Draft version May 6, 2024

Typeset using LATEX twocolumn style in AASTeX631

A quantification of the effects using different stellar population synthesis models for epoch of
reionization
Peiai Liu,1 Qingbo Ma ,1, 2 Yunkun Han,3 and Rongxin Luo1, 2
1 Schoolof Physics and Electronic Science, Guizhou Normal University, Guiyang 550001, PR China
2 Guizhou Provincial Key Laboratory of Radio Astronomy and Data Processing,
arXiv:2405.01821v1 [astro-ph.GA] 3 May 2024

Guizhou Normal University, Guiyang 550001, PR China

3 Yunnan Observatories, Chinese Academy of Sciences, 396 Yangfangwang, Guandu District, Kunming, 650216, PR China

(Received; Revised; Accepted)

ABSTRACT
The luminosity and spectral energy distribution (SED) of high-z galaxies are sensitive to the stellar
population synthesis (SPS) models. In this paper, we study the effects of different SPS models on the
measurements of high-z galaxies and the budget of ionizing photons during the epoch of reionization,
by employing each of them in the semi-analytical galaxy formation model L-Galaxies 2020. We
find that the different SPS models lead to ≲ 0.5 dex differences on the amplitudes of UV luminosity
functions, while the two modes of the same SPS model with and without the inclusion of binary stars
leads to similar UV luminosity functions at z ≥ 6. Instead, the binary stars produce ∼ 40% more
ionizing photons than the single stars, while such differences are smaller than those caused by different
SPS models, e.g. the BPASS model produces ∼ 100% more ionizing photons than other models.

Keywords: Reionization(1383), High-redshift galaxies (734), Galaxy evolution (594)

1. INTRODUCTION radiation of UV and X-ray sources, and the ionizing

Although the gas of intergalactic medium (IGM) is and heating of the IGM gas (Furlanetto et al. 2006).
observed highly ionised in today’s Universe, it was fully The galaxy formation relates to e.g. the cooling of hot
neutral in the past after the cosmic recombination, the gas, star formation, supernova feedback, active galac-
period of transition from neutral to ionised phase of the tic nuclei (AGN) feedback and galaxy mergers (Dayal &
Universe is called as the epoch of reionization (EoR, Ferrara 2018). Some theoretical models have been devel-
Furlanetto et al. 2006; Dayal & Ferrara 2018). EoR oped to describe the galaxy formation and evolution, e.g.
happened ∼ 1.3 million years after the big bang, when the halo occupation distribution model (HOD, Zheng
the first generation of stars and galaxies formed and et al. 2005), the sub-halo abundance matching model
radiated photons, which then ionised the hydrogen and (SHAM, Campbell et al. 2018) and the conditional lu-
helium in the IGM (Barkana 2016). The observations, minosity function model (CLF, Yang et al. 2003). These
e.g. Lyα absorption lines of quasars (Fan 2006), the models focus on the correlations between different physi-
optical depth of cosmic microwave background radiation cal quantities on the formation and evolution of galaxies,
(CMB, Planck Collaboration et al. 2020) and the Lyα while do not involve specific physical processes. The hy-
emitters (Weinberger et al. 2019), suggest that the EoR drodynamic simulations can include abundant physical
ended at z > 5. processes and provide the best description of galaxy for-
The evolution of EoR includes the physical processes mation processes (Kannan et al. 2022), while they are
of the formation of dark matter halos and large-scale very computing expensive. The semi-analytical model
structures, the formation of first stars and galaxies, the (SAM) basing on the merger trees from N-body simula-
tions can describe almost all the physical processes re-
lated to galaxy formation (Henriques et al. 2020), which
Corresponding author: Qingbo Ma, Yunkun Han is more efficient than the hydrodynamic simulations but
maqb@gznu.edu.cn, hanyk@ynao.ac.cn more precise than the theoretical models. The SAM
2

