then
and the recessional velocity of galaxies due to the ex-
express the transverse comoving distance DM :
pansion of the Universe at a given epoch is parametrized
through the Hubble parameter H(z), with Hubbles  
Constant H0 = H(0) its value at the present time. If we

c H0 K

H


sinh c D C if K > 0 ,
divide the energy density of the Universe, normalised

0 K

by the critical energy density, into its fractional com- DM (z) = DC   if K = 0 ,
ponents of matter (M ), radiation (R ), vacuum en- H0 |K |

c
sin DC if K < 0 ,


H0 |K | c
ergy ( ), and also include a term for spatial curva-
ture (K = 1 = 1 M R ), then we can (9)
express H(z) (Weinberg 2008): and finally the luminosity distance DL and angular di-
ameter distance DA :
H(z) = H0 E(z) , (4)
where
p DL (z) = (1 + z)DM (z) , (10)
E(z) = R (1 + z)4 + M (1 + z)3 + K (1 + z)2 + . DA (z) = DM (z)/(1 + z) . (11)
(5)
Note that the i values here correspond to those at z = As these distances are all inversely proportional to the
0, and that the above expression does not include any chosen value of the Hubble constant, this dependence
redshift-dependent factors alongside . If desired, a can be included explicitly via little h in the units, e.g.
model can also be included for a varying vacuum energy h1
C Mpc.
density, as discussed in Section 20.
The dependence of a given quantity on the Hubble
constant is often explicitly stated alongside the physical 5 Line-of-Sight Velocity
units of a measurement through the use of the dimen- The various different definitions of velocity offer a sig-
sionless Hubble constant (little h): nificant potential source of confusion, particularly given
H0 the historical use of velocity in HI spectral line studies
h . (6) to describe both rest frame motions (i.e. the motions
100 km s1Mpc1
of objects through space), Vpec , as well as a proxy for
The various uses of this quantity and its pitfalls are
redshift caused by the expansion of space, Vcos .
well-described in Croton (2013). In this paper, and to Beginning with the full relativistic expressions for
further aid the comparison of values contained in the these two velocities, for Vpec we have the special rel-
historical literature, we give all Hubble Constant depen-
ativity expression (Einstein 1905; Davis et al. 2003):
dencies in terms of hC , allowing for the practice some-
times used of specifying the precise value of H0 used in
the little h nomenclature: Vpec (zpec ) = VSR (zpec ) , (12)
H0 2
rest 2
obs
hC . (7) =c
C km s1 Mpc1 2
rest 2 ,
+ obs
(13)
An explanation of how to convert values from one value (1 + zpec )2 1
of the Hubble constant to another with a specified hC =c , (14)
(1 + zpec )2 + 1
dependence is given in Appendix A.
where these equations assume that motion is purely
along the line of sight relative to the observer (see
4 Cosmological Distances
Einstein for the relevant equations where the source has
There are a number of cosmological distances relevant a transverse velocity component), and that the observer
to the calculation and understanding of HI quanti- is in the same inertial reference frame as the source.
ties, most notably including the line-of-sight comov- In comparison, for Vcos we have the general relativis-
ing distance to a galaxy (DC ), the transverse comov- tic expression (Davis et al. 2003):
PASA (2017)
doi:10.1017/pas.2017.xxx
HI Beyond the Local Universe 3

Vcos (zcos , zref ) = VGR (zcos , zref ) , (15) obs =  hi  Vrest . (19)
Vrest 2


H(zref ) zcos dzcos
Z c(1 + zcos ) 1 2c
=c )
, (16)
1 + zref 0 H(zcos Given the non-relativistic assumption Vrest c, and
where zref is the redshift at which the velocity is to be again assuming that any systematic peculiar velocity
evaluated. In practice, as we are observing the object can be ignored (zobs = zcos = z), Equation 19 simpli-
at zcos at the current epoch, zref = 0. In this situation, fies to the basic relations used in this paper to convert
the above equation simplifies to (and as used in Cosmo- between Vrest and obs :
Calc):
c(1 + z) c
zcos
Vrest obs = obs . (20)
hi obs
Z
Vcos (zcos ) = c E 1 (z )dz . (17)
0 While discouraged, if you need to convert between ob-
Comparing the above expression with Equation 8, it served frame and rest frame velocities (such as might
can be seen that Vcos then is the velocity that restores need to be done when using some software which by
a basic Hubble Law relation Vcos (z) = DC (z)/H0 for all default will measure observed frame widths in terms of
redshifts. optical velocity), or similarly between observed frame
For small redshifts (z < 0.1), both the special rela- and rest frame frequency widths, this can be done us-
tivistic and general relativistic formulae can be approx- ing:
imated by the optical velocity convention (to distin-
guish this from the now deprecated radio velocity al- 1
Vrest = Vobs , (21)
ternative): 1+z
1
Vopt (z) = cz . (18) obs = rest . (22)
1+z
As a general rule to avoid confusion, the use of velocities As a trivial extension, Equation 20 can also be used
is best restricted to describe source rest frame motions to determine the rest frame velocity resolution for a
(e.g. galaxy rotation, peculiar velocities), and not as a fixed observed frame channel width (setting obs =
proxy for observed frame quantities (e.g. cosmological chan ).
redshift, distance, or observed frame frequency width).
To separate these two potential uses of velocity in the
7 Beam
rest of the paper, we refer to Vrest ( czpec , as in general
for HI, Vpec c and so this is a good approximation of For a telescope with a normalised main beam sensitiv-
VSR ) and for the non-recommended observed frame ve- ity response that can be represented as a 2D elliptical
locity we refer to Vobs ( czobs , which while only an ac- gaussian (and ignoring position angle):
curate approximation of VGR at low redshift, has the ad-
vantage that it is readily invertible to obtain the source 
x2 y2

