MNRAS 000, 1–11 () Preprint 2 August 2023 Compiled using MNRAS LATEX style file v3.

The morphology of the redshifted 21-cm signal from the Cosmic

Dawn

Raghunath
1
Ghara1,2★ , Satadru Bag3,4 , Saleem Zaroubi1,2 , Suman Majumdar5
Astrophysics Research Center of the Open University (ARCO), The Open University of Israel, 1 University Road, Ra’anana 4353701, Israel.
2 Kapteyn Astronomical Institute, University of Groningen, PO Box 800, 9700AV Groningen, The Netherlands.
3 Department of Physics, TUM School of Natural Sciences, Technical University of Munich, James-Franck-Straße 1, 85748 Garching, Germany
4 Max-Planck-Institut fur Astrophysik, Karl-Schwarzschild-Str. 1, 85748 Garching, Germany
5 Department of Astronomy, Astrophysics & Space Engineering, Indian Institute of Technology Indore, Indore 453552, India
arXiv:2308.00548v1 [astro-ph.CO] 1 Aug 2023

Accepted XXX. Received YYY; in original form ZZZ

ABSTRACT
The spatial fluctuations in the tomographic maps of the redshifted 21-cm signal from the
Cosmic Dawn (CD) crucially depend on the size and distribution of the regions with gas
temperatures larger than the radio background temperature. In this article, we study the mor-
phological characteristics of such emission regions and their absorption counterparts using
the shape diagnostic tool surfgen2. Using simulated CD brightness temperature cubes of the
21-cm signal, we find that the emission regions percolate at stages with the filling factor of the
emission regions 𝐹𝐹emi ≳ 0.15. Percolation of the absorption regions occurs for 𝐹𝐹abs ≳ 0.05.
The largest emission and absorption regions are topologically complex and highly filamentary
for most parts of the CD. The number density of these regions as a function of the volume
shows the power-law nature with the power-law indexes ≈ −2 and −1.6 for the emission and
absorption regions, respectively. Overall, the planarity, filamentarity and genus increase with
the increase of the volume of both emission and absorption regions.
Key words: radiative transfer - galaxies: formation - intergalactic medium - cosmology: theory
- dark ages, reionization, first stars - X-rays: galaxies

1 INTRODUCTION Observation of the redshifted 21-cm line radiation emitted

from the IGM’s neutral hydrogen (Hi ) atoms is the most promising
Our Universe became neutral during the epoch of recombination
probe of the physical processes during the CD and EoR (see e.g.,
about four hundred thousand years after the Big Bang. At a later
Madau et al. 1997; Furlanetto et al. 2006; Zaroubi 2013; Murmu
stage, radiation emitted from the first stars, galaxies, Quasars, High-
et al. 2022; Shaw et al. 2023; Bera et al. 2023). Many of the world’s
mass X-ray binaries (HMXBs), etc. heat up and ionize the cold and
largest radio telescopes have dedicated their valuable resources to
neutral gas in the inter-galactic medium (IGM). Theoretical studies
measuring the Hi 21-cm signal from these periods. One type of such
such as Pritchard & Furlanetto (2007); Mesinger et al. (2011); Ghara
observations uses radio interferometers such as the Low-Frequency
et al. (2016); Islam et al. (2019); Ross et al. (2019) suggest that the
Array (LOFAR)1 (Mertens et al. 2020), the Precision Array for
first X-ray sources such as HMXBs, and mini-Quasars heat up the
Probing the Epoch of Reionization (PAPER)2 (Kolopanis et al.
gas in the IGM during a period earlier than the stage when the
2019), the Murchison Widefield Array (MWA)3 (e.g. Wayth et al.
IGM became highly ionized. This period is known as the Cosmic
2018) and the Hydrogen Epoch of Reionization Array (HERA)4
Dawn (CD) of our Universe. The CD was succeeded by a period,
(DeBoer et al. 2017). Due to limited sensitivity, the ongoing radio
known as the Epoch of Reionization (EoR), in which the IGM’s
interferometer-based 21-cm observations of the CD and EoR aim to
primordial neutral gas became ionized by the UV radiation from the
measure the time evolution of the spatial fluctuations of the signal at
sources. The physical processes of these eras have major roles in
different scales (e.g., Gehlot et al. 2020; Mertens et al. 2020; Trott
shaping the current state of our Universes through various feedback
et al. 2020; The HERA Collaboration et al. 2022). The upcoming
mechanisms. Unfortunately, these epochs remain among the least
Square Kilometre Array (SKA)5 will have a higher sensitivity and
understood epochs of Cosmic history (e.g. Fan et al. 2006; Bañados
et al. 2018; Planck Collaboration et al. 2020; Mitra et al. 2015).
Many details about the timing, feedback mechanisms, morphology 1 http://www.lofar.org/
of the ionized and heated regions, etc. are still poorly known. 2 http://eor.berkeley.edu/
3 http://www.mwatelescope.org/
4 https://reionization.org/
★ E-mail: ghara.raghunath@gmail.com 5 http://www.skatelescope.org/

© The Authors
2 Ghara et al.
will further produce tomographic images of the distribution of the
Hi signal on the sky (Mellema et al. 2015; Ghara et al. 2017).
The second type of radio observations such as EDGES (Bowman &
Rogers 2010), EDGES2 (Monsalve et al. 2017; Bowman et al. 2018),
SARAS (Patra et al. 2015), SARAS2 (Singh et al. 2017), LEDA
(Price et al. 2018) use single radio antenna and aim to measure the
redshift evolution of the sky averaged Hi 21-cm signal from these
epochs.
The redshifted Hi 21-cm signal from the CD and EoR is en-
riched with information about the first sources of our Universe.
The measurements of the redshifted Hi signal from these epochs
are often used to infer the properties of the sources formed during
these epochs (e.g., Ghara et al. 2020; Greig et al. 2021; Mondal
et al. 2020; Abdurashidova et al. 2022), nature of dark matter (e.g.,
Barkana 2018; Fialkov et al. 2018; Muñoz & Loeb 2018; Nebrin
et al. 2019; Ghara & Mellema 2020; Ghara et al. 2022), etc. Further-
more, the same observations can be used to infer the properties of
the IGM such as the thermal and ionization states and morphology
of the 21-cm signal maps (e.g., Ghara et al. 2020; Ghara & Choud-
hury 2020; Ghara et al. 2021). The EoR 21-cm signal is expected
to be enriched with information about the ionized regions while the
CD 21-cm signal is expected to carry information about the charac-
teristics of the heated/emission regions. The heated regions around
the early X-ray emitting sources such as the HMXBs, mini-Quasars
are expected to appear as emissions in the Cosmic microwave back-
ground (CMB) subtracted 21-cm signal maps while the cold regions
will appear as absorption signal (see e.g., Ghara et al. 2016, 2017).
The radio interferometer-based observations probe multi-scale
properties of the signal and thus, are expected to contain more infor-
mation about the states of the IGM. The fluctuation in the observed
21-cm signal by the radio interferometers depends on the morphol-
ogy and distribution of the ionized/emission regions. Thus it is
important to study the morphology of the ionized/emission regions
of the signal to understand the connection between the measured
statistical quantities of the signal such as the power spectrum or the
sky-averaged signal. Recently several studies using methods such
as granulometry method (Kakiichi et al. 2017), Mean-free path
(Mesinger & Furlanetto 2007), image segmentation method (Giri
et al. 2018), percolation analyses (Iliev et al. 2006, 2014; Furlan-
etto & Oh 2016) and the Minkowski functionals (Friedrich et al.
2011; Yoshiura et al. 2017; Kapahtia et al. 2021), etc. have made
a significant effort in understanding the morphology of the ionized
and neutral regions during the EoR. These studies indicate that the
characteristics of the ionized/neutral regions e.g., the probability
density functions (PDF) of the bubble size distribution can be used
to distinguish different reionization scenarios and thus to constrain
the EoR source parameters, etc.
In this work, we thus focus on the morphological analysis of
the Hi 21-cm signal from the CD. Similar to the ionized/neutral
regions which are the focus of morphology studies of the 21-cm
signal from the EoR (e.g., Mesinger & Furlanetto 2007; Kakiichi
et al. 2017; Bag et al. 2018; Kapahtia et al. 2021), we consider the
emission/absorption regions and study the morphology and distri-
bution of these regions. Our study is based on simulated 21-cm
signal maps and SURFGEN2 algorithm (Bag et al. 2019, 2023)
which assesses the geometry, morphology and topology of a field
above/below certain thresholds. This algorithm models the surface
of a region through the Marchingcube 33 triangulation (Chernyaev
1987; Lorensen & Cline 1995). The accuracy in measuring the
Minkowski functionals and Shapefinders is much better compared to
the other existing methods, such as the cell counting based Crofton’s
formula (Crofton 1868). Previously, Bag et al. (2018, 2019); Pathak
et al. (2022); Dasgupta et al. (2023) studied the morphology of the
ionized regions during the EoR using the same algorithm. These
studies found that the time evolution of the Largest Cluster Statis-
tics (LCS) of the ionized regions is a robust means to study the
percolation of the ionized regions and can be used to distinguish
between different reionization scenarios. Here, we study the evolu-
tion of the emission regions LCS and study the percolation state for
the emission regions of the 21-cm signal during the CD.
This paper is structured in the following way. We first describe
the simulation used in this study in Section 2, then we discuss
the methodology to characterise the emission regions in Section
3. Section 4 shows the results from this study. We discuss and
conclude the findings of this study in Section 5. This study uses
cosmological parameters Ωm = 0.27, ΩΛ = 0.73, ΩB = 0.044,
ℎ = 0.7 (Wilkinson Microwave Anisotropy Probe (WMAP);
Hinshaw et al. 2013) which are the same as used in the N-body
simulation used in this study.
2 SIMULATION
In this section, we briefly present the simulation that leads towards
21-cm signal brightness temperature maps used for the morpholog-
ical study. Here, we use grizzly (Ghara et al. 2015, 2018) code
to simulate the redshifted 21-cm signal from the CD and EoR. The
algorithm is based on one-dimensional radiative transfer (RT) and is
an independent implementation of the previous 1D RT-based code
bears (Thomas et al. 2009; Krause et al. 2018).
The differential brightness temperature (𝛿𝑇b ) of the redshifted
Hi 21-cm signal emitted from a region with angular position x and
redshift 𝑧 can be expressed as (see e.g., Madau et al. 1997)
𝑇S (x, 𝑧) − 𝑇𝛾 (𝑧)
𝛿𝑇b (x, 𝑧) = [1 − exp{−𝜏𝑏 (x, 𝑧)}] , (1)
1+𝑧
where the Hi 21-cm optical depth 𝜏𝑏 can be denoted as (Bharadwaj
& Ali 2005; Datta et al. 2022),
4 mK
𝜏𝑏 (x, 𝑧) = 𝑥HI (x, 𝑧) [1 + 𝛿B (x, 𝑧)]
𝑇S (x, 𝑧)
  