models, e.g. the Santa Cruz semi-analytic model (Yung the galaxy formation and SPS models adopted in Sect
et al. 2019), ASTRAEUS (Hutter et al. 2021) and 2, present the results in Sect 3, and the conclusions are
MERAXES (Balu et al. 2023), have been applied to summarized in Sect 4.
study the high-z galaxies (z ≥ 6), which can explain
the high-z measurements e.g. the UV luminosity func- 2. METHODS
tions. Recently, the well-developed galaxy formation We apply the SAM model L-Galaxies 2020 (Hen-
SAM model L-Galaxies 2020 (Henriques et al. 2020) riques et al. 2020) in combination with the N-body dark
is also applied to explain the high-z observations and matter simulation Millennium-II (Boylan-Kolchin et al.
study the EoR process (Ma et al. 2023). 2009) to investigate the effects of different SPS models
The stellar population synthesis (SPS) model is a on the study of high-z galaxies and EoR. We will briefly
key component within galaxy formation models (Conroy describe the simulations and the SPS models here, while
2013), which relates the stellar mass, age, and metallic- the readers can refer to the original papers for more de-
ity to the luminosity and spectral energy distribution tails.
(SED) of galaxies (Henriques et al. 2015). Many SPS
models have been developed to explain the observations 2.1. Dark matter simulations
of galaxies, e.g. Bruzual & Charlot 2003 (named as The dark matter simulation and merger trees adopted
BC03), Maraston 2005 (named as M05), Yunnan evo- are from the Millennium-II simulation (MS-II, Boylan-
lutionary population synthesis model (YEPS, Zhang Kolchin et al. 2009), which was run with an up-
et al. 2004, 2005), and Binary Population and Spectral dated version of the GADGET code (Springel 2005)
Synthesis mode (BPASS, Eldridge et al. 2017; Stanway i.e. GADGET-3. The original box size of MS-II is
& Eldridge 2018). Since more than 50% of observed 100 Mpc/h with 21603 dark matter particles, and each
stars in galaxies and clusters are in binary systems, the particle is with mass of 6.89 × 106 h−1 M⊙ . The halos
YEPS and BPASS models also consider the effects of bi- were identified using the Friends-Of-Friends (FOF) algo-
nary stars (Zhang et al. 2005; Stanway & Eldridge 2018). rithm (More et al. 2011), then the SUBFIND algorithm
The interaction of binary stars can change the expected was applied to identify the self-bound substructures
SEDs of galaxies and produce more ionising photons within each FOF group. The halos are with at least 20
(Stanway et al. 2016; Götberg et al. 2020). The differ- particles, i.e. the minimal halo mass 1.38 × 108 h−1 M⊙ .
ences on the SPS models are expected to affect the prop- The simulations are scaled to the Planck cosmology
erties of high-z galaxies and the budget of ionizing pho- (Planck Collaboration et al. 2020) following the proce-
tons (Wilkins et al. 2016; Seeyave et al. 2023), which can dure of Angulo & Hilbert (2015), with the cosmologi-
be measured by the current and near future telescopes. cal parameters Ωm = 0.315, Ωb = 0.049, ΩΛ = 0.685,
For example, the high-z galaxies can be observed by the h = 0.673, σ8 = 0.826 and ns = 0.965.
Hubble Space Telescope (HST, Bouwens et al. 2015) MS-II simulation has outputs of 68 snapshots from
and the James Webb Space Telescope (JWST, Roberts- z = 127 to 0, while 22 of them are at z ≥ 6. The merger
Borsani et al. 2021), while the EoR can be measured by trees were constructed with the sub-halos found in these
the 21-cm signals from neutral hydrogen with the low- snapshots, in this case the merger trees of MS-II simu-
frequency radio telescope arrays such as the Low Fre- lation include ∼ 590 million sub-halos in total (Boylan-
quency Array (LOFAR1 ), the Square Kilometre Array Kolchin et al. 2009). The more or less snapshots should
(SKA2 ), the Murchison Widefield Array (MWA3 ), and not obviously change the conclusions presented in this
the Hydrogen Epoch of Reionization Array (HERA4 ). paper. We test it with the low resolution simulations
In this paper, we investigate the effects of different described in Ma et al. (2023).
SPS models on the observations of high-z galaxies and
the budget of ionizing photons using the SAM model 2.2. SAM model L-Galaxies 2020
L-Galaxies 2020 (Henriques et al. 2020) and the
We adopt the public SAM model L-Galaxies 2020
Millennium-II simulation (Boylan-Kolchin et al. 2009).
(Henriques et al. 2020) to evolve the galaxies with the
The paper is organized as the following: we describe
merger trees from MS-II simulation. This model has in-
cluded almost all the physical processes related to galaxy
1 http://www.lofar.org/ formation, such as gas cooling, star formation, galaxy
2 https://www.skatelescope.org/ merger, supernovae and AGN feedback. The application
3 http://www.mwatelescope.org/
of L-Galaxies 2020 on the high-z galaxies and EoR
4 https://reionization.org/
has been explored in Ma et al. (2023). The L-Galaxies
2020 model is the updated version of Henriques et al.
 3

(2015), which has a few differences with respect to the 2.3.1. BC03 Model
latter. In the new model, the galactic discs are spa-
The SED of BC035 covers the wavelength range from
tially resolved by dividing the discs into 12 concentric
91 Å to 3.6 × 108 Å with 2023 outputs. The age of stars
rings with radii ri = 0.01 × 2i h−1 kpc, i = 0, ..., 11 (Fu
is from 0 yr to 20 Gyr, with a grid of 221 steps. There
et al. 2013). All the properties and physical processes of
are seven available metallicities, i.e. Z = 10−4 , 4×10−4 ,
discs, such as star formation, chemical enrichment and
4×10−3 , 8×10−3 , 0.02, 0.05 and 0.1. This information is
gas ejection, are evolved for each ring separately. The
summarized in Tab. 1. From Fig. 1, the stellar mass nor-
star formation is linked to the H2 abundance of each ring
malized SEDs from BC03 are similar to the M05 model
that depending on the metallicity of gas (Krumholz et al.
at age ≤ 100 Myr, while closer to the BPASS model at 1
2009; McKee & Krumholz 2010).
Gyr. With the increasing of metallicities, the radiations
We adopt the default values for the free parameters
at hP ν > 13.6 eV are slightly reduced at age 10 Myr,
in the L-Galaxies 2020 code, which are the best fit
while this effect is not very significant at other ages.
values with the observations of galaxies at low-z and
the merger trees from MS-II simulation (Henriques et al.
2.3.2. M05 Model
2020). We only modify the input SPS model adopted in
the L-Galaxies 2020 code. The wavelength range of the SED from M056 is from
91 Å to 1.6 × 106 Å, with 1221 outputs. The age of stars
2.3. Stellar populations synthesis is 103 yr to 15 Gyr, with a grid of 67 steps. The outputs
are available at four metallicity values, i.e. Z = 10−3 ,
We adopt four popular SPS models in this paper, i.e. 0.01, 0.02 and 0.04. As showed in Fig. 1, the SEDs of
BC03 (Bruzual & Charlot 2003), M05 (Maraston 2005), M05 are roughly similar to the BC03 model at age ≤ 100
YEPS (Zhang et al. 2005) and BPASS (Stanway & El- Myr, while present obviously less high energy radiation
dridge 2018). The parameters of these SPS models, in- (hp ν > 13.6 eV) at 1 Gyr. Its SEDs are also similar to
cluding the wavelength range, age and metallicity, are the single star mode of YEPS and BPASS models (i.e.
summarized in Tab. 1. To make consistent comparison, YEPS SS and BPASS SS) at age 1 Myr and 10 Myr,
we employ the results of these models with the same while have less radiation at hp ν > 13.6 eV at age 100
Salpeter stellar initial mass function (IMF, Salpeter Myr and 1 Gyr.
1955) with a power index α of 1.35 and maximum ini-
tial mass 100 M⊙ . For the YEPS and BPASS models, 2.3.3. YEPS model
we consider both the ones including binary stars, which
are named as YEPS BS and BPASS BS respectively, The wavelength range of YEPS SED7 is 91 Å - 1.6 ×
and the ones with only single stars, which are named as 106 Å with 1221 outputs. The age of stars covers the
YEPS SS and BPASS SS respectively. range from 0.1 Myr to 15 Gyr, with 90 outputs. The
As a reference, in Fig. 1 we present the initial SEDs results at 7 metallicities are available, i.e. Z = 10−4 ,
from different SPS models at four metallicities (Z = 3 × 10−4 , 10−3 , 4 × 10−3 , 0.01, 0.02, 0.03. We adopt
10−4 , 10−3 , 4 × 10−3 and 0.02) and four ages (1 Myr, two modes of SEDs from YEPS model, one with the
10 Myr, 100 Myr and 1 Gyr). Note that the SEDs in binary stars (YEPS BS) and one with only the single
Fig. 1 are only in the frequency range [6.2, 200] eV (i.e. stars (YEPS SS). From Fig. 1, the SEDs of single star
UV band), while the Fig. 6 in Appendix A shows one mode (YEPS SS) are similar to other models at 1 Myr
example of the full SEDs. As showed in Fig. 1, the and 10 Myr, while close to the BPASS model at 100
initial SEDs from all four SPS models are not very sen- Myr and 1 Gyr. Compared to YEPS SS, the differences
sitive to the metallicities, while the amplitudes reduce caused by the binary stars (YEPS BS) are significant at
obviously with the increasing ages. The different SPS age 10 Myr and slightly at 100 Myr, but not too much at
models present roughly consistent SEDs, especially at 1 Myr and 1 Gyr. The inclusion of binary stars obviously
hP ν < 13.6 eV, while some differences are still visible. increases the amplitudes of SEDs at hp ν > 13.6 eV at
The inclusion of binary stars in the YEPS and BPASS age 10 Myr, while not too much at other ages.
models (i.e. YEPS BS and BPASS BS) leads to harder
SEDs at hP ν > 13.6 eV compared to those with single 2.3.4. BPASS model
stars (i.e. YEPS SS and BPASS SS), and thus expects
to increase the emission of ionizing photons. In the fol-
5 http://www.bruzual.org/
lowing, we will briefly describe these SPS models. For
6 http://www.icg.port.ac.uk/∼maraston/
more comparisons about the SPS models, one can also
7 http://www1.ynao.ac.cn/∼zhangfh/YN SP.html
refer to e.g. Chen et al. (2010) and Han & Han (2019).
4