2 + 2
redshift if desired, and is the form that has traditionally Pn (x, y) = e 2x 2y
, (23)
been used in the literature). For reference, a compari-
its noise-equivalent field of view, or equivalent solid
son plot showing the differences between VSR (z), VGR (z)
angle sampled at full sensitivity, is given by (e.g. Kraus
and Vopt (z) can be found in Davis & Lineweaver (2001).
1986):

6 Emission Profile Width Z Z

bm = Pn (x, y) dxdy , (24)
Along with redshift, another key parameter that can
be measured from HI profiles is their frequency width, = 2x y , (25)
which provides a line-of-sight measure of velocity dif- ab
ferences in the source material, predominantly caused = . (26)
4 ln(2)
by galaxy rotation. In a similar vein to the discussion p
of the previous section, to avoid confusion it is recom- where a (= 2 2 ln(2) px ) is the main beam angular ma-
mended that widths in the source rest frame be specified jor axis, and b (= 2 2 ln(2)y ) is the angular minor
in terms of velocity, Vrest , while those in the observed axis, both measured at the half power point.
frame be given in frequency, obs . The relation be- Note that the half power points a, b obs (1 + z),
tween the two can be derived from Equation 3. Using so for a synthesised beam with no frequency-dependent
the non-relativistic approximation for rest frame veloc- weighting, or a non-compound primary beam, the beam
ity width, zpec = Vpec /c = Vrest /(2c), this gives: solid angle will vary as:
PASA (2017)
doi:10.1017/pas.2017.xxx
4 Meyer et al.

tial to underestimate the source flux if the pixel is not

exactly centred on the source), or weighting the pixel
bm (z) = (1 + z)2 bm (0) . (27) values by the telescope beam response (which avoids
In practice, the illumination of a telescope is frequency this problem, but increases the effective angular area
dependent, so the beam area may not scale exactly as over which the measurement is being made compared
(1 + z)2 as given above. A broadband feed will gener- to the single pixel method, and as such can increase the
ally under-illuminate a dish at the highest frequencies, impact of source confusion).
making the beam larger (and the efficiency lower).
9 Flux
8 Flux Density
For a source at redshift z, with a rest frame total HI
The observed flux density, S , for a source with a luminosity of L, its observed flux S will be:
rest frame luminosity density of Lrest = L(1+z)obs =
L(1+z) will be (Peacock 1999): Z
S= S dobs , (32)
L(1+z) Z
Lrest
S = (1 + z) 2 . (28) = (1 + z)
4DL 4DL 2 dobs , (33)

Lrest
Z
The preferred unit for S is Jy (= 1026 W m2 Hz1 ). =
4DL 2 drest , (34)
The maximum of S for a given source is often referred
to an objects peak flux density. L
As a side note, the intensity scale of HI images is S = 2 . (35)
4DL
usually presented in units of Jy per beam, reflecting
the fact that a telescope will measure any flux within The preferred unit for this quantity is Jy Hz (=
its beam when pointed a source, i.e. that the specific 1026 W m2 ). In HI studies, reference is often made to
intensity observed, Iobs , is the real specific intensity at a similar, but dimensionally different quantity defined
the location of the observer, I , convolved with the nor- as the integral of flux density as a function of velocity
malised telescope beam, Pn . The flux density, S of the rather than frequency:
source is then given by (Kraus 1986):
Z
Z SV = S dV , (36)
S = I (, ) d , (29)
src
the units for which are Jy km s1 (=
1
Z
= I (, ) Pn (, ) d , (30) 1026 W m2 Hz1 km s1 ). This is a poor quan-
bm src
tity for cosmological measurements as it introduces
1
Z
= I obs (, ) d . (31) some uncertainty about exactly what velocity (or
bm src pseudo-velocity) has been used in its calculation, as
discussed earlier. It is preferable to use S rather than
R To recover
obs
the correct flux for an extended region,
S V wherever possible, and particularly so if referring
I
src (, ) d is measured by summing values of the
to a measurement of an objects observed total flux
pixels in the source region (in Jy per beam) multiplied
given the dimensional difference of S V from that of a
by the pixel area, from which S can then recovered by
natural flux value.
dividing by the area of the telescope beam. Or expressed
The conversion between S and S V depends on the ve-
alternatively, S can be measured by simply converting
locity convention used. If S V is measured in the optical
the specific intensity values of the image to Jy per pixel,
observed frame:
and then summing over the region of interest.
Note that if the ratio of beam area to pixel area is
constant as a function of frequency (i.e. both pixel area 
S Vobs

c(1 + z)2

S

and beam area are scaling as 1 + z), then the correction = , (37)
Jy km s1 hi Jy Hz
factor for the extended source flux sum will be constant, 
S

but if not, e.g. the image cube has a fixed angular pixel 2.11 104 (1 + z)2 . (38)
Jy Hz
sale, a varying correction will be required.
For a point source, while this approach could also be
taken, it results in a sub-optimal signal-to-noise mea- Alternatively, if the velocity used for the calculation of
surement compared to either just taking the value of the S V is a source rest frame velocity (for which Vopt VSR
pixel centred on the source (although this has the poten- at the velocities relevant for galaxy rotation), then:
PASA (2017)
doi:10.1017/pas.2017.xxx
HI Beyond the Local Universe 5