ΩB ℎ2 0.7 𝐻0
× (1 + 𝑧) 3 . (2)
0.02 ℎ 𝐻 (𝑧)
Here, 𝑇𝛾 is the brightness temperature of the radio background. In
the absence of exotic sources of radio emission (e.g., see Mondal
et al. 2020; Ghara et al. 2022), 𝑇𝛾 is expected to be the CMB
brightness temperature 𝑇𝛾 (𝑧) = 2.73 ×(1 + 𝑧) K. The quantities 𝑥HI
and 𝛿B are the neutral fraction and the density contrast of hydrogen
in that region, respectively. The quantity 𝑇S stands for the spin
temperature of the neutral hydrogen which quantifies the relative
population of hydrogen in the two hyperfine states of the hydrogen
atom’s ground state.
The grizzly algorithm uses dark matter halo catalogues and
density and velocity fields on grids as derived from N-body simula-
tions. Given these inputs, grizzly generates 𝛿𝑇b maps for a source
model. Here, we use the outcome of a dark-matter-only N-body sim-
ulation from the results of the PRACE6 project PRACE4LOFAR.
The simulation box is of volume (500 ℎ −1 Mpc) 3 while the density
and velocity fields are smoothed into 3003 grids. The dark matter
halo catalogues contain all halos with masses larger than 109 M ⊙
6 Partnership for Advanced Computing in Europe: http://www.
prace-ri.eu/
MNRAS 000, 1–11 ()
 IGM during the Cosmic Dawn 3
1.0 − 500 TK (K)
xHII
fheat fheat = 0.05 fheat = 0.1 fheat = 0.3 104
0.8 400

Y (h−1 Mpc)
103
xHII, fheat

0.6 300

0.4 200 102

−

0.2 100
101
0.0 0
8 10 12 14 16 18 100 200 300 400 500 100 200 300 400 500 100 200 300 400 500
z X (h−1 Mpc) X (h−1 Mpc) X (h−1 Mpc)

Figure 1. Left panel: evolution of the volume fraction of the ionized region (i.e., volume averaged ionization fraction 𝑥¯ HII ) and heated regions (i.e., heated
fraction 𝑓heat ) of our simulated Cosmic Dawn (CD) and Epoch of Reionization (EoR) as functions of redshift. We define a ‘heated region’ as a region with
a gas temperature larger than the CMB temperature. 2D maps: Left to right maps show the gas temperature at three different stages of the CD with heated
fractions 0.05, 0.1 and 0.3, respectively. The corresponding redshifts are 14.7, 14.3 and 13.5, respectively. The colour bar shows the temperature in Kelvin.

500 50
fheat = 0.05 fheat = 0.3 fheat = 0.6 fheat = 0.9
400 0
Y (h−1 Mpc)

300 −50

−100
200

−150
100
−200
0 δTb(mK)
100 200 300 400 500 100 200 300 400 500 100 200 300 400 500 100 200 300 400 500
X (h−1 Mpc) X (h−1 Mpc) X (h−1 Mpc) X (h−1 Mpc)

Figure 2. Slices through the 𝛿𝑇b cubes at different stages of the Cosmic Dawn. From left to right panels correspond to redshifts 14.7, 13.5, 12.9 and 12.3 with
heated fractions 0.05, 0.3, 0.6, and 0.9, respectively. These maps correspond to 𝑇𝛾 = 𝑇CMB .

which is ≈ 25 times the dark matter particles resolution. The sim- be found in Ghara et al. (2015); Islam et al. (2019). These references
ulation saved the gridded fields and halo catalogues at different also provide the details of the method to simulate the coeval cubes
redshifts while the time gap between two successive redshifts is of neutral fraction 𝑥HI and 𝑇K at different redshifts.
11.4 mega-years. Details of the N-body simulation can be found in The solid line of the left panel of Figure 1 shows the redshift
Dixon et al. (2016); Giri et al. (2019b). Here we use 59 redshift evolution of the ionization fraction while the dashed curve shows
snapshots between 𝑧 ∼ 20 − 6.5 to simulate the CD and EoR for the the evolution of the heated fraction 𝑓heat i.e., the volume fraction
following source model. of regions with gas temperature larger than the CMB brightness
The astrophysical sources dependence in equation 1 comes temperature. The Figure shows that the IGM is significantly heated
through 𝑥HI and 𝑇S . Both these quantities depend on the emissivity by redshift ∼ 13 while the ionization fraction remains less than
of UV, X-ray, radio and Ly𝛼 photons, as well as the population 0.1. Figure 1 also shows slices of the gas temperature distribution
number density of the sources (see e.g., Ross et al. 2019; Reis et al. at three different stages of heating. From left to right 𝑇K maps
2020). The source model adopted in this study assumes that each of represent redshifts 14.7, 14.3 and 13.5 with 𝑓heat 0.05, 0.1 and 0.3,
the dark matter halos on the halo catalogues emits UV, X-ray and respectively. The slices show that significant overlap between heated
Ly𝛼 photons. The emission rates of these photons are proportional regions already started as early as a stage with redshift 14.3. The
to the stellar mass (𝑀★) of the sources which we assume to be Ly𝛼 photon flux fields at different redshifts are generated assuming a
 related to the dark matter halo mass 𝑀halo , i.e., 𝑀★ =
linearly 1/𝑅 2 fall of the number density of Ly𝛼 photons with radial distance
ΩB
𝑓★ Ω m
𝑀halo . We fix the star formation efficiency 𝑓★ = 0.02 in 𝑅 from the source. These Ly𝛼 photon flux fields are later used to
this study. The emission rate of the UV photons per stellar mass generate the 𝑇S fields. These 𝑥HI , 𝑇S cubes and the density fields
𝑁¤ 𝑖 is controlled by the Ionization efficiency parameter 𝜁 as 𝑁¤ 𝑖 = are then used to generate the 𝛿𝑇b maps using equation 1. Finally,
𝜁 × 2.85 × 1045 s −1 M −1⊙ . We use 𝜁 = 0.1 for which the EoR ends we incorporate the effect of the peculiar velocities of the gas or so-
around 𝑧 ∼ 6.5. The X-ray spectral energy distribution of the source called redshift-space distortion using a cell movement method (see
is assumed to be a power-law with energy 𝐸 as 𝐼 𝑋 (𝐸) ∝ 𝐸 − 𝛼 with e.g., Mao et al. 2012; Ross et al. 2021). Figure 2 shows the 𝛿𝑇b slices
spectral index 𝛼 = 1.2. The emission rate of the X-ray photons per at four different stages of the CD. The size of the heated/emission
stellar mass 𝑁¤ 𝑋 = 𝑓 𝑋 ×1042 s −1 M −1
⊙ where 𝑓 𝑋 is the X-ray heating regions (with 𝛿𝑇b ≥ 0) increases as heating progress. The panels
efficiency parameter. The X-ray band spans from 100 eV to 10 keV. also show that, above 𝑓heat ≳ 0.5, the distribution of the absorption
We choose 𝑓 𝑋 = 100 for the heating model considered in this study. regions (with 𝛿𝑇b < 0) is more relevant compared to the emission
The details of the spectral energy distribution and source model can regions.