Table 1. Parameters of four SPS models applied in the L-Galaxies 2020 model.

Models Wavelength range Nwavelength Age range Nages Nmetallicites

BC03 91 Å-3.6 × 108 Å 2023 0 yr - 20 Gyr 221 7
M05a 91 Å-1.6 × 106 Å 1221 103 yr - 15 Gyr 67 4
YEPSb 91 Å-1.6 × 106 Å 1221 0.1 Myr - 15 Gyr 90 7
BPASS 1 Å-1.0 × 105 Å 100,000 1 Myr - 100 Gyr 51 13
a The version incorporating the red horizontal-branch morphology was employed.
b The version employed in this paper is Yunnan-II, while the updated Yunnan-III version (Zhang et al.
2013) is only available for solar-metallicity.

Z = 10 −4 Z = 10 −3 Z = 4 × 10 −3 Z = 0.02
1020 1 Myr
1018
1016
1014
1012
SED[erg s−1 Hz−1 M−¯1 ]

1020 10 Myr
1018
1016
1014
1012
1020 100 Myr
1018
1016
1014
1012
1020 YEPS BS
YEPS SS
1 Gyr
BPASS BS
1018 BPASS SS
BC03
1016 M05
1014
1012 101 102 101 102 101 102 101 102
hP ν[eV] hP ν[eV] hP ν[eV] hP ν[eV]

Figure 1. Stellar mass normalized SEDs of four SPS models, i.e. BC03 (dash-dotted cyan), M05 (dotted blue), YEPS (solid
black) and BPASS (dashed magenta), at metallicity Z = 10−4 , 10−3 , 4 × 10−3 and 0.02 (from left to right), and age 1 Myr,
10 Myr, 100 Myr and 1 Gyr (from top to bottom). The thin lines of YEPS (YEPS BS) and BPASS (BPASS BS) models are
the SEDs including binary stars, while the thick lines (i.e. YEPS SS and BPASS SS) are with single stars. Note that in the
BC03 model, the SED output of Z = 10−3 is obtained by interpolation. In the M05 model, the SED output of Z = 4 × 10−3 is
obtained by interpolation, and at Z = 10−4 its SED is not shown because the age of star is only in the range of 1 Gyr-15 Gyr.
The vertical gray lines denote the location of 13.6 eV.
 5