12 HI Mass Function

S Vrest

c(1 + z)

S
 The HI mass function, (Mhi ), gives the number of
= , (39) sources per unit volume as a function of HI mass, i.e.
Jy km s1 hi Jy Hz
  the volume density of sources, n, having masses between
S
2.11 104 (1 + z) . (40) M1 and M2 will be given by:
Jy Hz
Z 2
M
n= hi )dM
(M hi , (48)

M1
10 Number of HI Atoms where for convenience in the equations that follow we
3
Taking of HI atoms to be in the upper hyperfine state, have expressed this in terms of dimensionless mass
4
M hi Mhi /M , M being one of the Schechter func-
with a spontaneous emission rate of Ahi , an emitted hi hi
photon energy of hhi , and an HI source with luminosity tion parameters traditionally used to parametrize the
L to be optically thin, the number of HI atoms, Nhi , mass function. Clearly the mass function could also
will be given by: be expressed without this change of variable. The
Schechter function form is given by (Schechter 1976;
Zwaan et al. 2005):
L
Nhi = 3 , (41)
4 hhi Ahi hi )dM
hi = M
hi
M hi ,
2
(M e hi dM (49)
16DL S
= , (42) (50)
3hhi Ahi
where provides the overall normalisation, gives

the slope of the low mass power-law, and Mhi gives
2 
the characteristic turnover mass, above which the num-
   
Nhi DL S
5.91 1058 . (43)
h2
C
1
hC Mpc Jy Hz ber of sources exponentially declines. Alternatively, the
number of sources is often expressed as an integral over
log(M hi ) rather than M
hi , ie:

2)
log(M
11 HI Mass
Z
n= hi )d log M
(M hi , (51)
Using the above relation, the HI mass of a source is 1)
log(M
trivially given by (caveat corrections that need to be which gives (Springob et al. 2005; Martin et al. 2010):
applied if the source is not optically thin):
hi )d log M
(M hi = ln(10)M
+1 eMhi d log M
hi .
hi
(52)
Mhi = Nhi mH , (44)
In practice, the HI mass function is normally presented
as log() vs. log(M hi ). In this space, the functional
   2   form of the Schechter function is given by:
Mhi DL S
49.7 . (45) log = log( ln 10) + (1 + ) log M hi M
hi log(e) .
h2
C M h1
C Mpc
Jy Hz
(53)

Which compares to the relation expressed in terms of For evolutionary studies, the mass function is best ex-
traditional observed frame velocity integrated flux: pressed in comoving coordinates to remove normalisa-
tion changes caused purely by the expansion of the Uni-
verse.
2 
2.35 105 S Vobs Note also that the Schechter form of the HI mass func-
   
Mhi DL
.(46)
2
hC M (1 + z)2 1
hC Mpc Jy kms1 tion is not consistently defined in the literature, and so
you may need to pay attention to the particular defini-
tion used when comparing results.
Or rest frame velocity integrated flux:

2  13 Cosmological Mass Density

2.35 105 S Vrest
   
Mhi DL
2
1
.(47) The cosmological mass density of HI can be calculated
hC M 1+z hC Mpc Jy kms1
by carrying out the mass-weighted integral over the HI
mass function:
PASA (2017)
doi:10.1017/pas.2017.xxx
6 Meyer et al.

hi will also change, even if the comoving density of

Z HI does not. To remove this effect, the practice is often
hi = Mhi (Mhi )dMhi . (54) used of replacing crit (z) with crit (z = 0) in the above
0
expressions, ie:
This can either be simply summed directly from the
measured HI mass function data points (providing a suf- hi (z)
hi (z) = , (61)
ficient mass range is spanned and making sure to correct crit (z = 0)
values to those appropriate for the log Mhi binwidths which also offers the convenience of pcrit (z = 0) =
actually used), or through the following analytic solu- ccrit (z = 0), but be aware of the difference between this
tion expressed in terms of the fitted Schechter parame- and the above derivations, and be sure to specify which
ters and the complete Gamma function, (Zwaan et al. definition of hi (z) is being used.
2003):

hi
 14 Brightness Temperature
=
hC M Mpc3 The brightness temperature corresponding to an ob-
served flux density S , measured with a telescope of
   