MNRAS 000, 1–11 ()

4 Ghara et al.
500 1
fheat = 0.05 fheat = 0.1 fheat = 0.3 fheat = 0.4
400 0.8
Y (h−1 Mpc)

300 0.6

200 0.4

100 0.2

0 0
500 1
fheat = 0.6 fheat = 0.8 fheat = 0.9 fheat = 0.95
400 0.8
Y (h−1 Mpc)

300 0.6

200 0.4

100 0.2

0 0
100 200 300 400 500 100 200 300 400 500 100 200 300 400 500 100 200 300 400 500
−1 −1 −1 −1
X (h Mpc) X (h Mpc) X (h Mpc) X (h Mpc)

Figure 3. Slices through the binary cubes 𝛿𝑇b bin of the Hi 21-cm brightness temperature at different stages of the CD. From left-top to right-bottom panels
corresponds to redshifts 14.7, 14.3, 13.5, 13.2, 12.9, 12.6 12.3 and 12.0 with heated fractions 0.05, 0.1, 0.3, 0.4, 0.6, 0.8, 0.9 and 0.95, respectively. The binary
cubes 𝛿𝑇b bin (x, 𝑧) assume 𝛿𝑇b bin (x, 𝑧) = 1 if 𝛿𝑇b (x, 𝑧) ≥ 0, and zero otherwise. These maps correspond to 𝑇𝛾 = 𝑇CMB .

3 MORPHOLOGY ANALYSIS (i) volume: 𝑉,

(ii) surface area: 𝑆,
There exists no unique way to characterize the morphology and (iii) integrated mean curvature (IMC):
size distribution of a complex spatial structure such as the distribu- ∮
1
tion of 𝛿𝑇b emission regions in the IGM. Here we use surfgen2 𝐼 𝑀𝐶 = (𝜅 + 𝜅2 ) 𝑑𝑆 , (3)
2 𝑆 1
algorithm (Bag et al. 2019, 2023) to assess the morphology of in-
dividual heated regions in the 3D coeval 𝛿𝑇b boxes generated using (iv) integrated Gaussian curvature or Euler characteristic:
∮
grizzly. Given a coeval cube of the 𝛿𝑇b , we first generate a binary 1
𝜒= (𝜅 𝜅 )𝑑𝑆 . (4)
field 𝛿𝑇b bin (x, 𝑧) where 𝛿𝑇b bin (x, 𝑧) = 1 if 𝛿𝑇b (x, 𝑧) ≥ 0, and zero 2𝜋 𝑆 1 2
otherwise. These binary maps are then used in surfgen2 for the
Here 𝜅1 and 𝜅2 represent the two principle curvatures at the surface
morphological analysis of the emission regions. The code imple-
element 𝑑𝑆 on the surface 𝑆. The integrated Gaussian curvature 𝜒
ments periodic boundary conditions on the simulated 𝛿𝑇bbin (x, 𝑧)
is related to the genus (𝐺) of a closed surface as 𝐺 = 1 − 𝜒/2.
boxes and uses the friends-of-friends (FoF) algorithm to identify
Physically 𝐺 ≡ (no. of tunnels) − (no. of isolated surfaces) + 1.
individual emission and absorption regions. Figure 3 shows slices
Note that while 𝑉 and 𝑆 of the emission and absorption regions will
of such binary fields at different stages of our simulated CD as
always be positive, 𝐼 𝑀𝐶 and 𝐺 might also be negative in certain
described in the previous section. The white regions in the slices
cases.
represent emission regions for 𝑇𝛾 = 𝑇CMB . For 𝑓heat > 0.5, the size
To measure the shape of a region directly, we calculate the
distribution of the absorption regions is more meaningful than the
following three ratios of the Minkowski functionals,
emission regions. This is due to the inside-out nature of the heating
around the X-ray sources. Clearly, the individual emission regions 𝑇 = 3𝑉/𝑆 , 𝐵 = 𝑆/|𝐼 𝑀𝐶 | , 𝐿 = |𝐼 𝑀𝐶 |/(4𝜋). (5)
grow in size over time and overlap to create a large region that dom-
Note that each of 𝑇, 𝐵 and 𝐿 are of the dimension of length and are
inates the total emission segment volume, while the morphology of
spherically normalized, i.e. 𝑉 = (4𝜋/3)𝑇 𝐵𝐿. These ratios were first
the absorption regions towards the end of the CD appears different
introduced in Sahni et al. (1998) as ‘Shapefinders’ since they assess
in nature. Visually the morphology of these binary emission regions
the extent of a region in three dimensions. Here, 𝐿, 𝐵, 𝑇 represent
is similar to the morphology of ionized regions during an inside-out
the length, breadth and thickness of a region, respectively. A sphere
reionization scenario (see e.g., Bag et al. 2018; Pathak et al. 2022).
of radius 𝑅 will have 𝑇 = 𝐵 = 𝐿 = 𝑅. These quantities should
The calculation of the Minkowski functionals and thereafter satisfy natural ordering 𝐿 ≥ 𝐵 ≥ 𝑇. However, this ordering might
the Shapefinders of a connected (emission/absorption) region in not be maintained naturally in some rare cases. In those cases, we
surfgen2 is based on modelling the surface of that region through define the largest value among the three in Equation 5 as 𝐿 and the
the advanced Marching Cube 33 triangulation scheme (Chernyaev smallest value as 𝑇, to maintain the order 𝐿 ≥ 𝐵 ≥ 𝑇.
1987; Lorensen & Cline 1995). We refer the readers to Sheth et al. Further, the planarity (𝑃) and filamentarity (𝐹) of a region
(2003); Bag et al. (2019) papers for the details of the surfgen2 are defined as the following dimensionless quantities based on the
algorithm. The algorithm first determines the four basic Minkowski Shapefinders (see Sahni et al. 1998),
functionals (Mecke et al. 1994) that well describe the morphology 𝐵−𝑇 𝐿−𝐵
of an individual region. These are 𝑃= , 𝐹= . (6)
𝐵+𝑇 𝐿+𝐵

MNRAS 000, 1–11 ()

 IGM during the Cosmic Dawn 5
−2 1 1
10

−3
10 0.8 0.8

FFemi=0.15

FFemi=0.95
10−4
−3

0.6
Mpc)

0.6
10−5
FF

LCS
−1
Φ (h

0.4
fT=1
10−6 fT=3
0.4
fT=6 fT=1
fT=10 0.2 fT=3
10−7 fT=6
Absorption 0.2
Emission fT=10
−8 0 Emission
10
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Absorption
fheat fheat
0
0 0.2 0.4 0.6 0.8 10 0.2 0.4 0.6 0.8 1
fheat FFemi

Figure 4. Evolution of the number density of emission and absorption re-

gions (left panel) and their filling factors 𝐹𝐹 (right panel) as functions of
the heated fraction, i.e., the volume fraction of regions with gas tempera-
ture larger than 𝑇CMB . Different types of lines represent different threshold
parameter ( 𝑓𝑇 ) values for the radio background temperature as chosen in
this study. The black circular dots mark different stages of the CD where
the morphology of the emission regions has been estimated. The emission
and absorption regions at different stages of the CD are determined using
surfgen2 algorithm.
Figure 5. Evolution of the Largest cluster statistics (LCS) as a function of
the heated fraction (left panel) and filling factor (right panel) of the emission
regions. The dashed curves represent emission regions while the solid curves
represent the absorption regions of the 21-cm signal 𝛿𝑇b coeval cube. The
largest emission and absorption regions at different redshifts of the CD are
determined using surfgen2 algorithm.

