MNRAS 000, 1–12 (2023) Preprint 24 March 2023 Compiled using MNRAS LATEX style file v3.

Stellar cluster formation in a Milky Way-sized galaxy at 𝑧 > 4 – II. A

hybrid formation scenario for the nuclear star cluster and its connection to
the nuclear stellar ring
Floor van Donkelaar,1★ Lucio Mayer,1 Pedro R. Capelo,1 Tomas Tamfal,1 Thomas R. Quinn2
and Piero Madau3
1 Center for Theoretical Astrophysics and Cosmology, Institute for Computational Science, University of Zurich,
arXiv:2303.12828v1 [astro-ph.GA] 22 Mar 2023

Winterthurerstrasse 190, CH-8057 Zürich, Switzerland

2 Astronomy Department, University of Washington, Seattle, WA 98195, USA
3 Department of Astronomy and Astrophysics, University of California, 1156 High Street, Santa Cruz, CA 95064, USA

Accepted XXX. Received YYY; in original form ZZZ

ABSTRACT
Nuclear star clusters (NSCs) are massive star clusters found in the innermost region of the majority of galaxies. While recent
studies suggest that low-mass NSCs in dwarf galaxies form largely out of the merger of globular clusters and NSCs in massive
galaxies have assembled most of their mass through central star formation, the formation channel of the Milky Way’s NSC is still
uncertain. In this work, we use GigaEris, a very high resolution 𝑁-body hydrodynamical cosmological “zoom-in” simulation, to
investigate NSC formation in the progenitor of a Milky Way-sized galaxy, as well as its relation to the assembly and evolution of
the galactic nuclear region. We study the possibility that bound, young, gas-rich, stellar clusters within a radius of 1.5 kpc of the
main galaxy’s centre at 𝑧 > 4 are the NSC predecessors (NSCPs). We identify 53 systems which satisfy our criteria, with a total
baryonic mass of 107.7 M . They have a relatively low mean stellar metallicity (−0.47 . [Fe/H] . −0.11) in comparison to
the present-day stars in the Milky Way’s NSC. The NSCPs with a ‘born thin-disc’ star fraction, 𝐹thin , higher than 0.5 are older
and display slightly different properties than the clusters with 𝐹thin ≤ 0.5. We demonstrate that both stellar cluster accretion and
in-situ star formation will contribute to the formation of the NSC, providing evidence for an hybrid formation scenario for the
first time in an 𝑁-body, hydrodynamical, cosmological “zoom-in” simulation. Furthermore, we also identify a nuclear stellar
ring in the simulation, with properties similar to those of the Milky Way’s nuclear stellar disc.
Key words: galaxies: formation – galaxies: high-redshift – galaxies: nuclei – methods: numerical