Mhi
(2 + ) . (55) main beam solid angle bm , is the blackbody tempera-
h3C Mpc3 2
hC M
ture an extended object would need to have to produce
the observed flux in the Rayleigh-Jeans limit (h kT ;
For an HI mass function measured using comoving vol- Wilson et al. 2009). From the Rayleigh-Jeans law, the
ume, this will similarly yield a cosmic HI density in co- luminosity emitted by a blackbody per unit area into a
moving coordinates, and so hi should remain constant unit solid angle, B is given by:
with redshift/lookback time if no evolution is occurring
aside from the expansion of the Universe. 2k 2 T
The cosmic mass density of HI can also be expressed B = , (62)
c2
as a fraction of the critical mass density: where k is Boltzmanns constant. In a Euclidean geom-
etry, I = B (surface brightness conservation; I is the
hi (z) received specific intensity), giving the traditional local-
hi (z) = , (56)
crit (z) Universe relation for HI brightness temperature:
where in proper (physical) coordinates crit (z) is given
2
by: S 2khi T
B = I = = 2
, (63)
3H 2 (z) bm c
= pcrit (z)
, (57) c2 S
8G TB = . (64)
2
2khi
or in comoving coordinates: bm

1 3H 2 (z) However, in a relativistic Universe, we have (Peacock

ccrit (z) = . (58) 1999):
(1 + z)3 8G
Evaluating these expressions for hi gives: B(1+z) Brest
I = = , (65)
! (1 + z)3 (1 + z)3
3.60 108 H 2 (z)
 
hi (z)
=
h1
C h2 1
C M Mpc
3
(59)
giving:
p
hi (z)
 
,
hC M Mpc3 S 2k 2 T
I = = 2 hi 3 , (66)
or: bm c (1 + z)

hi (z)
 c2 S
TB = (1 + z)3 2 . (67)
h1 2khi bm
C
!
3.60 108 H 2 (z) chi (z) Evaluating the constants gives:

1
= 3
.
(1 + z)3 h2 1
C M Mpc hC M Mpc3
1
(60)
   
TB S bm
= 6.86 105 (1 + z)3 , (68)
K Jy arcsec2
Note that for an expanding Universe, if there exist con-
tributors to the total energy density other than matter, or expressed alternatively as a function of the beam
ccrit will evolve with redshift along with pcrit . As such, angular major and minor axes a and b:
PASA (2017)
doi:10.1017/pas.2017.xxx
HI Beyond the Local Universe 7

 
1 Nhi
= 1.25 1024 (1 + z)3
   
TB 5 3 S ab (77)
= 6.06 10 (1 + z) . (69) cm2
K Jy arcsec2 1
S Vrest
 
bm
,
Jy kms1 arcsec2
15 Column Density
 
Nhi
= 1.10 1024 (1 + z)3 (78)
cm2
The HI column density gives the number of atoms per 1
S Vrest
 
unit area along the line of sight through an astronomical ab
.
object. The column density Nhi for a flux S measured Jy kms1 arcsec2
over solid angle is given by (using Equation 42 for the
number of atoms, and the small angle approximation to The above column density expressions can be readily
calculate the source area; material is again assumed to expressed as HI surface mass densities by multiplying
be optically thin): mH and converting to the desired units, eg.:
   
2 HI Nhi
= 8.01 1021
  
Nhi 16DL S 1 , (79)
Nhi = = 2 , (70) M pc2 cm2
area 3hhi Ahi DA bm   1
S bm
16 S = 2.12 (1 + z)4 , (80)
= (1 + z)4 , (71) JyHz arcsec2
3hhi Ahi bm
= 1.00 104 (1 + z)2 (81)
1
S Vobs
 
bm
,

Nhi
 Jy kms1 arcsec2

cm2 = 1.00 104 (1 + z)3 (82)
(72) 1
S Vrest
 

S

bm
1 bm
= 2.64 1020 (1 + z)4 . .
JyHz arcsec2 Jy kms1 arcsec2

Or in terms of the beam angular major and minor axes Column density can also be expressed as a function
a and b: of brightness temperature, TB . Following Wilson et al.
(2009); Giovanelli & Haynes (1988); Brinks (1990), for
a cloud with spin temperature TS and optical depth
2
  
16DL S 4 ln(2) embedded in a radiation field having brightness temper-
Nhi = 2 ab , (73)
3hhi Ahi DA ature TR , its brightness temperature, TB , will be given
by:
 
Nhi
TB = TR e + TS (1 e ) . (83)
cm2
(74)
  1 If the cloud is optically thin ( 1), and assuming that
S ab
= 2.33 1020 (1 + z)4 . background radiation field has negligible impact, then:
JyHz arcsec2

For comparison, the expressions in terms of observed TB = TS . (84)

frame velocity integrated flux are: The optical depth per unit frequency can be related to
the density of hydrogen atoms per unit length along the
  line of sight:
Nhi
= 1.25 1024 (1 + z)2 (75)
cm2
1 3c2 h
S Vobs
 
bm d = Ahi nhi ()dl , (85)
, 32hi kTs
Jy kms1 arcsec2

Nhi

where nhi refers to the volume density of hydrogen
= 1.10 1024 (1 + z)2 (76) atoms, and is theR line shape function reflecting the
cm2
1 natural line width ( ()d = 1). Integrating this over
S Vobs
 
ab
. and s gives:
Jy kms1 arcsec2

32khi
Z Z
And for rest frame velocity integrated flux: nhi dl = Ts d . (86)
3Ahi hc2
PASA (2017)
doi:10.1017/pas.2017.xxx
8 Meyer et al.