As their names suggest, they measure the degree of planarity and

filamentarity of a region. By definition, 𝑃 and 𝐹 values are bounded
by 0 and 1. Note that, for a spherical object, 𝑇 = 𝐵 = 𝐿 leads to
𝑃 = 0 = 𝐹. For a planar object, such as a sheet, 𝐿 ∼ 𝐵 ≫ 𝑇 results
in 𝑃 ≫ 𝐹. On the other hand, 𝑇 ∼ 𝐵 ≪ 𝐿 for a filament and thus
𝐹 ≫ 𝑃.
In this work, we employ the above-mentioned Minkowski func-
tionals and Shapefinders to statistically study the morphology of the
emission and absorption regions of our simulated CD 21-cm signal.
In addition, we consider the morphological properties of the largest
emission and absorption regions which provides crucial information
about the percolation state of these regions.
4 RESULTS
In this paper, we explore the morphological distribution of the emis-
sion and absorption regions of the 21-cm signal during the CD. We
consider simulated 𝛿𝑇b coeval maps at different redshifts generated
from grizzly code (see section 2). Unlike the ionization maps as
considered previously in studies like Bag et al. (2018); Pathak et al.
(2022), the fall of the signal amplitude at the boundary of the emis-
sion regions where 𝛿𝑇b = 0 is not very sharp. Furthermore, the
nature of the signal, either for emission or absorption, depends cru-
cially on the radio background temperature 𝑇𝛾 which is uncertain
at the frequencies of our interest (e.g., Fixsen et al. 2011). Thus,
we choose a number of different thresholds 𝛿𝑇b,thre on the 𝛿𝑇b (x, 𝑧)
maps to study the morphology of the regions with 𝛿𝑇b larger than
𝛿𝑇b,thre which we refer to as emission segment. In addition, we
also consider the complementary regions, i.e. the absorption re-
gions with 𝛿𝑇b < 𝛿𝑇b,thre . We have seen from Figures 2 and 3
that the distribution of these absorption regions is more relevant
for 𝑓heat ≳ 0.5. Different 𝛿𝑇b,thre values also represent different
strengths of radio background at 1420 MHz at the redshift of our
interest. We parameterize the 𝛿𝑇b,thre in terms of radio background
temperature 𝑇𝛾 = 𝑓𝑇 × 2.725(1 + 𝑧) K with the threshold parameter
𝑓𝑇 .
The number density and size of the emission regions vary with
the radio background temperature 𝑇𝛾 and thus also with threshold
parameter 𝑓𝑇 . The left panel of Figure 4 shows the evolution of
the number density (Φ) of isolated emission (dashed curves) and
absorption (solid curves) regions with 𝑓heat for four different 𝑓𝑇
values. Φ of the emission regions first increases with 𝑓heat until
𝑓heat reaches ∼ 0.1 due to the formation of new heated regions.
For 𝑓heat ≳ 0.1, Φ of the emission regions gradually falls before
sharply falling to 1 when 𝑓heat → 1. This happens due to overlaps
between individual emission regions. Both the maximum value of
emission Φ and the corresponding 𝑓heat (at which the maximum
occurs) remain smaller for a smaller value of 𝑓𝑇 . This is expected
as by increasing 𝑓𝑇 , a large emission region at a later stage of
heating can break into multiple smaller emission regions. The fall
of Φ for emission regions is faster for a smaller value of 𝑓𝑇 . In fact,
Φ changes very little in the range 0.1 ≲ 𝑓heat ≲ 0.9 for 𝑓𝑇 = 10.
This implies that the overlap between the emissions regions happens
at a later stage for a larger 𝑓𝑇 value.
The number density of absorption regions, on the other hand,
gradually increases with 𝑓heat since larger absorption regions break
into smaller ones as heating progresses. Contrary to the Φ of the
emission regions, the Φ of the absorption regions (solid curves
of Figure 4) decreases with the increase of 𝑓𝑇 . However, Φ of
absorption regions sharply drops to zero at the end of the heating
of the IGM, as expected.
The filling factor (which is defined as the volume fraction of
emission or absorption segment), 𝐹𝐹, of emission or absorption
regions also depends on 𝑓𝑇 . The right panel of Figure 4 shows how
𝐹𝐹s of the emission (dashed) and absorption (solid) regions evolve
with 𝑓heat for different 𝑓𝑇 values. As expected, 𝐹𝐹 of the emission
regions (𝐹𝐹emi ) is roughly the same as the 𝑓heat for 𝑓𝑇 = 1. For a
certain 𝑓heat , 𝐹𝐹emi decreases for a larger value of 𝑓𝑇 . On contrary
to 𝐹𝐹emi , 𝐹𝐹 of the absorption regions (𝐹𝐹abs ) decreases as heating
progresses since 𝐹𝐹abs = 1 − 𝐹𝐹emi .

MNRAS 000, 1–11 ()

6 Ghara et al.
1 108 106 106
0.9 7 5 105
10 10
0.8

V [d3], S [d2], IMC [d=h−1 Mpc]

104
0.7 106 104
103

T, B, L (h−1 Mpc)
0.6 5 3
FFHII=0.15

10 10 102

P, F, G
T P
LCS

0.5 B F
104
10 2 L 101 G
0.4
100
0.3 103
101
−1
0.2 10
102 V 100
0.1 A 10−2
xHII 1 IMC −1
0 10 10 10−3
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
FFHII FFHII FFHII FFHII

Figure 6. Left: evolution of the Largest cluster statistics (LCS) of the ionized regions as a function of the filling factor of the ionized regions. Evolution of
different shape-defining quantities of the largest cluster of ionized regions in the IGM as functions of the filling factor of the ionized regions. The left Panel
shows volume (𝑉), surface area (𝑆) and integrated mean curvature (𝐼 𝑀𝐶). The middle panel shows the thickness (𝑇), breadth (𝐵) and length (𝐿) of the largest
ionized region. The right panel shows the planarity (𝑃), filamentarity (𝐹) and genus (𝐺) of the largest ionized region.