1 INTRODUCTION disc-like, structures in NSCs that are well-aligned with the disc of
their host galaxies (Seth et al. 2006). The NSCs with a higher stellar
The Galactic Centre of the Milky Way (MW) is an excellent lab-
mass tend to be more flattened than lower-mass NSCs (e.g. Spengler
oratory for studying phenomena and physical processes that may
et al. 2017; Georgiev & Böker 2014).
be occurring in many other galactic nuclei. The MW’s nuclear star
cluster (NSC) and nuclear stellar disc (NSD) are the main features NSCs exist in very different host environments (Neumayer et al.
of the Galactic Centre. Nevertheless, their observation is hampered 2020), which raises the question of whether NSC formation is con-
by the extreme source crowding and high extinction. Hence, their trolled by similar processes in all galaxy types, or if NSCs follow
relation and formation scenario are not fully clear yet (Schödel 2021; evolutionary paths that depend on the properties of their host galaxy.
Nogueras-Lara et al. 2021). As the NSC is one the main features of the MW’s Galactic Centre,
NSCs are extremely dense and massive star clusters occupying its formation path is of exceptional interest for the understanding
the innermost region of a majority of galaxies of all types (e.g. of physical processes that occur in the central region of the MW.
Carollo et al. 1997; Matthews et al. 1999; Böker et al. 2002; Côté The MW’s NSC extends up to hundreds of arcseconds across from
et al. 2006). They are more luminous than globular clusters (GCs; the central supermassive black hole and is believed to have a mass
e.g. Böker 2010) and have masses of the order of ∼104 –109 M of ∼107.4 M (e.g. Launhardt et al. 2002; Schödel et al. 2014;
(Walcher et al. 2005; Fahrion et al. 2020, 2021) and effective radii Feldmeier-Krause et al. 2017).
of the order of 1–20 pc (see Neumayer et al. 2020, and references The formation and evolution path of the MW’s NSC are still un-
therein). Many NSCs appear to be non-spherical. This is supported known. There are two main hypotheses that have been suggested for
by observations of edge-on spirals which identified elongated, i.e. the formation of the cluster: (a) through GC accretion; (b) through in-
situ star formation (SF; Neumayer et al. 2020). In the GC-accretion
scenario, the NSC forms out of the gas-free merger of GCs that spiral
★ [email protected] into a galaxy’s centre due to dynamical friction (e.g. Tremaine et al.