The left-hand side of this

R expression is the desired HI give the sensitivity in Jy:
column density Nhi = nhi dl, while the right-hand side
can now be written as a function of brightness temper-
 
S 1952.5
=
ature using Equation 84: Jy
p
s N (N 1)
(96)
 1   1   1
Ae /Tsys t 2 2
32khi
Z
2
.
Nhi = TB drest , (87) m K 1 s Hz
3Ahi hc2
  Z   
Nhi TB drest The corresponding flux sensitivity scaling is:
= 3.85 1020 .(88)
cm2 K MHz
Note that the frequency integral here is in the source S = S , (97)
rest frame. As both Nhi and TB are also rest frame s
quantities, there are no (1 + z) factors in the above ex- SEF D
= , (98)
pression. Alternatively, Equation 87 can also expressed s 2N (N 1)t
as an integral over rest frame velocity: s
kTsys 2
= . (99)
2
s Ae N (N 1)t
32khi
Z
Nhi = TB dVrest , (89)
3Ahi hc3 Or:
  Z   
Nhi 18 TB dVrest
= 1.82 10 .(90) 
S

1952.5
cm2 K kms1 =
JyHz
p
s N (N 1)
To recover the previous expression for HI column den- (100)
 1   1  1
sity (Equation 71), Equations 67 and 87 give: Ae /Tsys t 2 2
.
m2 K1 s Hz
32khi c2 S
Z
3 Observationally, if the flux density sensitivity aver-
Nhi = (1 + z) 2 drest , (91)
3Ahi hc2 2khi bm aged over channel interval chan is known, Schan , the
16 1
Z
= (1 + z)4 S dobs , (92) flux and flux density sensitivities over interval can
3hhi Ahi bm be simply calculated using:
16 S
= (1 + z)4 , as before. (93)
3hhi Ahi bm r
chan
S = Schan , (101)

16 Flux and Flux Density Sensitivity p
S = Schan chan . (102)
If not known observationally, the flux density sensi-
tivity of a single-pointing, naturally-weighted, Stokes- As observed frame quantities, these have no inher-
I, non-primary beam corrected interferometric image ent scaling as a function of redshift, however telescope
can be estimated using (Taylor et al. 2008; Kraus 1986; performance may varying significantly as a function of
Obreschkow et al. 2011): frequency, often encoded as variations in the Ae /Tsys
term of Equations 95 & 99 (and system efficiency s if
desired), introducing a redshift scaling of:
SEF D
S = p , (94)
s 2N (N 1)t
s S (z) = f (z)S (0) , (103)
kTsys 2
= , (95) S (z) = f (z)S (0) , (104)
s Ae N (N 1)t
where:
where SEFD is the System Equivalent Flux Density of
s Ae
an antenna in the array (= 2kTsys /Ae ; assumed to be Tsys (z = 0)
the same for all elements in the above relation), is f (z) = s Ae
. (105)
the frequency width of interest, t is the integration Tsys (z)
time, Tsys is the system temperature, N is the number
As will become apparent in the next sections concerning
of antennae, Ae is the effective area of each antenna, s
is the system efficiency, and k is Boltzmanns constant. brightness temperature and column density sensitivity,
It should be noted that in practice a number of fac- it can also be useful to understand how these relations
tors may reduce the achieved observational sensitivity vary as a function of rest frame velocity interval Vrest .
from the theoretical value above, such as the application Substituting Equation 20 into Equations 101 & 102
of weighting and tapering. Evaluating the constants to gives:
PASA (2017)
doi:10.1017/pas.2017.xxx
HI Beyond the Local Universe 9

c2
s
c(1 + z)chan TB = (1 + z)3 S , (118)
S = Schan , (106) 2k 2hi bm
hi Vrest
s
1 
hi Vrest chan   
TB
 bm S
S = Schan , (107) = 6.86 105 (1 + z)3 ,
c(1 + z) K arcsec2 Jy
(119)
and a redshift scaling of (with Ae /Tsys variations):
which combining with the previous expression for flux
V const 1 V const density sensitivity (Equation 95) yields:
S rest (z) = f (z)(1 + z) 2 S rest (0) , (108)
c2 Tsys
const const
Vrest 1 Vrest 1
S (z) = f (z)(1 + z) 2 S (0) . (109) TB = (1 + z)3 2
, (120)
hi Ae bm s 2N (N 1)t
p

Widefield Single-Beam Mosaic:

For a widefield mosaic in which many overlapping  1.34 109 (1 + z)3

TB
primary beam pointings have been used to image a =
K
p
s N (N 1)
large area, an additional factor needs to be considered.  1  1   1   1
For a fixed pointing pattern, the primary beam area bm Ae /Tsys t 2 2
2 2
.
for each pointing will also increase as (1 + z)2 , which arcsec m K 1 s Hz
will in turn lead to the noise decreasing as (1 + z)1 (121)
(Abdalla & Rawlings 2005), giving:
Using Equation 101 this can also be expressed as a func-
tion of the flux density sensitivity per channel chan
S (z) = f (z)(1 + z)1
S (0) , (110) (width chan ):
1

S (z) = f (z)(1 + z)
S (0) , (111) r

V const
S rest (z) = f (z)(1 + z) 21
SVrest (0) ,
(112) c2 3 chan
TB = (1 + z) 2 Schan , (122)
const
Vrest 3
2khi bm

S SVrest (0) .
(z) = f (z)(1 + z) 2 (113)
 