4.1 Largest cluster statistics 𝑓heat ∼ 1. This suggests that the absorption segment in the IGM
remains percolated for a large extent of the CD and only breaks
The ‘largest cluster statistics’ (LCS) for a segment (emission or
into isolated absorption regions towards the end period of the CD.
absorption) is defined as the volume fraction enclosed by the largest
When plotted against 𝐹𝐹, the LCS of the absorption regions shows
region of that segment,
a sharp decrease around 𝐹𝐹emi ∼ 0.95 for all the thresholds cho-
volume of the largest emission/absorption region sen. The evolution of the LCS of the emission regions with 𝐹𝐹emi is
LCS = . (7) almost identical for the different choices of 𝑓𝑇 . The sharp increase
total volume of all the emission/absorption regions
of LCS of the emission regions at 𝐹𝐹emi ∼ 0.15 is similar to the
It has been established that the LCS carries robust information about percolation case of the ionized regions. The percolation of the ion-
the percolation process in different fields at cosmological scales, e.g. ized regions occurs at a stage with 𝑥 HII ∼ 0.1 − 0.15 (see e.g., Iliev
the large-scale matter distribution in the Universe (Shandarin 1983; et al. 2006; Chardin et al. 2012; Furlanetto & Oh 2016; Bag et al.
Klypin & Shandarin 1993; Yess & Shandarin 1996) and the neutral 2018). To demonstrate the same with grizzly simulation outputs,
hydrogen field in the Epoch of Reionization (EoR) (Iliev et al. 2006, we present the LCS as a function of the filling factor of the ionized
2014; Furlanetto & Oh 2016; Bag et al. 2018). For the ionization regions (𝐹𝐹HII ) in the left panel of Figure 6. The ionization history
bubbles, the percolation happens early with volume averaged ion- is presented in the left panel of Figure 1.
ization fraction 𝑥 HII ∼ 0.1 (see e.g., Furlanetto & Oh 2016; Bag
et al. 2018). Here, we study the characteristics of the largest emis-
sion and absorption regions in the IGM at different stages of CD
4.2 Evolution of the largest emission and absorption regions
and the percolation transitions in the respective segment.
In the left panel of Figure 5, we present the evolution of LCS Next, we study the evolution of shape and morphology of the largest
of the emission and absorption regions as a function of 𝑓heat for the region, separately for the emission and absorption segments, in
four different choices of 𝑓𝑇 . At the initial stages of the heating (e.g., terms of Shapefinders. The left panel of Figure 7 shows the evolu-
with 𝐹𝐹emi ≲ 0.1), the emission regions are fairly isolated with tion of the first three Minkowski functionals 𝑉, 𝑆 and 𝐼 𝑀𝐶 of the
nominal overlaps with each other. As heating progresses, many new largest emission (dashed) and absorption (solid) regions as func-
emission regions appear around the X-ray emitting sources and the tions of 𝐹𝐹emi for 𝑓𝑇 = 1. All these three quantities evolve rapidly
emission regions themselves grow in size resulting in a gradual in- around the percolation transition of the corresponding segment (at
crease in the LCS of the emission regions with 𝑓heat (see the dashed 𝐹𝐹emi ≈ 0.15 and 0.95 for the emission and absorption segments
curves in the left panel of Figure 5). Soon the emission regions respectively). In fact, this panel demonstrates how the largest emis-
start to merge significantly. Due to the vigorous merging of smaller sion/absorption region grows rapidly at the onset of their respective
regions, an extremely large connected emission region that extends percolation transitions and this is manifested in the evolution of the
throughout the IGM appears abruptly at this stage. In our case, it LCS in Figure 5. Also note that the evolution of 𝑉, 𝑆 and 𝐼 𝑀𝐶
stretches throughout the simulation box. We refer to this period as of the largest emission region around the percolation transition is
the ‘percolation transition’ in the emission regions. The LCS of the similar to that of the same quantities of the largest ionized region in
emission regions increases sharply during the percolation transition the ionization field (see the second panel from the left in Figure 6;
since the largest region abruptly accumulates most of the emission see also Bag et al. (2018)).
segment volume. The critical 𝑓heat at which percolation transition Let us now focus on the largest emission region. As expected,
takes place depends on 𝑓𝑇 values. However, percolation transition its 𝑉 always increases with the increase of 𝐹𝐹emi . Its 𝑆, on the
roughly occurs at similar values of the filling factor 𝐹𝐹emi ∼ 0.15 other hand, increases with 𝐹𝐹emi until 𝐹𝐹emi ∼ 0.6 and slowly
(marked by the vertical dotted line in the right panel of Figure 5) for decrease thereafter. This shows at stages with 𝐹𝐹emi ≳ 0.6, the
all the threshold values chosen (see the dashed curves of the right largest emission region in the simulation box becomes less complex
panel of Figure 5). in shape with less amount of absorption regions surface inside it,
The LCS of the absorption regions, on the other hand, re- although its volume increases with time. While 𝑉 and 𝑆 are always
mained ≈ 1 for a long time before sharply dropping to zero around positive quantities, 𝐼 𝑀𝐶 can be negative also. The change in the

MNRAS 000, 1–11 ()

 IGM during the Cosmic Dawn 7

108 106 105

T
V [d3], S [d2], IMC [d=h−1 Mpc]

107 B 104
105 L
6
103

T, B, L (h−1 Mpc)
10
104
105 102

P, F, G
P
103 F
104 101 G
102
103 V 100
2 S 101 −1
10 IMC 10
1 V−Emission 0
10 10 10−2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
FFemi FFemi FFemi

Figure 7. Evolution of different shape defining quantities of the largest region of the emission (dashed) and absorption (solid) segments in the IGM as functions
of filling factor of the emission regions correspond to 𝑓𝑇 = 1. The x-axis is equivalent to the heated fraction of the simulation. The left panel shows the volume
(𝑉), surface area (𝑆) and integrated mean curvature (𝐼 𝑀𝐶) of the largest emission and absorption region. The middle panel shows the thickness (𝑇), breadth
(𝐵) and length (𝐿) of the largest regions, while the rightmost panel shows their planarity (𝑃), filamentarity (𝐹) and genus (𝐺).

10−4 10−4 emission region negative 7 . The evolution of the largest absorption
FFabs=0.05 V−1.6
0.1
Absorption regions

10−6 region exhibits similar behaviour but in the opposite direction along
10−5 0.2
0.3 the 𝐹𝐹emi axis.
dN/dV (h Mpc−1)6
N(V) (h Mpc−1)3

0.6 10−8 The middle panel of Figure 7 shows the evolution of the three
10−6
10−10 Shapefinders – ‘thickness’ (𝑇), ‘breadth’ (𝐵) and ‘length’ (𝐿) – of the
10−7 largest emission and absorption regions as functions of 𝐹𝐹emi , again
10−12
for 𝑓𝑇 = 1. The ‘length’ (𝐿) of the largest emission or absorption
−8
10 10−14 region increases rapidly at the onset of the percolation transition
(in the respective segment, i.e. 𝐹𝐹emi ≈ 0.15 and 0.95 for the
10−9 10−16
101 102 103 104 105 106 107 101 102 103 104 105 106 107 emission and absorption segment respectively) as compared to its
V (h−1 Mpc)3 V (h−1 Mpc)3 𝐵 and 𝑇.This is also similar to the evolution of 𝑇, 𝐵 and 𝐿 of the
10−3 10−4 largest ionized region in our simulation box as functions of 𝐹𝐹HII ,
FFemi=0.05 V−2.
0.1 as shown in the second right panel of Figure 6; see Fig 6 of Bag
10−4 0.2 10−6
Emission regions

0.3 et al. (2018) (also Pathak et al. 2022; Dasgupta et al. 2023). The
dN/dV (h Mpc−1)6
N(V) (h Mpc−1)3

10−5 0.4
10 −6
10−7
10−8
10−9
101 102 103 104 105 106 107
V (h−1 Mpc)3
Figure 8. Left: Cumulative number of emission (bottom panel) and ab-
sorption (top panel) regions as a function of the volume at different stages
of the Cosmic Dawn. Right: The size distribution of the emission (bottom
panel) and absorption (top panel) regions as a function of volume at the
same stages of Cosmic Dawn. Different curves in the bottom panels are for
𝐹𝐹emi = 0.05, 0.1, 0.2, 0.3 and 0.4 which corresponds to redshifts 14.7,
14.3, 13.9, 13.6 and 13.2, respectively. On the other hand, different curves in
the top panels are for 𝐹𝐹abs = 0.05, 0.1, 0.2, 0.3 and 0.6 which corresponds
to redshifts 12, 12.3, 12.6, 12.9 and 13.2, respectively. The list of emission
and absorption regions is generated from a simulated grizzly Cosmic Dawn
scenario using surfgen2 algorithm.
sign of the 𝐼 𝑀𝐶 occurs around 𝐹𝐹emi ∼ 0.65 for this model CD.
The negative part of the 𝐼 𝑀𝐶 is shown in the light-blue colour in
the left panel of Figure 7. For 𝐹𝐹emi ≳ 0.7, one can think of the
IGM as a distribution of several absorption regions in an emission
region background. At these stages, the sum of 𝐼 𝑀𝐶s of all the
absorption regions is positive which makes the 𝐼 𝑀𝐶 of the largest
10−8 evolution of 𝐿 of the largest emission region merely follows the
10−10
evolution of its |𝐼 𝑀𝐶 | (see equation 5).
The right panel of Figure 7 shows the evolution of morpho-
10−12 logical quantities, 𝑃 and 𝐹, and the topology, 𝐺, of the largest
10−14 emission (dashed curves) and absorption (solid curves) regions. 𝐿
of the largest emission region is more than 2 orders of magnitude
10−16
101 102 103 104 105 106 107 larger than its 𝐵 around the percolation transition stage. This leads
V (h−1 Mpc)3 to the largest emission region being highly filamentary with 𝐹 ∼ 1
(see equation 6) at percolation. The same is also applicable to the
largest absorption region. These findings also agree well with the
filamentary nature of the largest ionized region during ionization
percolation (see right panel of Figure 6). Also, the genus values
of both the largest emission and absorption regions increase at the
onset of the percolation transition in the respective segments (see
the blue curves), and remain stable at post percolation stages. This
implies that the largest regions become more multiply connected
with complex topology as they grow and percolate. Note that, at
some stage, well after percolation, the genus value of the largest
region (emission or absorption) abruptly becomes negative. In this
phase where the genus is negative, we show its absolute value by the
light-blue coloured curves (solid and dashed). The sign flip, which
indicates the change in the topology of the largest region, manifests
in negative IMC values (see left panel of Figure 7) which in turn
lead to the sharp dip/rise in different Shapefinders (as evident from
the middle panel and explained above) at this stage. Indeed, well af-
7 We also see a similar sign change for the 𝐼 𝑀𝐶 of the ionized regions as
shown in the second left panel of Figure 6.