We adopt the results of version 2.2.1 of BPASS8 at higher energy band. The BPASS model shows obvi-
model, including the modes with binary stars ously higher amplitudes of SEDs than other models at
(BPASS BS) and only single stars (BPASS SS). The hp ν = 13.6 − 50 eV. Within the same band, the BC03
wavelength range is from 1 Å to 105 Å with 105 outputs. model is slightly lower than the BPASS model, while
The age of stars is from 1 Myr to 100 Gyr with 51 steps. the YEPS model has the lowest amplitude, and the M05
There are 13 available metallicities: Z = 10−5 , 10−4 , model is between BC03 and YEPS models. Differently,
10−3 , 2 × 10−3 , 3 × 10−3 , 4 × 10−3 , 6 × 10−3 , 8 × 10−3 , the YEPS model has the highest SED of galaxies at
0.01, 0.014, 0.02, 0.03, 0.04. As showed in Fig. 1, the hp ν > 50 eV and z ≥ 7, while becomes similar to the
SEDs of BPASS model have slightly higher amplitudes BPASS model at z = 6. Within the same band, the
than other models at hp ν < 50 eV and age 1 Myr. At 10 SEDs of galaxies with M05 model are roughly consis-
Myr, its SEDs with single stars (BPASS SS) are similar tent with the BPASS model at z ≥ 7, while obviously
to other models. At 100 Myr and 1 Gyr, the SEDs of lower than the latter at z = 6. The SEDs of galaxies
BPASS SS are slightly lower than the YEPS model, but with BC03 model at z ≥ 7 are similar to those with
have more high energy radiation (hp ν > 13.6 eV) than M05 and BPASS models at Mvir < 1010 M⊙ , while be-
the BC03 and M05 models. The inclusion of binary stars come similar to those with YEPS models with the in-
(BPASS BS) increases the SEDs at hp ν > 13.6 eV at age creasing halo mass e.g. at Mvir = 1011 − 1012 M⊙ . The
10 Myr, but not too much at other ages, similar to that inclusion of binary stars (BPASS BS) does not obviously
of YEPS model. change the SEDs of BPASS model at hp ν < 50 eV, while
increases the high energy radiation at hp ν > 50 eV.
3. RESULTS
Such effect is weaker on the SEDs of massive halos (e.g.
In this section, we present the properties of high-z Mvir = 1011 − 1012 M⊙ ) than the less massive ones, due
galaxies and the budgets of ionizing photon from L- to the increasing contributions of old stars (age > 100
Galaxies 2020 with different SPS models. Myr). With the same reason, the SEDs of galaxies at
3.1. SEDs of high-z galaxies z = 6 has no significant features of binary stars. Instead,
the inclusion of binary stars in YEPS model (YEPS BS)
The SEDs of galaxies from L-Galaxies 2020 are has no obvious effects on the SEDs of high-z galaxies,
computed with the SPS results (i.e. the Fig. 1) and the except slightly higher amplitude at hp ν > 40 eV and
history of star formation and metal enrichment within z = 6.
each galaxy. More specifically, with the information of Note that, due to the abundant neutral hydrogen
star formation and metal enrichment history produced, during EoR, it will be hard to measure the SEDs at
the L-Galaxies 2020 post-processes the SEDs of each > 13.6 eV, although the uncertainties of different SPS
galaxy by linearly interpolated the input SEDs from dif- models are mostly at such band. The measurements
ferent SPS models at specific ages and metallicities, and and comparisons at rest-frame UV band e.g. the UV lu-
then multiply the star mass. The final SED of galaxies minosity function (ϕ showed in Fig. 3) can exclude the
is the sum of the results at all ages. Fig. 2 shows the uncertainties of physical processes except the SPS mod-
average rest-frame SEDs of high-z galaxies within the els, while the results of SED fitting, e.g. the ionizing
same halo mass range obtained from L-Galaxies 2020 photon production efficiency ζion , can help to distinguish
with four SPS models at different zs. To compare with the different SPS models (Seeyave et al. 2023).
the input SEDs showed in Fig. 1, the results of Fig. 2 are
normalized by the stellar mass of galaxies. By compar- 3.2. UV luminosity function of galaxies
ing with the Fig. 1, we can see that the SEDs of high-z Fig. 3 shows the UV luminosity function ϕ at the rest-
galaxies within UV band are dominated by the young frame wavelength λ = 1600 Å from four SPS models at
stars, i.e. those with age < 100 Myr. Due to the same different zs. The absolute magnitude of galaxy luminos-
reason, the SEDs of high-z galaxies do not evolve too ity at λ = 1600 Å is computed by:
much with the decreasing redshift. The SEDs of mas- 5

F1600

sive halos are slightly lower than the less massive ones, M1600,AB = − log10 − 48.6 (1)
2 4πR2
since the latter ones have higher ratio of star formation
rate over stellar mass (Henriques et al. 2020). where F1600 is the brightness of galaxy at λ = 1600Å,
Four SPS models predict similar SEDs of high-z galax- and R = 10 pc. As a comparison, we also present some
ies at hp ν < 13.6 eV, while some differences are obvious of recent observations of ϕ from HST (Bouwens et al.
2021) and JWST (Donnan et al. 2023; Adams et al.
2023; Finkelstein et al. 2023; Donnan et al. 2024) tele-
8 http://bpass.auckland.ac.nz scopes. Note that the results of Bouwens et al. (2021)
6

Mvir = 10 8 − 10 9 M ¯ Mvir = 10 9 − 10 10 M ¯ Mvir = 10 10 − 10 11 M ¯ Mvir = 10 11 − 10 12 M ¯

1020 z=9
1018 YEPS BS
YEPS SS
10 16 BPASS BS
BPASS SS
10 14 BC03
M05
10 12
SED[erg s−1 Hz−1 M−¯1 ]

1020 z=8
1018
1016
1014
1012
1020 z=7
1018
1016
1014
1012
1020 z=6
1018
1016
1014
1012 101 102 101 102 101 102 101 102
hP ν[eV] hP ν[eV] hP ν[eV] hP ν[eV]

Figure 2. Stellar mass normalized SEDs of high-z galaxies with halo mass (from left to right) Mvir = 108 − 109 M⊙ , 109 −
1010 M⊙ , 1010 − 1011 M⊙ and 1011 − 1012 M⊙ with four SPS models i.e. BC03 (dash-dotted cyan), M05 (dotted blue), YEPS
(solid black) and BPASS (dashed magenta). From top to bottom, the results are at z = 9, 8, 7 and 6. The thin lines of YEPS
(YEPS BS) and BPASS (BPASS BS) models are the SEDs including binary stars, while the thick lines (i.e. YEPS SS and
BPASS SS) are with single stars. The vertical gray lines denote the location of 13.6 eV. The semitransparent gray region denotes
the frequency ranges that can be measured by the JWST telescope.

are at λ = 1600Å, while others that observed by the L-Galaxies 2015 (Clay et al. 2015). Although the dust
JWST telescope are at λ = 1500 Å. Such differences are model is not included to precisely match the results from
not significant on the UV luminosity functions. Table 2 the MS-II simulation and L-Galaxies 2020 code with
shows the χ2 of ϕ from different SPS models compared the observations, it indeed affects the UV luminosity of
to the observations at three zs, which is defined as: the bright galaxies at low redshifts (Yung et al. 2020a;
X (ϕobs − ϕsim )2 Bhagwat et al. 2023), but not too much on the ones at
χ2 = 2 (2) high redshifts and the faint ones.
σobs The UV luminosity function ϕ from four SPS mod-
where ϕobs is the observational ϕ, ϕsim is the ϕ from els are roughly consistent with the observations at six
simulations, and σobs is the 1-σ error of ϕobs . Due to the zs. The differences of four SPS models on ϕ are ≲ 0.5
lack of bright galaxies from MS-II simulation, the χ2 is dex, smaller than the uncertainties (i.e. error bars) of
computed only for the galaxies with M1600,AB > −21. the current measurements of ϕ. Specifically, the BPASS
The UV luminosities of galaxies can be reduced by the model has higher amplitudes of ϕ than other models.
extinction models of dust and molecular clouds. We test The BC03 model is globally similar to the M05 model,
the dust model within L-Galaxies 2020, which does which ϕs have amplitudes lower than the YEPS and
not change too much on the Fig. 3, especially at z > 7. BPASS models. The YEPS model is similar to the
This is different to previous studies, e.g. the results of BPASS model at M1600,AB > −18, while closer to the
 7

Table 2. χ2 of UV luminosity function ϕ from four SPS models compared to the observations at z = 10, 9 and 8.