TB
Phased Array Feed Observations: = 6.86 105 (1 + z)3
K
(123)
For a phased array feed (PAF), the noise-equivalent field
r 1 
Schan
 
chan bm
of view (FoVNE ) will not increase as (1 + z)2 as in the .
arcsec2 Jy
single beam case above, due to correlated noise effects
between adjacent formed beams. Indeed, once the cen- Or alternatively expressed per desired rest frame ve-
tral part of the field is fully sampled, FoVNE will not locity interval Vrest rather than per observed frame
continue to rise with redshift apart from some growth frequency width using Equation 106:
at the edges. Leaving FoVNE (z) as a generic function,
and applying the same inverse square root scaling of c5/2
r
7/2 chan
this to derive the observed noise in a surveyed area, as TB = (1 + z) 5/2
S ,(124)
2khi bm Vrest chan
above, gives:
  7
TB
= 9971 (1 + z) 2
1 K

S (z) = f (z)FoVNE2 (z)
S (0) , (114) 1  1   1 
Schan
 
1
bm chan 2 Vrest 2

S (z) = f (z)FoVNE2 (z)
S (0) , (115) .
arcsec2 Hz km s1 Jy
V const 1 1
(125)
SVrest (0) , (116)
S rest (z) = f (z)FoVNE2 (z)(1 + z) 2

const
Vrest 1 1

S SVrest (0) .(117)
(z) = f (z)FoVNE2 (z)(1 + z) 2 In the instance of a comparison being made for obser-
vations with a given telescope at different redshifts, but
where the configuration has been changed between ob-
servations to yield the same synthesised beam solid an-
gle, this will give a redshift scaling relation of:
17 Brightness Temperature Sensitivity
The relativistically derived relation between brightness
TB (z) = f (z)(1 + z)3 TB (0) , (126)
temperature and flux density (Equation 67), gives a cor-
V const V const
responding sensitivity relation of: TB rest (z) = f (z)(1 + z)7/2 TB rest (0) , (127)
PASA (2017)
doi:10.1017/pas.2017.xxx
10 Meyer et al.

where f (z) captures redshift variations in s Ae /Tsys as  N 

hi = 5.15 1023 (1 + z)4

before (Equation 105). If instead we now consider how cm 2
p
s N (N 1)
the brightness temperature sensitivity will change for  1  1   1  1
bm Ae /Tsys t 2 2
a fixed configuration observation across a range of fre- .
arcsec2 m2 K1 s Hz
quencies, the synthesised beam solid angle will increase
as bm (1 + z)2 (ignoring frequency dependent illu- (137)
mination effects) giving:
Or from Equation 102, scaling from known channel
noise:

TB (z) = f (z)(1 + z)TB (0) , (128)
const
Vrest const
Vrest

(z) = f (z)(1 + z)3/2 (0) . (129) 16 chan
TB TB Nhi = (1 + z)4 Schan , (138)
3hhi Ahi bm
Taking these last equations as a starting point, we can
also derive the redshift scaling equations for a widefield  N 
mosaic and for a phased array feed as per the previ- hi = 2.64 1020 (1 + z)4
cm2
ous section allowing for overlapping primary (formed-) 1  1  1 
2 chan 2 Schan
 
beam effects: bm
.
arcsec2 Hz Hz Jy
Widefield Single Beam Mosaic: (139)

TB (0) ,
TB (z) = f (z) (130) And finally as a function of rest frame velocity interval
V const V const using Equation 107:
TB rest (z)
= f (z)(1 + z)1/2
TB rest (0) . (131)

Phased Array Feed Observations: 7/2 16 Vrest chan
Nhi = (1 + z) 1/2
Schan ,(140)
1
3hhi Ahi bm c1/2
TB (z) =
f (z)FoVNE2 (z) (132)
TB (0) ,
(1 + z)  N 
hi = 1.82 1022 (1 + z) 27

const
Vrest 1 cm 2

TB (z) = f (z)FoVNE2 (z) (133) 1  1  1 
Vrest 2 chan 2 Schan
 
V const
bm
(1 + z)3/2
TB rest (0) . .
arcsec2 km s1 Hz Jy
(141)
18 Column Density Sensitivity
Looking at the redshift scaling relations as before, we
Taking the previous expression for column density in have for a changing configuration with fixed synthesised
terms of flux and beam size (Equation 71) gives the beam solid angle (with f (z) defined as before in Equa-
corresponding sensitivity relation of: tion 105):

16 Nhi (z) = f (z)(1 + z)4 Nhi (0) ,

Nhi = (1 + z)4 S , (134) (142)
3hhiAhi bm const const
Vrest Vrest
N (z) = f (z)(1 + z)7/2 N (0) . (143)
hi hi
 N 
hi = 2.64 1020 (1 + z)4 (135) For a fixed configuration with synthesised beam solid
cm2    angle increasing as (1 + z)2 :
S bm
.
JyHz arcsec2
Nhi (z) = f (z)(1 + z)2 Nhi (0) , (144)
V const V const
and substituting in Equation 99 for a theoretical sensi- N rest (z) = f (z)(1 + z)3/2 N rest (0) . (145)
hi hi
tivity estimate:
And finally for the widefield mosaic and phased array
feed cases:
16kTsys
Nhi = (1 + z)4 (136)
3hhi Ahi bm s Ae Widefield Single-Beam Mosaic:
s
2
,
Nhi (z) = f (z)(1 + z)
Nhi (0) , (146)
N (N 1)t
V const V const