MNRAS 000, 1–11 ()

8 Ghara et al.

Emission Absorption the top right panel of Figure 8). As the smaller emission regions
0.6 0.6 overlap and form larger emission regions, the values of 𝑑𝑁/𝑑𝑉
FFemi=0.05 FFabs=0.05
0.1 0.1
of the emission regions at the smaller 𝑉 decreases as the heating
0.5 0.2 0.5 0.2 progress. Roughly, 𝑑𝑁/𝑑𝑉 of the emission regions shows a power-
0.3 0.3

Mean free path

0.4 0.4 0.4 0.6 law dependence on the volume, i.e., 𝑑𝑁/𝑑𝑉 (𝑉) ∝ 𝑉 𝜏 with 𝜏 ≈ −2.
The same is also true for the absorption regions while 𝜏 ≈ −1.6 for
Rdn/dR

0.3 Rdn/dR 0.3 that. The curves show that a single power-law dependence of N (𝑉)
0.2 0.2
become loose at the lower 𝑉 side. This shows that the dependencies
of the 𝑑𝑁/𝑑𝑉 on 𝑉 are not presentable by a single power law
0.1 0.1 and there exit characteristics sizes of the emission (also for the
absorption regions) which in this case is 𝑉 ∼ 20 − 30 ℎ −1 Mpc.
0.0 0.0
10 100 10 100 As there is no unique method to estimate the size distribution
R (h−1 Mpc) R (h−1 Mpc) of the ionized/emission regions, we also consider two other methods
namely the mean free path (MFP) method (Mesinger & Furlanetto
Emission Absorption 2007) and the granulometry method (Kakiichi et al. 2017) to ob-
0.8 1.0 tain the side distributions of the emission and absorption regions
FFemi=0.05 FFabs=0.05
0.7 0.1 0.1 and compare the results. Note that the interpretation of the size
0.2 0.8 0.2
0.6 0.3 0.3 distribution varies with the methodology.
0.4 0.6
Granulometry

For the estimation of the size distribution of these regions using

RdF(<R)/dR

RdF(<R)/dR

0.5 0.6
the MFP method, we shoot 107 rays around randomly selected points
0.4
inside the emission (or absorption) regions in random directions
0.3 0.4
and record the lengths of the rays until those reach the edge of the
0.2 regions. The top panels of Figure 9 show the probability density
0.2
0.1 function (PDF) 𝑅𝑑𝑛/𝑑𝑅 of the recorded ray lengths at different
0.0 0.0 stages of the CD. Here, 𝑛 is the number of rays whose sizes are
0 5 10 15 20 0 2 4 6 8 10 12 14 between 𝑅 and 𝑅 + 𝑑𝑅. The top left panel show the PDFs for the
R (h−1 Mpc) R (h−1 Mpc) emission regions while the top right panel represents the absorption
regions. The overlap of the smaller emission regions at a later stage
of heating is prominent by the characteristic scales or the peaks of
Figure 9. Size distribution of the emission (left panels) and absorption (right the PDFs. The peak of the PDFs of the emission regions moves
panels) regions during the CD. The size distributions in the top panels are toward the larger values as the heating progresses. The PDFs show
estimated using the mean-free path method, while the bottom panels are for log-normal type behaviour which is similar to the distribution of the
the granulometry method. Here, 𝑛 is the number of rays used in the MFP
ionized regions as well (see e.g., Mesinger & Furlanetto 2007). The
method whose sizes are between 𝑅 and 𝑅 + 𝑑𝑅. On the other hand, the
quantity 𝐹 (< 𝑅) from the granulometry method is the volume fraction of
size distribution of the absorption regions also shows log-normal
the emission (or absorption) regions whose size is smaller than a radius 𝑅. type nature and is similar to the distribution of the neutral regions
during the EoR (see e.g., Giri et al. 2019a).
The bottom panels of Figure 9 show the PDFs 𝑅𝑑𝐹 (< 𝑅)/𝑑𝑅
of the size distribution as obtained by using a granulometry method.
ter percolation in the emission (absorption) segment, many isolated
Here, the quantity 𝐹 (< 𝑅) is the volume fraction of the emission
small pockets of absorption (emission) regions emerge inside the
(or absorption) regions whose size is smaller than a radius 𝑅. We
percolated largest emission (absorption) region that brings down
follow the method of Kakiichi et al. (2017) which is based on
the latter’s genus value and eventually rendering it negative.
Minkowski subtraction and addition steps to estimate the PDFs.
Note that, this method is based on applying spherical filters, and
4.3 Size Distribution is sensitive to the smallest dimension of a complex-shaped region.
Thus, the algorithm does not find any structure with a radius as large
The right panels of Figure 8 show the distribution of the number as the largest ray length found in the MFP method. In our case, e.g.,
density 𝑑𝑁/𝑑𝑉 of absorption (top panel) and emission (bottom for 𝐹𝐹 ≲ 0.6 the granulometry method does not find any spherical
panel) regions as functions of volume at different stages of the CD. structure of the emission or absorption regions with a radius more
The left panels of Figure 8 show the cumulative number density
∫ ∞ 𝑑𝑁 than ∼ 20 ℎ −1 Mpc.
of the regions N (𝑉) ≡ 𝑉 𝑑𝑉 ′
′ 𝑑𝑉 . The individual regions in

these distributions are identified in the surfgen algorithm using

𝑓𝑇 = 1. It should be realized that the size distribution of the emission
regions is more meaningful when 𝐹𝐹emi is not very large. We show
the distribution of the emission regions for 𝐹𝐹emi ≲ 0.5. For the
other half range, i.e., 𝐹𝐹emi ≳ 0.5, we show the distribution of the
absorption regions which are more relevant at those stages of the
CD.
Similar to the distribution of the ionized region (see e.g., Bag
et al. 2018), the size distribution of the emission regions also shows
a fall of 𝑑𝑁/𝑑𝑉 towards larger sizes. After the percolation stages, the
distribution of the emission regions becomes bimodal. The bimodal
nature of 𝑑𝑁/𝑑𝑉 is also visible for the absorption regions (see
4.4 Shape distribution of emission and absorption regions
In this subsection, we study how the shape (together with topol-
ogy and morphology) of the emission and absorption regions is
distributed over their volume and how these distributions evolve
over time. The top panels of Figure 10 show the distributions of the
𝑇, 𝐵 and 𝐿 of the emission (dashed curves) and absorption (solid)
regions as functions of their volume at different states of the CD.
For both types of regions, 𝐿 is significantly larger compared to their
𝑇 and 𝐵 while the difference increases with the increase of vol-
ume of the individual regions. E.g., for 𝐹𝐹 ∼ 0.1, 𝐿/𝑇 or 𝐿/𝐵 is
∼ 1 − 2 at 𝑉 ∼ 102 (ℎ −1 Mpc) 3 , while the ratios increases to ∼ 20

MNRAS 000, 1–11 ()