© 2023 The Authors

2 F. van Donkelaar et al.
1975; Capuzzo-Dolcetta 1993; Capuzzo-Dolcetta & Miocchi 2008; and a nuclear spiral (Schödel 2021). Nevertheless, the fact that the
Agarwal & Milosavljević 2011; Arca-Sedda & Capuzzo-Dolcetta MW has not had any major merger in the past ∼10 Gyr (e.g Wyse
2014; Gnedin et al. 2014, see also Clarke et al. 2019 for clump ac- 2001; Helmi et al. 2018; Renaud et al. 2021; Sotillo-Ramos et al.
cretion). An NSC formed through GC accretion is expected to reflect 2022) makes the existence of a nuclear disc, such as those observed
properties typical of GCs, which are in general characterised by sim- in S0 galaxies, still a possibility in the MW. The bulk of stars in the
ple SF histories (SFHs), low metallicity, and a high fraction of old MW’s NSD is old and formed at least 8 Gyr ago, followed first by a
stars. The alternative formation path, in-situ SF, considers an NSC phase of quiescence and then by recent SF activity (about 1 Gyr ago,
to form directly at the galactic centre out of star-forming gas (e.g. when 5 per cent of the mass of the NSD was formed very quickly;
Milosavljević 2004; McLaughlin et al. 2006; Bekki 2007). In this Nogueras-Lara et al. 2019, 2020a, 2021, 2023; Schödel et al. 2020).
nuclear SF scenario, gas falls into the nucleus and then transforms Galactic bars can lead to the creation of substructures in the nuclear
into stars (Loose et al. 1982). As SF can proceed in several episodes region of disc galaxies such as NSDs and NSRs by redistributing
during a galaxy’s evolution, an NSC formed in this way will exhibit angular momentum (Combes & Gerin 1985). There is now growing
a more complex SFH and is expected to have a low mass fraction evidence that these substructures in the galactic centre are built from
of old and metal-poor stellar populations in comparison to NSCs gas that was funnelled to the centre by the bar (e.g. Contopoulos &
that formed through GC accretion (e.g. Antonini & Perets 2012; Grosbol 1989; Binney et al. 1991; Knapen 1999; Kim et al. 2011;
Feldmeier-Krause et al. 2015; Arca Sedda et al. 2020; Fahrion et al. Fragkoudi et al. 2016). Another formation scenario of NSDs could
2020, 2021, 2022). be galaxy mergers (e.g. Mayer et al. 2008); these, however, fail to
Both processes may contribute with different weights in differ- reproduce the sizes of typically observed NSDs in nearby galaxies
ent galaxies, meaning that both mechanisms could contribute to the (Schultheis et al. 2021).
build-up of NSCs over a Hubble time (see, e.g. Guillard et al. 2016), Moreover, since we do not detect an age-metallicity relation in the
through a hybrid formation scenario. The infall of GCs alone is not a solar neighbourhood, clusters of stars are expected to undergo radial
viable formation scenario for the NSCs in massive early-type galax- migration (Sellwood & Binney 2002; Haywood 2008; Roškar et al.
ies with stellar mass 𝑀★ > 109 M , as the observed metallicity is 2008; Schönrich & Binney 2009). Therefore, one could expect that
too high and the SFH shows an ongoing process (Walcher et al. 2005; gas-rich stellar clusters migrated from the outer parts of the Galaxy
Fahrion et al. 2019; Pinna et al. 2021). However, mergers and accre- towards the Galactic Centre. For example, in Schönrich & Binney
tion of gas-rich young stellar clusters (Figer et al. 2002; Paudel & (2009), the stars are trapped on to a resonant co-rotation with spiral
Yoon 2020) could still provide a viable formation mechanism. Such arms and may migrate inwards and outwards along the spiral waves.
a formation pathway can lead to similar stellar population properties The thin disc is expected to start forming the earliest assembly stage
as those observed in the MW’s NSC. of a galaxy (e.g. Tamfal et al. 2022, see also Agertz et al. 2021;
In spite of its proximity, the observation of the MW’s NSC is re- Beraldo e Silva et al. 2021; Kretschmer et al. 2022; Van Donkelaar
strained by the extreme source crowding and the high interstellar ex- et al. 2022b for the early formation of a thin-disc component through