N rest (z) = f (z)(1 + z)1/2
N rest (0) . (147)
hi hi
PASA (2017)
doi:10.1017/pas.2017.xxx
HI Beyond the Local Universe 11

Phased Array Feeds Observations: for the fact that source flux is distributed over multi-
ple beams, increasing the observed noise (Duffy et al.
12
Nhi (z) = f (z)FoVNE (z)
(148) 2012). Note that this correction factor only disappears
(1 + z) 2
Nhi (0) , if the source is a true point source, rather than just be-
ing smaller than the (synthesised) beam. The estima-
V const 1

N rest (z) = f (z)FoVNE2 (z) (149) tion of signal-to-noise using this correction factor also
hi
const
Vrest assumes the optimal extraction of the source signal in
(1 + z)3/2
N (0) . both spatial and frequency coordinates, which for low
hi
(150) signal-to-noise sources may not be possible.
For an extended source, a generally more appropri-
ate calculation is to determine the expected signal-to-
19 Signal-to-Noise noise from the desired column density sensitivity. For
The significance of a measurement can be determined instance, considering the case of single pointing with
by simply taking the ratio of the observed quantity beam bm at redshift z, the signal-to-noise in the cen-
to the measured noise in that quantity over the same tre of the primary beam will be given by (using Equa-
region. As such, the equations of the previous sections tion 140):
trivially lend themselves to predicting the likely out- Nhidesired
comes of observations. A few of the most important S/NNhi = Vrest
,
N
are detailed below. hi
Nhidesired
 
2
5.50 1023 (1 + z) 7
Peak signal-to-noise: cm2
 1   1 
Schan 1
  
bm Vrest 2 chan 2
.
S,peak arcsec2 km s1 Hz Jy
S/Npeak = . (151)
Schan (157)

Integrated signal-to-noise for a point source:

20 HI Fidelity Calculator
S As a complementary online tool, the HI Fidelity (HiFi)
S/Nint = , (152)
Schan chan calculator makes available many of the central for-
S mulae identified in this paper for ready application
= , (153) to the analysis of, or planning for, HI observations
chan chan Nchan
(http://hifi.icrar.org). These include the conver-
where the width of the profile is specified as either a fre- sion of observed frame quantities to rest frame equiv-
quency units () or the number of channels (Nchan ). It alents and vice versa, including: frequency width and
is worth
p noting in the above equations that the product velocity width, flux and mass, flux density and bright-
chan chan is independent of the particular choice of ness temperature, and flux and column density. Also
channel width used, and as such S/Nint is independent included are calculators for the estimation of observed
of channel choice. To predict the S/N for a point source noise and signal-to-noise, combining observed frame
of known mass and rest frame velocity width, the above measurement characteristics with rest frame source
can also be alternatively expressed: properties.
The calculator makes use of the Celestial R package3
for the calculation of cosmological distances. In addition
s
S c(1 + z)
S/Nint = , (154) to the non-evolving vacuum energy density calculation
Schan hi chan Vrest outlined in Section 4, this package, and the HiFi calcu-
(155) lator, also allow for an evolving vacuum energy model
through the use of the w and w parameters, replacing
which substituting for S using Equation. 45 gives:
in Equation 5 with:
 2
1 DL
S/Nint = 2.92 104 (1 + z) 2 (3+3w0 +6w )
Mpc 
1
e6w (1 1+z ) , (158)
1
   1   1 
Schan
1
Mhi Vrest 2 chan 2
1+z
.
M km s1 Hz Jy
where w0 = 1 and w = 0 return the default non-
(156)
evolving vacuum energy density.
If the source
p is extended, a further correction of
S/Nint 1/ (1 + Agalaxy /Abeam is required to allow 3 adsabs.harvard.edu/abs/2016ascl.soft02011R,cran.r-project.org/web/packa

PASA (2017)
doi:10.1017/pas.2017.xxx
12 Meyer et al.