 IGM during the Cosmic Dawn 9

FF=0.01 FF=0.05 FF=0.1 FF=0.3

102 102 102 102
T−emi
B−emi
Shapefinders (h−1 Mpc)

L−emi
T−abs
B−abs
L−abs
101 101 101 101

100 100 100 100

102 103 104 102 103 104 102 103 104 102 103 104
−1
Mpc)3 −1
Mpc)3 −1
V (h V (h V (h Mpc)3 V (h−1 Mpc)3

FF=0.01 FF=0.05 FF=0.1 FF=0.3

102 102 102 102
Shapefinders (h−1 Mpc)

101 101 101 101

100 100 100 100

10−1 P−emi 10−1 10−1 10−1

F−emi
G−emi
10−2 P−abs 10−2 10−2 10−2
F−abs
G−abs
10−3 10−3 10−3 10−3
102 103 104 102 103 104 102 103 104 102 103 104
V (h−1 Mpc)3 V (h−1 Mpc)3 V (h−1 Mpc)3 V (h−1 Mpc)3

Figure 10. The top panels show the Shapefinders ‘length’ (𝐿), ‘breadth’ (𝐵) and ‘thickness’ (𝑇) of the emission (dashed) and the absorption (solid) regions
functions of volume (𝑉) for the threshold parameter 𝑓𝑇 = 1 at different stages of the CD. The bottom panels show the ‘planarity’ (𝑃), ‘filamentarity’ (𝐹) and
‘genus’ (𝐺) of the emission and the absorption regions as functions of 𝑉 for 𝑓𝑇 = 1 at different stages of the CD. 𝐹𝐹 stands for the filling factor corresponding
to the type of region, i.e. 𝐹𝐹 stands for 𝐹𝐹emi (𝐹𝐹abs ) when we consider the emission (absorption) regions.

for 𝑉 ∼ 104 (ℎ −1 Mpc) 3 . This implies that small-volume emis-
sion/absorption regions are more spherical while the larger regions
are more filament-type. The same conclusions can be made from
the bottom panels of Figure 10 as well which show the distribu-
tion of 𝑃, 𝐹 and 𝐺 of the emission (dashed curves) and absorption
(solid) regions as functions of their volume for different 𝐹𝐹. While
𝐹 of the emission regions is ∼ 0.2 for 𝑉 ∼ 102 (ℎ −1 Mpc) 3 , it in-
creases to ∼ 1 at 𝑉 ∼ 2 × 103 (ℎ −1 Mpc) 3 . For the same range of 𝑉,
the change of 𝑃 is smaller compared to that of 𝐹. From the genus
distribution (blue dashed and solid curves) it is evident that larger
emission/absorption regions tend to be more multiply connected
with more complex topology. Strikingly, the shape (and morphol-
ogy) distributions of the absorption regions remain similar to that
of the emission regions when both segments are compared at the
same filling factor.
5 SUMMARY & DISCUSSION
The spatial fluctuation of the redshifted Hi 21-cm signal from the
400 million years to the first billion years of our Universe’s history
carries information about the first radiating sources, the ionization
and thermal states of the IGM. The observations of this signal from
the Epoch of Reionization (EoR) and the Cosmic Dawn (CD) using
the world’s largest radio interferometers have the ability to reveal
many unknown facts about these epochs. Theoretical studies suggest
that the spatial fluctuations in the signal from the Cosmic Dawn
depend on the distribution and properties of the heating sources.
An alternative interest focuses on the distribution of the emission
regions around the heating sources. Such a morphological study is
important to understand the size and distribution of such emission
regions or the remaining part of the sky which is seen in absorption.
In this study, we focus on the shape and morphology of the
emission and absorption regions during the Cosmic Dawn. This
study is based on a simulation of the Cosmic Dawn 21-cm sig-
nal using grizzly (Ghara et al. 2015) code and analyzing it using
surfgen2 algorithm (Bag et al. 2019) to produce the geometrical
properties of the regions. The main findings of our study are the
following.
• If the radio background is the CMB, the number density of the
emission regions gets a maximum at a stage when the volume frac-
tion of regions with gas temperature larger than the CMB tempera-
ture (heated fraction 𝑓heat ) is ∼ 0.1. In case the radio background is
stronger than CMB, the peak of the number density vs 𝑓heat shifts
towards a larger 𝑓heat value and can occur between 𝑓heat = 0.1
and 0.9 depending on the radio background strength. The peak of
the number density of the absorption regions as a function of 𝑓heat
occurs at a stage when 𝑓heat ≳ 0.9.
• The percolation transition in the emission regions roughly oc-
curs at a filling factor (defined as the volume of the emission seg-
ments) of 𝐹𝐹emi ≈∼ 0.15. For the absorption regions, the perco-
lation transition occurs at the filling factor 𝐹𝐹abs ≈ 0.05 which
corresponds to 𝐹𝐹emi ≈∼ 0.95. The evolution of the ‘largest cluster
statistics’ LCS (defined as the ratio of the largest region volume and
the total volume of all regions with the same type) of the emission,
as well as the absorption regions, as a function of the filling fac-
tor, are roughly identical for the different radio background levels
chosen in this study.
• The overall evolution of the Minkowski functionals – volume
(𝑉), surface area (𝑆), integrated mean curvature (𝐼 𝑀𝐶) and genus

MNRAS 000, 1–11 ()