tinction that limits its analysis to the infrared regime (e.g. Nishiyama co-formation of the discs), therefore a significant population of old
et al. 2008; Schödel et al. 2010; Nogueras-Lara et al. 2018, 2020b). thin-disc stars could have influenced the formation of the NSC and
Consequently, the relation between the NSC and the NSD is also NSD.
not well understood yet (see, e.g. Launhardt et al. 2002; Schödel From a theoretical point of view, hydrodynamic simulations of
et al. 2020; Nogueras-Lara et al. 2020a). There is some evidence MW analogs have shown that the formation of the bar can trigger gas
that the NSC and the NSD may host different stellar populations funnelling to the centre of the MW, forming a kinematically cold, ro-
with different SFHs (e.g. Nogueras-Lara et al. 2020a; Schödel et al. tating NSD (e.g. Fux 1999; Li et al. 2015; Ridley et al. 2017; Sormani
2020; Schultheis et al. 2021). Both components seem to have a pre- & Barnes 2019; Tress et al. 2020; Moon et al. 2021; Sormani et al.
dominantly old stellar population, with the initial starburst followed 2022). Additionally, a large number of simulations and semi-analytic
by several billion years of quiescence. Nevertheless, there is some models have calculated the efficiency of dynamical friction for a
evidence for a ∼3 Gyr old intermediate-age population in the NSC, range of starting conditions for GC systems and their host galaxies,
which cannot be found in the NSD. On the other hand, there is evi- confirming that dynamical friction provides a plausible mechanism
dence for a ∼1 Gyr old SF event associated to the NSD, that is not to form an NSC (e.g. Capuzzo-Dolcetta 1993; Lotz et al. 2001). De-
found when analysing the stellar population of the NSC (see Schödel tailed 𝑁-body simulations of the infall of GCs through dynamical
et al. 2020; Nogueras-Lara et al. 2020a, 2021). Because of the differ- friction have shown consistency with the flattening and the kinematic
ences, understanding the formation paths of both systems can help properties of observed NSCs, but that matching the kinematics and
us interpret the physical processes that occur in galactic nuclei. luminosity function of the stellar clusters likely requires roughly half
This was initiated by Becklin et al. (1982), who showed that the of the NSC mass to come from in-situ SF and accreted gas (e.g.
region within 2 pc of the MW’s Galactic Centre is largely devoid Hartmann et al. 2011; Antonini & Perets 2012; Tsatsi et al. 2017).
of interstellar matter and is surrounded by a dust ring or disc. More Furthermore, semi-analytic models that follow the evolution of GC
recent observations and simulations have determined this to be an systems with dynamical friction to track the growth of the NSCs have
NSD which includes stars and is star-forming (e.g. Launhardt et al. been able to reproduce some of the properties, like mass and radius,
2002; Kim & Lee 2012; Schultheis et al. 2021). The NSD of the MW of both the present-day NSCs and the GC systems (e.g. Antonini
extends up to a radius of 220 pc with a scale height of ∼50 pc (Pierce- 2013; Gnedin et al. 2014). Bekki (2010) carried out 𝑁-body simula-
Price et al. 2000; Launhardt et al. 2002; Nogueras-Lara et al. 2020a; tions of the orbital decay of stellar clusters in the background of field
Gallego-Cano et al. 2020). NSDs are also detected in extragalactic stars in a disc galaxy embedded in a dark matter (DM) halo, finding
systems and are quite common in early-type galaxies (Pizzella et al. that NSCs could have formed from stars delivered by inspiralling
2002; Gadotti et al. 2019). However, given that no NSDs have been stellar clusters.
clearly identified in late-type galaxies and provided that the MW is a In this work, we aim to pinpoint the formation channels of the
barred spiral galaxy, it is plausible that the NSD of the MW is rather a MW’s NSC with an 𝑁-body, hydrodynamical, cosmological “zoom-
mixture of a nuclear star-forming ring (the nuclear stellar ring; NSR) in” simulation of unprecedented resolution, GigaEris (Tamfal et al.