z YEPS BS YEPS SS BPASS BS BPASS SS BC03 M05

10 16.49 20.26 26.39 18.10 28.81 32.63
9 22.02 29.01 26.02 14.62 40.75 45.30
8 25.75 31.80 11.83 16.27 43.03 43.99

Finkelstein2023 z = 11 Adams2023 z = 10 To properly compute the number of ionizing photons

10 1 Donnan2024 Donnan2023
φ [cMpc −3 mag −1 ]

Donnan2024 from high-z galaxies, we rerun the L-Galaxies 2020

10 3
SAM simulations with the integrated SED (iSED) over
10 5 the age of stars (i.e. the time from the birth of stars
to the output zs) for four SPS models. The iSED can
10 7
easily include the effects of star formation history (Ma
Bouwens2021 z=9 Bouwens2021 z=8 et al. 2023). Meanwhile, as showed in Ma et al. (2023),
10 1 Adams2023 Adams2023
φ [cMpc −3 mag −1 ]

Finkelstein2023 Donnan2023 after normalized by the stellar mass the iSED of high-
10 3 Donnan2023
z galaxies is sensitive neither to the galaxy formation
10 5 models nor to the output zs. The number of ionizing
photons (nion ) from galaxies is then calculated by:
10 7
Z
Bouwens2021 z=7 Bouwens2021 z=6 Lν
10 1 nion = dν (3)
13.6 eV P ν
h
φ [cMpc −3 mag −1 ]

10 3
YEPS BS
YEPS SS where Lν is the iSED of galaxies computed by L-
10 5 BPASS BS
BPASS SS
Galaxies 2020, hP is the Planck constant, and ν is
BC03 the frequency of photons. The integration is done with
10 7 M05
the full iSED at hp ν > 13.6 eV. The cosmic volume
24 22 20 18 16 14 24 22 20 18 16 14
M1600, AB M1600, AB averaged ionizing photons Nion (i.e. number density of
ionizing photon) is expressed as:
Figure 3. UV luminosity function ϕ at the rest-frame
wavelength λ = 1600 Å from four SPS models i.e. BC03 P
nion
(dash-dotted cyan), M05 (dotted blue), YEPS (solid black) Nion = (4)
and BPASS (dashed magenta). The thin lines of YEPS
Vbox
(YEPS BS) and BPASS (BPASS BS) models are the results where the sum
P
is for all the selected galaxies, and
including binary stars, while the thick lines (i.e. YEPS SS
Vbox is the comoving volume of MS-II simulation.
and BPASS SS) are with single stars. From left to right and
top to bottom, the six panels are the results at z = 11, 10, Fig. 4 shows the Nion of galaxies within different halo
9, 8, 7 and 6, respectively. The observational data points mass Mvir range from four SPS models as functions of
are from Bouwens et al. 2021 (blue up triangle), Adams zs. As a comparison, we also present the minimal budget
et al. 2023 (red circle), Finkelstein et al. 2023 (green down of ionizing photon number density to fully ionize neutral
triangle), Donnan et al. 2023 (yellow square) and Donnan hydrogen and first ionizing of helium by assuming 75% of
et al. 2024 (magenta diamond). Note that the observations baryon is hydrogen and 25% is helium, and show as the
of Bouwens et al. 2021 are at λ = 1600 Å, while others are
horizontal gray lines. From Fig. 4, the ionizing photons
at λ = 1500 Å.
are mostly from the galaxies with Mvir > 109 M⊙ , and ∼
half of them are from the galaxies with Mvir > 1010 M⊙ .
BC03 and M05 models at M1600,AB < −18. The inclu-
The galaxies with Mvir < 109 M⊙ only slightly increase
sion of binary stars has negligible effects on ϕ, both for
the Nion . This is partly due to the incomplete sample
the YEPS and BPASS models, consistent with the SEDs
of halos with Mvir < 109 M⊙ from MS-II simulation.
of high-z galaxies showed in Fig. 2. With the χ2 showed
Meanwhile, the supernovae and radiation feedback also
in Table 2, at z = 10 the YEPS BS model fits better
suppresses the star formation on such low mass halos
with the observations, then is the BPASS SS models.
(Hutter et al. 2021; Legrand et al. 2023). In Fig. 7 of
At z = 9, the BPASS SS model is the best fit one. At
Appendix B, we show one sample of the distributions of
z = 8, the best fit model is the BPASS BS model, then
ionizing photons at z = 7 versus halo mass.
is the BPASS SS model.
Since the different SPS models and the binary stars
mostly affect the radiation of galaxies at high energy
3.3. Budget of ionizing photons
band (hp ν > 13.6 eV) as showed in Fig. 2, the four
8

Mvir > 10 10 M ¯ Mvir > 10 9 M ¯ YEPS BS Hsiao2023

YEPS SS
102 z = 11 z = 10
Nion [10 65 × cMpc −3 ]

26.0

log(ζion [erg −1 Hz])

BPASS BS
BPASS SS
BC03
25.5 M05
101
25.0
100 24.5
Mvir > 10 8 M ¯ All halos Tang2023 Tang2023
Whitler2024 Whitler2024
102 z=9 z=8
Nion [10 65 × cMpc −3 ]

26.0

log(ζion [erg −1 Hz])

YEPS BS
YEPS SS
25.5
101 BPASS BS
BPASS SS 25.0
BC03
M05
1006 7 8 9 10 11 12 7 8 9 10 11 12 24.5
Tang2023 Sun2023
z z Castellano2022
26.0 z=7 z=6

log(ζion [erg −1 Hz])