Acknowledgments Peacock J. A., 1999, Cosmological Physics. Cambridge

University Press
Parts of this research were conducted by the Aus-
Schechter P., 1976, ApJ, 203, 297
tralian Research Council Centre of Excellence for All-
sky Astrophysics (CAASTRO), through project number Springob C. M., Haynes M. P., Giovanelli R., 2005, ApJ,
CE110001020. 621, 215
Staveley-Smith L., Oosterloo T., 2015, Advancing
Astrophysics with the Square Kilometre Array
(AASKA14), p. 167
REFERENCES Taylor G. B., Carilli C. L., Perley R. A., eds, 2008, Syn-
Abdalla F. B., Rawlings S., 2005, MNRAS, 360, 27 thesis Imaging in Radio Astronomy II Astronomical
Baldry I. K., et al., 2012, MNRAS, 421, 621 Society of the Pacific Conference Series Vol. 180
Blyth S., et al., 2015, Advancing Astrophysics with the Weinberg S., 2008, Cosmology. Oxford University Press
Square Kilometre Array (AASKA14), p. 128 Wilson T. L., Rohlfs K., H uttemeister S., 2009, Tools
Brinks E., 1990, in Thronson Jr. H. A., Shull J. M., eds, of Radio Astronomy. Springer-Verlag
Astrophysics and Space Science Library Vol. 161, The Zwaan M. A., et al., 2003, AJ, 125, 2842
Interstellar Medium in Galaxies. pp 3965 Zwaan M. A., Meyer M. J., Staveley-Smith L., Webster
Croton D. J., 2013, PASA, 30, e052 R. L., 2005, MNRAS, 359, L30
Davis T. M., Lineweaver C. H., 2001, in Durrer R.,
Garcia-Bellido J., Shaposhnikov M., eds, American
Institute of Physics Conference Series Vol. 555, Cos- Appendix A: litte h conversions
mology and Particle Physics. pp 348351
Davis T. M., Scrimgeour M. I., 2014, MNRAS, If you need to determine the value of some quantity for
442, 1117 a different Hubble constant, and the hC dependencies
Davis T. M., Lineweaver C. H., Webb J. K., 2003, Amer- are given, this can be done by simply calculating the
ican Journal of Physics, 71, 358 value of hC for your new Hubble constant and then
evaluating the expression given for the quantity. For
Duffy A. R., Meyer M. J., Staveley-Smith L., Bernyk
example, to calculate the value of an HI mass given as
M., Croton D. J., Koribalski B. S., Gerstmann D.,
Mhi = 109.7 h2
73 M for a Hubble constant of H0 = 100:
Westerlund S., 2012, MNRAS, 426, 3385
Einstein A., 1905, Annalen der Physik, 322, 891
Giovanelli R., Haynes M. P., 1988, Extragalactic neutral H0
h73 = , (159)
hydrogen. pp 522562 73 km s1 Mpc1
Giovanelli R., Haynes M. P., 2016, A&ARv, 24, 1 100 km s1 Mpc1
= , (160)
Hogg D. W., 1999, arXiv:9905116, 73 km s1 Mpc1
Holwerda B. W., Blyth S.-L., Baker A. J., 2012, in Tuffs = 1.37 . (161)
R. J., Popescu C. C., eds, IAU Symposium Vol. 284,
then,
The Spectral Energy Distribution of Galaxies - SED
2011. pp 496499
Jaffe Y. L., Poggianti B. M., Verheijen M. A. W., Mhi = 109.7 h2
73 M , (162)
Deshev B. Z., van Gorkom J. H., 2013, MNRAS, = 10 1.372 h2
9.7
100 M , (163)
431, 2111
= 109.4 h2
100 M . (164)
Kim H.-S., Wyithe J. S. B., Power C., Park J., Lagos
C. d. P., Baugh C. M., 2015, MNRAS, 453, 2315 Alternatively running this in reverse, and this time us-
Kim H.-S., Wyithe J. S. B., Baugh C. M., Lagos C. d. P., ing the abbreviation h = h100 :
Power C., Park J., 2016, preprint,
Kraus J. D., 1986, Radio Astronomy. Cygnus-Quasar Mhi = 109.4 h2 M , (165)
Martin A. M., Papastergis E., Giovanelli R., Haynes
= 10 0.732 h2
9.4
73 M , (166)
M. P., Springob C. M., Stierwalt S., 2010, ApJ, 9.7 2
723, 1359 = 10 h73 M . (167)
Obreschkow D., Heywood I., Klockner H.-R., Rawlings A final point to note is that the little h dependencies
S., 2009, ApJ, 702, 1321 for a quantity can be different, depending on how it
Obreschkow D., Heywood I., Rawlings S., 2011, ApJ, was determined. A classic example being the differences
743, 84 that often arise in little h exponents between observed
Obreschkow D., Glazebrook K., Kilborn V., Lutz K., and simulated quantities. The important thing to do
2016, ApJL, 824, L26 if you are wanting to compare two such quantities is
PASA (2017)
doi:10.1017/pas.2017.xxx
HI Beyond the Local Universe 13

to just make sure both values are valid for the same
Hubble constant, rather than worrying about the
exponents of hC , which are inherently different.

Appendix B: Symbols, Units, Constants &

Glossary
Units:

1 Jy = 126 W m2 Hz1 (168)

30
M = 1.98855 10 kg (169)
HI constants used in this work:
hi = 1.420405751786 109 Hz (170)
Ahi = 2.86888 1015 s1 (171)
mH = 1.673533 1027 kg (172)
Glossary of quantities used in this paper and their units:

= frequency
(Hz)
V = velocity
(km s1 )
S = Sobs = received flux density
(1026 W m2 Hz1 = Jy)
S = Sobs = received flux
(1026 W m2 = Jy Hz)
SV V
= Sobs = received velocity integrated flux
(1026 W m2 Hz1 km s1 = Jy km s1 )
L = Lrest = emitted luminosity density
(W Hz1 )
L = Lrest = emitted luminosity
(W)
I = Iobs = received specific intensity density
(W m2 sr1 Hz1 )
I = Iobs = received specific intensity
(W m2 sr1 )
B = Brest = emitted surface brightness density
(W m2 sr1 Hz1 )
B = Brest = emitted surface brightness
(W m2 sr1 )
Nhi = number of HI atoms
Nhi = column density of HI atoms (cm2 )
nhi = volume density of HI atoms (cm3 )

PASA (2017)
doi:10.1017/pas.2017.xxx