10 Ghara et al.
(𝐺) – and Shapefinders – thickness (𝑇), breadth (𝐵), length (𝐿),
planarity (𝑃) and filamentarity (𝐹) – of the largest emission region
(when studied functions of the filling factor of the emission regions)
during the CD is found to be similar to the evolution of the same
characteristics of the largest ionized region (when studied against
ionized filling factor) during the EoR.
• The number density of emission regions 𝑑𝑁/𝑑𝑉 as a function
of their volume 𝑉, estimated from the surfgen2 algorithm, shows a
rough power-law dependence on the volume, i.e., 𝑑𝑁/𝑑𝑉 (𝑉) ∝ 𝑉 𝜏
with 𝜏 ≈ −2. We also see similar behaviour for the number density of
the absorption regions while 𝜏 ≈ −1.6 for that. After the percolation
transition, the largest region appears as a disjoint part of the size
distribution of these regions.
• The size distribution of the emission regions when estimated
using the mean free path method shows a log-normal type behaviour
as a function of the length. A similar type of size distribution can
be seen for the ionized regions during the EoR. On the other hand,
the size distribution of the absorption regions is similar to the size
distribution of the neutral regions during the EoR. A similar con-
clusion can also be made when the size distribution is estimated
using the granulometry method.
• We find that the distributions of the Shapefinders (and genus)
of the emission regions, with respect to their volumes, are quite
similar to those of the absorption regions when both segments are
compared at the same filling factor. In particular, 𝐿 of both emission
and absorption type regions is significantly larger compared to 𝑇
and 𝐵. This difference increases with the increase of the volume of
these regions resulting in higher values 𝐹. Therefore, we find that
larger emission/absorption regions tend to be more filamentary with
higher values of 𝐺 as well.
This study has been done in a very idealistic situation. In re-
ality, the 21-cm signal as observed by the radio telescopes will be
contaminated by the astrophysical foreground, system noise, radio
frequency interference (RFI), etc. Even for the most optimistic sce-
nario with a full 𝑢𝑣 coverage of the instrument, an efficient method
to mitigate the foregrounds, RFI, etc., a perfect calibration tech-
nique, the presence of the system noise from a realistic observation
(e.g., 1000 hours with the SKA) will affect the recovery of such
high-resolution images as we have used in this study. Thus, the re-
covered size distribution will be affected. Studying the impact of
the noise of the system on the recovery of the size distribution of
the CD emission regions is beyond the scope of this work. We will
address this issue in a follow-up paper.
ACKNOWLEDGEMENTS
SB thanks Varun Sahni, Prakash Sarka, Santanu Das for their con-
tributions in developing surfgen2 in its initial phase. SB acknowl-
edges the funding provided by the Alexander von Humboldt Foun-
dation.
DATA AVAILABILITY
The derived data generated in this research will be shared upon
reasonable request to the corresponding author.
REFERENCES
Abdurashidova Z., et al., 2022, ApJ, 924, 51
Bañados E., et al., 2018, Nature, 553, 473
Bag S., Mondal R., Sarkar P., Bharadwaj S., Sahni V., 2018, MNRAS, 477,
1984
Bag S., Mondal R., Sarkar P., Bharadwaj S., Choudhury T. R., Sahni V.,
2019, MNRAS, 485, 2235
Bag S., Liivamägi L. J., Einasto M., 2023, MNRAS, 521, 4712
Barkana R., 2018, Nature, 555, 71
Bera A., Ghara R., Chatterjee A., Datta K. K., Samui S., 2023, Journal of
Astrophysics and Astronomy, 44, 10
Bharadwaj S., Ali S. S., 2005, MNRAS, 356, 1519
Bowman J. D., Rogers A. E. E., 2010, Nature, 468, 796
Bowman J. D., Rogers A. E. E., Monsalve R. A., Mozdzen T. J., Mahesh N.,
2018, Nature, 555, 67
Chardin J., Aubert D., Ocvirk P., 2012, A&A, 548, A9
Chernyaev E. V., 1987, ACM SIGGRAPH Computer Graphics, 21, 163
Crofton M. W., 1868, Phil. Trans. R. Soc. Lond, 158, 181
Dasgupta S., et al., 2023, J. Cosmology Astropart. Phys., 2023, 014
Datta K. K., Ghara R., Hoque A., Majumdar S., 2022, MNRAS, 509, 945
DeBoer D. R., et al., 2017, Publications of the Astronomical Society of the
Pacific, 129, 045001
Dixon K. L., Iliev I. T., Mellema G., Ahn K., Shapiro P. R., 2016, Monthly
Notices of the Royal Astronomical Society, 456, 3011
Fan X., et al., 2006, AJ, 132, 117
Fialkov A., Barkana R., Cohen A., 2018, Physical Review Letters, 121,
011101
Fixsen D. J., et al., 2011, ApJ, 734, 5
Friedrich M. M., Mellema G., Alvarez M. A., Shapiro P. R., Iliev I. T., 2011,
MNRAS, 413, 1353
Furlanetto S. R., Oh S. P., 2016, MNRAS, 457, 1813
Furlanetto S. R., Oh S. P., Briggs F. H., 2006, Phys. Rep., 433, 181
Gehlot B. K., et al., 2020, MNRAS, 499, 4158
Ghara R., Choudhury T. R., 2020, MNRAS, 496, 739
Ghara R., Mellema G., 2020, MNRAS, 492, 634
Ghara R., Choudhury T. R., Datta K. K., 2015, MNRAS, 447, 1806
Ghara R., Choudhury T. R., Datta K. K., 2016, MNRAS, 460, 827
Ghara R., Choudhury T. R., Datta K. K., Choudhuri S., 2017, MNRAS, 464,
2234
Ghara R., Mellema G., Giri S. K., Choudhury T. R., Datta K. K., Majumdar
S., 2018, MNRAS, 476, 1741
Ghara R., et al., 2020, MNRAS, 493, 4728
Ghara R., Giri S. K., Ciardi B., Mellema G., Zaroubi S., 2021, MNRAS,
503, 4551
Ghara R., Mellema G., Zaroubi S., 2022, J. Cosmology Astropart. Phys.,
2022, 055
Giri S. K., Mellema G., Ghara R., 2018, MNRAS, 479, 5596
Giri S. K., Mellema G., Aldheimer T., Dixon K. L., Iliev I. T., 2019a,
MNRAS, 489, 1590
Giri S. K., D’Aloisio A., Mellema G., Komatsu E., Ghara R., Majumdar S.,
2019b, Journal of Cosmology and Astro-Particle Physics, 2019, 058
Greig B., et al., 2021, MNRAS, 501, 1
Hinshaw G., et al., 2013, ApJS, 208, 19
Iliev I. T., et al., 2006, MNRAS, 371, 1057
Iliev I. T., Mellema G., Ahn K., Shapiro P. R., Mao Y., Pen U.-L., 2014,
MNRAS, 439, 725
Islam N., Ghara R., Paul B., Choudhury T. R., Nath B. B., 2019, MNRAS,
487, 2785
Kakiichi K., et al., 2017, MNRAS, 471, 1936
Kapahtia A., Chingangbam P., Ghara R., Appleby S., Choudhury T. R.,
2021, J. Cosmology Astropart. Phys., 2021, 026
Klypin A., Shandarin S. F., 1993, ApJ, 413, 48
Kolopanis M., et al., 2019, ApJ, 883, 133
Krause F., Thomas R. M., Zaroubi S., Abdalla F. B., 2018, New Astron., 64,
9
Lorensen W. E., Cline H. E., 1995, Technical Report CERN-CN-95-17
Madau P., Meiksin A., Rees M. J., 1997, ApJ, 475, 429
Mao Y., Shapiro P. R., Mellema G., Iliev I. T., Koda J., Ahn K., 2012,
MNRAS, 422, 926
Mecke K. R., Buchert T., Wagner H., 1994, A&A, 288, 697
MNRAS 000, 1–11 ()
 IGM during the Cosmic Dawn 11
Mellema G., Koopmans L., Shukla H., Datta K. K., Mesinger A., Majumdar
S., 2015, Advancing Astrophysics with the Square Kilometre Array
(AASKA14), p. 10
Mertens F. G., et al., 2020, MNRAS, 493, 1662
Mesinger A., Furlanetto S., 2007, ApJ, 669, 663
Mesinger A., Furlanetto S., Cen R., 2011, MNRAS, 411, 955
Mitra S., Choudhury T. R., Ferrara A., 2015, MNRAS, 454, L76
Mondal R., et al., 2020, MNRAS, 498, 4178
Monsalve R. A., Rogers A. E. E., Bowman J. D., Mozdzen T. J., 2017, ApJ,
835, 49
Muñoz J. B., Loeb A., 2018, Nature, 557, 684
Murmu C. S., Ghara R., Majumdar S., Datta K. K., 2022, Journal of Astro-
physics and Astronomy, 43, 104
Nebrin O., Ghara R., Mellema G., 2019, J. Cosmology Astropart. Phys.,
2019, 051
Pathak A., Bag S., Dasgupta S., Majumdar S., Mondal R., Kamran M.,
Sarkar P., 2022, J. Cosmology Astropart. Phys., 2022, 027
Patra N., Subrahmanyan R., Sethi S., Udaya Shankar N., Raghunathan A.,
2015, ApJ, 801, 138
Planck Collaboration et al., 2020, A&A, 641, A6
Price D. C., et al., 2018, MNRAS, 478, 4193
Pritchard J. R., Furlanetto S. R., 2007, MNRAS, 376, 1680
Reis I., Fialkov A., Barkana R., 2020, MNRAS, 499, 5993
Ross H. E., Dixon K. L., Ghara R., Iliev I. T., Mellema G., 2019, MNRAS,
487, 1101
Ross H. E., Giri S. K., Mellema G., Dixon K. L., Ghara R., Iliev I. T., 2021,
MNRAS, 506, 3717
Sahni V., Sathyaprakash B. S., Shandarin S. F., 1998, ApJ, 495, L5
Shandarin S. F., 1983, Pisma v Astronomicheskii Zhurnal, 9, 195
Shaw A. K., et al., 2023, J. Astrophys. Astron., 44, 4
Sheth J. V., Sahni V., Shandarin S. F., Sathyaprakash B. S., 2003, MNRAS,
343, 22
Singh S., et al., 2017, ApJ, 845, L12
The HERA Collaboration et al., 2022, arXiv e-prints, p. arXiv:2210.04912
Thomas R. M., et al., 2009, MNRAS, 393, 32
Trott C. M., et al., 2020, MNRAS, 493, 4711
Wayth R. B., et al., 2018, Publ. Astron. Soc. Australia, 35, 33
Yess C., Shandarin S. F., 1996, ApJ, 465, 2
Yoshiura S., Shimabukuro H., Takahashi K., Matsubara T., 2017, MNRAS,
465, 394
Zaroubi S., 2013, in Wiklind T., Mobasher B., Bromm V., eds, Astro-
physics and Space Science Library Vol. 396, The First Galaxies. p. 45
(arXiv:1206.0267), doi:10.1007/978-3-642-32362-1_2

This paper has been typeset from a TEX/LATEX file prepared by the author.

MNRAS 000, 1–11 ()