Figure 4. Redshift evolution of volume averaged ionizing
photon Nion from four SPS models i.e. BC03 (dash-dotted 25.5
cyan), M05 (dotted blue), YEPS (solid black) and BPASS
(dashed magenta). The thin lines of YEPS (YEPS BS) and 25.0
BPASS (BPASS BS) models are the results including binary
stars, while the thick lines (i.e. YEPS SS and BPASS SS) are
24.56 7 8 9 6 7 8 9
with single stars. From left to right and top to bottom, the log 10 (M ∗ [M ¯ ]) log 10 (M ∗ [M ¯ ])
four panels are the Nion of galaxies with Mvir > 1010 M⊙ , Figure 5. Ionizing photon production efficiency ζion from
> 109 M⊙ , > 108 M⊙ and all halos. As a reference, the four SPS models i.e. BC03 (dash-dotted cyan), M05 (dotted
horizontal gray line is the minimal budget of ionizing photon blue), YEPS (solid black) and BPASS (dashed magenta).
number density to fully ionize neutral hydrogen and first The thin lines of YEPS (YEPS BS) and BPASS (BPASS BS)
ionizing of helium models are the results including binary stars, while the thick
. lines (i.e. YEPS SS and BPASS SS) are with single stars.
From left to right and top to bottom, the six panels are
SPS models have much larger differences on the bud- the results at z = 11, 10, 9 8, 7 and 6, respectively. The
get of ionizing photons (i.e. Nion ) than that on the UV observational data points are from Hsiao et al. 2023 (green
luminosity function showed in Fig. 3. Specifically, the down triangle), Whitler et al. 2024 (red up triangle), Tang
et al. 2023 (yellow square), Castellano et al. 2022 (blue star)
BPASS model has the highest Nion , which is ∼ 2 times
and Sun et al. 2023 (cyan cycle).
that of YEPS model. Comparing with the minimal bud-
get of ionizing photon number density to fully ionize
neutral hydrogen and first ionizing of helium, this can SPS models, which is defined as:
lead to redshift difference δz ∼ 1 on the end redshift of ṅion
EoR. Note that, the precise end redshift of EoR should ζion = (5)
F1600
be computed with the radiative transfer simulations, i.e.
where ṅion is the ionizing photon emissivity of galax-
the redshift difference estimated might be different due
ies. Our results are roughly consistent with the previous
to the ionizing and recombination models. The M05
studies e.g. Wilkins et al. (2016); Yung et al. (2020b);
model has Nion similar to that of YEPS model, which
Seeyave et al. (2023). We only present some results of re-
is slightly lower than that of BC03 model. With the in-
cent measurements about ζion , while for more ones one
clusion of binary stars, the BPASS model (BPASS BS)
can refer to the recent paper Simmonds et al. (2024).
can have ∼ 40% more budget of Nion than that with
From Fig. 5, the ζion from our simulations are lower
single stars (BPASS SS), consistent with the results of
than the measurements by Tang et al. (2023); Whitler
e.g. Ma et al. (2022), while the YEPS model (YEPS BS)
et al. (2024), while roughly consistent with other obser-
has ∼ 28% higher Nion than YEPS SS. The differences
vations.
on Nion caused by the binary stars are much smaller
than that due to different SPS models. Such results are 4. DISCUSSION AND CONCLUSION
consistent with the SEDs of galaxies showed in Fig. 2.
The stellar population synthesis (SPS) model is an
To compare with the observations, in Fig. 5 we show
important component within galaxy formation models.
the ionizing photon production efficiency ζion from four
The uncertainties on the predictions of SPS models
will affect the theoretical studies of galaxy formation
 9

and reionization process with the observations by the sumptions of IMF, stellar evolution, stellar atmosphere
telescopes. In this paper, we use the semi-analytical and/or binary star model will also change the predic-
galaxy formation model L-Galaxies 2020 (Henriques tions on the properties of high-z galaxies and the budget
et al. 2020) and the N-body dark matter simulation of ionizing photons (Seeyave et al. 2023). The specific
Millennium-II (MS-II, Boylan-Kolchin et al. 2009) to modeling of each SPS and the detailed explanation for
explore the effects of four popular SPS models, i.e. their difference are beyond the scope of this paper.
BC03 (Bruzual & Charlot 2003), M05 (Maraston 2005), Although the L-Galaxies 2020 code includes most
YEPS (Zhang et al. 2005) and BPASS (Stanway & El- of the physical processes related to galaxy formation,
dridge 2018), on the properties of high-z galaxies and while not the effects of radiation feedback, which can
the budget of ionizing photons during epoch of reioniza- suppress the star formation on halos < 109 M⊙ (Hut-
tion (EoR), including both the modes with binary stars ter et al. 2021; Legrand et al. 2023). However, we do
and only single stars. not expect it will obviously change the results in this
We find the uncertainties of different SPS models on paper, as the star formation of halos < 109 M⊙ is al-
the SEDs of galaxies are mostly at the high energy band, ready reduced by the supernovae feedback. We adopt
i.e. hP ν > 13.6 eV. With this reason, the four SPS mod- the best-fit parameters in the L-Galaxies 2020 model
els have not significant differences on the UV luminosity with the MS-II simulation and the low-z observations,
functions which are measured at the rest-frame wave- i.e. assuming that all the models at low-z are still avail-
length λ = 1600 Å, while predict obviously different able for the high-z ones. However, we can expect that
budget of ionizing photon number density (Nion ), con- different dark matter simulations and SAM codes might
sistent with the conclusions of e.g. Wilkins et al. (2016); lead to different results for high-z galaxies.
Yung et al. (2020b). Specifically, the BPASS model As a summary, although the different SPS models and
has the higher amplitudes of SED than other models binary stars will not significantly change the high-z UV
at hP ν < 50 eV, which predicts Nion ∼ 2 times that of luminosity functions, they obviously affect the predic-
YEPS model. The BC03 and M05 models predict simi- tion of the budget of ionizing photons during EoR. Thus,
lar SEDs, and thus have similar Nion , which are slightly the proper selection of SPS model is important to study
higher than the YEPS model but much lower than the the EoR process, especially when including the galaxy
BPASS model. The inclusion of binary stars does not formation process in the modeling of EoR.
visibly change the UV luminosity functions, while pre-
dicts ∼ 40% more ionizing photons with the BPASS 1 This work is supported by the National SKA Program
model and ∼ 28% more with the YEPS model. We 2 of China (grant No. 2020SKA0110402), National Natu-
note that, the BPASS model adopts a detailed stellar 3 ral Science Foundation of China (Grant No. 12263002),
evolution calculation with the Cambridge STARS code 4 and GZNU 2019 Special projects of training new aca-
instead of the approximate and rapid stellar evolution 5 demics and innovation exploration. Yunkun Han grate-
model applied in e.g. the YEPS model (Han & Han 6 fully acknowledges the support from the National Key
2019), which results should be more credible. 7 R&D Program of China (Nos. 2021YFA1600401 and
Considering that the differences of different SPS mod- 8 2021YFA1600400), the National Science Foundation of
els on the high-z galaxies are mostly on the ionizing band 9 China (grant nos. 11773063, 12288102) the China
which can be highly absorbed by the neutral hydrogen, 10 Manned Space Project (grant nos. CMS-CSST-2021-
we do not expect the JWST observations can directly 11 A02, CMS-CSST-2021-A04, CMS-CSST-2021-A06), the
distinguish different SPS models. The indirect measure- 12 ’Light of West China’ Program of Chinese Academy of
ments e.g. the ionizing photon production efficiency ζion 13 Sciences, the Yunnan Ten Thousand Talents Plan Young
measured by the SED fitting might help to distinguish 14 & Elite Talents Project, the Natural Science Foundation
different SPS models (Seeyave et al. 2023). 15 of Yunnan Province (No. 202201BC070003), and the
In this paper, we only focus on the effects of using dif- 16 International Centre of Supernovae, Yunnan Key Lab-
ferent SPS models. All the initial SEDs from four SPS 17 oratory (No. 202302AN360001). The tools for biblio-
models are the public data from their official websites. 18 graphic research are offered by the NASA Astrophysics
We take the ones with the same initial mass functions 19 Data Systems and by the JSTOR archive.
(IMF) i.e. Salpeter IMF. We note that the different as-

APPENDIX

A. ONE EXAMPLE OF FULL SED FROM FOUR SPS models have similar SEDs at λ ≳ 900 Å, while
SPS MODELS
Fig. 6 shows one example of full SED from four SPS
models at metallicity Z = 10−3 and age 10 Myr. Four
10

Z = 10 −3 , 10 Myr 0.5
1035 YEPS BS
YEPS SS
1029 BPASS BS
0.4

Distribution of ionizing photons

1023 BPASS SS
BC03
SED[erg s−1 Å−1 M−¯1 ]

1017 M05
0.3
1011
105 YEPS BS 0.2
YEPS SS
10 1 BPASS BS
BPASS SS
10 7 BC03 0.1
M05
100 101 102 103 104 105 106
λ[Å] 8 9 10 11 12
log 10 (Mvir [M ¯ ])

Figure 6. Stellar mass normalized SED from SPS model

Figure 7. Distributions of ionizing photons at z = 7 from
BC03 (dash-dotted cyan), M05 (dotted blue), YEPS (solid
SPS model BC03 (dash-dotted cyan), M05 (dotted blue),
black) and BPASS (dashed magenta), with wavelength λ as
YEPS (solid black) and BPASS (dashed magenta) as func-
the x-axis. The results are at metallicity Z = 10−3 and age
tions of halo mass Mvir . The thin lines of YEPS (YEPS BS)
10 Myr. The thin lines of YEPS (YEPS BS) and BPASS
and BPASS (BPASS BS) models are the results includ-
(BPASS BS) models are the SEDs including binary stars,
ing binary stars, while the thick lines (i.e. YEPS SS and
while the thick lines (i.e. YEPS SS and BPASS SS) are with
BPASS SS) are with single stars.
single stars.

show some differences at shorter λ. The inclusion of

binary stars obviously increases the amplitudes of SED
at λ ≲ 900 Å, both with the YEPS and BPASS models.

B. DISTRIBUTIONS OF IONIZING PHOTONS AT

Z = 7 AS FUNCTIONS OF HALO MASS
Fig. 7 shows the distributions of ionizing photons
at z = 7 from four SPS models, which is computed
P
by ∆(nion )/∆(log10 (Mvir ))/ nion , where ∆(nion ) is
sum of nion from the halos P within a mass bin width
∆(log10 (Mvir )) = 0.1, and nion is the total number of
ionizing photons from all halos. From Fig. 7, the ioniz-
ing photon number increases with halo mass decreasing
until Mvir ≈ 3 × 109 M⊙ , while decreases with halo mass
decreasing at lower Mvir . This conclusion is not very
sensitive to the SPS models except the M05 one.

REFERENCES
Adams, N. J., Conselice, C. J., Austin, D., et al. 2023, Bhagwat, A., Costa, T., Ciardi, B., Pakmor, R., & Garaldi,
arXiv e-prints, arXiv:2304.13721, E. 2023, arXiv e-prints, arXiv:2310.16895,
doi: 10.48550/arXiv.2304.13721 doi: 10.48550/arXiv.2310.16895
Bouwens, R. J., Illingworth, G. D., Oesch, P. A., et al.
Angulo, R. E., & Hilbert, S. 2015, MNRAS, 448, 364,
2015, ApJ, 803, 34, doi: 10.1088/0004-637X/803/1/34
doi: 10.1093/mnras/stv050
Bouwens, R. J., Oesch, P. A., Stefanon, M., et al. 2021, AJ,
Balu, S., Greig, B., Qiu, Y., et al. 2023, MNRAS, 520,
162, 47, doi: 10.3847/1538-3881/abf83e
3368, doi: 10.1093/mnras/stad281
Boylan-Kolchin, M., Springel, V., White, S. D. M., Jenkins,
Barkana, R. 2016, PhR, 645, 1, A., & Lemson, G. 2009, MNRAS, 398, 1150,
doi: 10.1016/j.physrep.2016.06.006 doi: 10.1111/j.1365-2966.2009.15191.x
 11

Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000, Ma, Q.-B., Fiaschi, S., Ciardi, B., Busch, P., & Eide, M. B.
doi: 10.1046/j.1365-8711.2003.06897.x 2022, MNRAS, 513, 1513, doi: 10.1093/mnras/stac1018
Campbell, D., van den Bosch, F. C., Padmanabhan, N., Ma, Q.-B., Ghara, R., Ciardi, B., et al. 2023, MNRAS, 522,
et al. 2018, MNRAS, 477, 359, doi: 10.1093/mnras/sty495 3284, doi: 10.1093/mnras/stad1203
Castellano, M., Pentericci, L., Cupani, G., et al. 2022, Maraston, C. 2005, MNRAS, 362, 799,
A&A, 662, A115, doi: 10.1051/0004-6361/202243348 doi: 10.1111/j.1365-2966.2005.09270.x
Chen, X. Y., Liang, Y. C., Hammer, F., et al. 2010, A&A, McKee, C. F., & Krumholz, M. R. 2010, ApJ, 709, 308,
515, A101, doi: 10.1051/0004-6361/200913894 doi: 10.1088/0004-637X/709/1/308
Clay, S. J., Thomas, P. A., Wilkins, S. M., & Henriques, B. More, S., Kravtsov, A. V., Dalal, N., & Gottlöber, S. 2011,
M. B. 2015, MNRAS, 451, 2692, ApJS, 195, 4, doi: 10.1088/0067-0049/195/1/4
doi: 10.1093/mnras/stv818 Planck Collaboration, Aghanim, N., Akrami, Y., et al.
Conroy, C. 2013, ARA&A, 51, 393, 2020, A&A, 641, A6, doi: 10.1051/0004-6361/201833910
doi: 10.1146/annurev-astro-082812-141017 Roberts-Borsani, G., Treu, T., Mason, C., et al. 2021, ApJ,
Dayal, P., & Ferrara, A. 2018, PhR, 780, 1, 910, 86, doi: 10.3847/1538-4357/abe45b
doi: 10.1016/j.physrep.2018.10.002 Salpeter, E. E. 1955, ApJ, 121, 161, doi: 10.1086/145971
Donnan, C. T., McLeod, D. J., Dunlop, J. S., et al. 2023, Seeyave, L. T. C., Wilkins, S. M., Kuusisto, J. K., et al.
MNRAS, 518, 6011, doi: 10.1093/mnras/stac3472 2023, MNRAS, 525, 2422, doi: 10.1093/mnras/stad2487
Donnan, C. T., McLure, R. J., Dunlop, J. S., et al. 2024,
Simmonds, C., Tacchella, S., Hainline, K., et al. 2024,
arXiv e-prints, arXiv:2403.03171,
MNRAS, 527, 6139, doi: 10.1093/mnras/stad3605
doi: 10.48550/arXiv.2403.03171
Springel, V. 2005, MNRAS, 364, 1105,
Eldridge, J. J., Stanway, E. R., Xiao, L., et al. 2017, PASA,
doi: 10.1111/j.1365-2966.2005.09655.x
34, e058, doi: 10.1017/pasa.2017.51
Stanway, E. R., & Eldridge, J. J. 2018, MNRAS, 479, 75,
Fan, X. 2006, NewAR, 50, 665,
doi: 10.1093/mnras/sty1353
doi: 10.1016/j.newar.2006.06.077
Stanway, E. R., Eldridge, J. J., & Becker, G. D. 2016,
Finkelstein, S. L., Leung, G. C. K., Bagley, M. B., et al.
MNRAS, 456, 485, doi: 10.1093/mnras/stv2661
2023, arXiv e-prints, arXiv:2311.04279,
Sun, F., Egami, E., Pirzkal, N., et al. 2023, ApJ, 953, 53,
doi: 10.48550/arXiv.2311.04279
doi: 10.3847/1538-4357/acd53c
Fu, J., Kauffmann, G., Huang, M.-l., et al. 2013, MNRAS,
Tang, M., Stark, D. P., Chen, Z., et al. 2023, MNRAS, 526,
434, 1531, doi: 10.1093/mnras/stt1117
1657, doi: 10.1093/mnras/stad2763
Furlanetto, S. R., Oh, S. P., & Briggs, F. H. 2006, PhR,
Weinberger, L. H., Haehnelt, M. G., & Kulkarni, G. 2019,
433, 181, doi: 10.1016/j.physrep.2006.08.002
MNRAS, 485, 1350, doi: 10.1093/mnras/stz481
Götberg, Y., de Mink, S. E., McQuinn, M., et al. 2020,
A&A, 634, A134, doi: 10.1051/0004-6361/201936669 Whitler, L., Stark, D. P., Endsley, R., et al. 2024, MNRAS,
Han, Y., & Han, Z. 2019, ApJS, 240, 3, 529, 855, doi: 10.1093/mnras/stae516
doi: 10.3847/1538-4365/aaeffa Wilkins, S. M., Feng, Y., Di-Matteo, T., et al. 2016,
Henriques, B. M. B., White, S. D. M., Thomas, P. A., et al. MNRAS, 458, L6, doi: 10.1093/mnrasl/slw007
2015, MNRAS, 451, 2663, doi: 10.1093/mnras/stv705 Yang, X., Mo, H. J., & van den Bosch, F. C. 2003, MNRAS,
Henriques, B. M. B., Yates, R. M., Fu, J., et al. 2020, 339, 1057, doi: 10.1046/j.1365-8711.2003.06254.x
MNRAS, 491, 5795, doi: 10.1093/mnras/stz3233 Yung, L. Y. A., Somerville, R. S., Finkelstein, S. L.,
Hsiao, T. Y.-Y., Abdurro’uf, Coe, D., et al. 2023, arXiv Popping, G., & Davé, R. 2019, MNRAS, 483, 2983,
e-prints, arXiv:2305.03042, doi: 10.1093/mnras/sty3241
doi: 10.48550/arXiv.2305.03042 Yung, L. Y. A., Somerville, R. S., Finkelstein, S. L., et al.
Hutter, A., Dayal, P., Yepes, G., et al. 2021, MNRAS, 503, 2020a, MNRAS, 496, 4574, doi: 10.1093/mnras/staa1800
3698, doi: 10.1093/mnras/stab602 Yung, L. Y. A., Somerville, R. S., Popping, G., &
Kannan, R., Garaldi, E., Smith, A., et al. 2022, MNRAS, Finkelstein, S. L. 2020b, MNRAS, 494, 1002,
511, 4005, doi: 10.1093/mnras/stab3710 doi: 10.1093/mnras/staa714
Krumholz, M. R., McKee, C. F., & Tumlinson, J. 2009, Zhang, F., Han, Z., Li, L., & Hurley, J. R. 2004, A&A, 415,
ApJ, 693, 216, doi: 10.1088/0004-637X/693/1/216 117, doi: 10.1051/0004-6361:20031268
Legrand, L., Dayal, P., Hutter, A., et al. 2023, MNRAS, —. 2005, MNRAS, 357, 1088,
519, 4564, doi: 10.1093/mnras/stac3760 doi: 10.1111/j.1365-2966.2005.08739.x
12

Zhang, F., Li, L., Han, Z., Zhuang, Y., & Kang, X. 2013, Zheng, Z., Berlind, A. A., Weinberg, D. H., et al. 2005,
MNRAS, 428, 3390, doi: 10.1093/mnras/sts280 ApJ, 633, 791, doi: 10.1086/466510