MNRAS 000, 1–12 (2023)

 Hybrid nuclear star cluster formation 3
2022). Additionally, we explore the relation between the MW’s NSC discussion), using tabulated rates from Cloudy (Ferland et al. 1998,
and its NSD, and how their formation may have been connected. For 2013) and following the method described in Shen et al. (2010, 2013).
that reason, we delve into the properties of young, gas-rich, stellar Feedback from supernovae SNae Type Ia is implemented by in-
clusters at 𝑧 > 4 in the nuclear region of an MW-sized galaxy. The jecting energy and a fixed amount of mass and metals, independent
layout of this paper is the following: Section 2 briefly summarises of the progenitor mass, into the surroundings (see Thielemann et al.
the simulation setup and describes how we identified the clusters. In 1986; Stinson et al. 2006), whereas SNae Type II (SNII) feedback is
Section 3, we present the simulation results, first with a focus on the implemented following the delayed-cooling recipe of Stinson et al.
properties of the possible NSC predecessors (NSCPs) at 𝑧 = 4.4 and (2006), with metals and energy, 𝜖 SF = 1051 erg, being injected per
the possibility that these stellar clusters will form an NSC with the event into the interstellar medium as thermal energy, according to
properties of the MW’s NSC, and then presenting the NSR within the ‘blastwave model’ of Stinson et al. (2006). For each SNII event, a
the simulation and its link to the NSC. Finally, we discuss our results given amount of oxygen and iron mass, dependent on the mass of the
in Section 4 and conclude in Section 5. star, is injected into the surrounding gas (Woosley & Weaver 1995;
Raiteri et al. 1996). The stars with masses between 8 and 40 M  will
explode as SNII, whereas stars with masses between 1 and 8 M  do
not explode as SNae but release part of their mass as stellar winds,
2 METHODS with the returned gas having the same metallicity of the low-mass
stars.
2.1 Simulation code and initial conditions
We base our analysis on an 𝑁-body, hydrodynamical, cosmological
“zoom-in” simulation of a MW-like galaxy, GigaEris (Tamfal et al. 2.2 Cluster finding
2022), carried out with the 𝑁-body smoothed-particle hydrodynam-
We use the adaptive mesh grid AMIGA Halo Finder (AHF; Gill
ics (SPH) code ChaNGa (Jetley et al. 2008, 2010; Menon et al.
et al. 2004; Knollmann & Knebe 2009) to identify the gas-rich stellar
2015). A brief summary of the numerical recipes is provided below;
substructures that possibly could be connected to the formation of
one can find a more detailed discussion on the set-up in Tamfal et al.
the NSCs in the final simulation step at 𝑧 = 4.4. The clusters were se-
(2022).
lected in an identical way as in Van Donkelaar et al. (2022a; hereafter
The GigaEris simulation follows a Galactic-scale halo identified in
Paper I), namely with a minimum threshold of 64 baryonic particles
a low-resolution, DM-only simulation at 𝑧 = 0 in a periodic cube of
in the virial radius per cluster and 0 subclusters inside the identified
side 90 cMpc. It was chosen to have a similar mass as that of the MW
cluster. Furthermore, to validate that the clusters are bound, we have
and a rather quiet late merging history. This method is similar to the
calculated the binding energies of the identified clusters and unbound
way the galaxy halo was selected in the original Eris suite in Guedes
clusters were removed from the set. Hereafter the word cluster in this
et al. (2011). Next, the selected halo was re-simulated at several orders
paper refers to a bound object found by AHF.
of magnitude higher resolution than the DM-only simulation, adding
gas particles as well as the necessary short-wavelength modes. The
initial conditions were generated with the MUSIC code (Hahn & Abel
2011), with 14 levels of refinement and the cosmological parameters 3 RESULTS
Ωm = 0.3089, Ωb = 0.0486, ΩΛ = 0.6911, 𝜎8 = 0.8159, 𝑛s = 0.9667,
and 𝐻0 = 67.74 km s−1 Mpc−1 (see Planck Collaboration et al. 3.1 Classifying stellar systems
2016). The gravitational softening of all particles is set to a constant To investigate the possibility that gas-rich stellar clusters at 𝑧 > 4
in physical coordinates (𝜖 c = 0.043 kpc) for redshifts smaller than are NSCP candidates within our simulation, we extract all AHF
𝑧 = 10 and otherwise evolves as 𝜖 = 11𝜖 c /(1 + 𝑧). For the final identified substructures in the simulated box with a baryonic mass
snapshot at 𝑧 = 4.4, the DM, gas, and stellar particle numbers in the range 104 –108 M . For each of the substructures, the total baryonic
entire simulation box are 𝑛DM = 5.7 × 108 , 𝑛gas = 5.2 × 108 , and mass fraction,
𝑛★ = 4.4 × 107 , respectively.
Each star particle is created stochastically with an initial mass
𝑀★ + 𝑀gas
of 𝑚★ = 1026 M using a simple gas density and temperature 𝐹b = , (2)
threshold criterion (Stinson et al. 2006), with 𝑛SF > 100 𝑚 H cm-3 𝑀total
and 𝑇SF < 3 × 104 K, and the gas particle that spawns the new star and stellar mass fraction,
has its own mass reduced accordingly. A star particle represents an
entire stellar population with its own Kroupa (2001) initial stellar 𝑀★
mass function (IMF). The SF proceeds at a rate which is given by 𝐹★ = , (3)
𝑀★ + 𝑀gas

d𝜌★ 𝜌gas are calculated, where 𝑀total is the sum of the baryonic and DM
= 𝜖 SF , (1) masses, 𝑀★ the stellar mass, and 𝑀gas the gas mass, all computed
dt 𝑡 dyn
at half the virial radius. This approach is similar to that of Paper I.
with 𝜌★ indicating the stellar density, 𝜌gas the gas density, 𝑡dyn the The distance away from the galactic centre of the main galaxy of
local dynamical time, and 𝜖 SF the SF efficiency, which is set to 0.1. the identified clusters is plotted against their baryonic mass fraction
The code solves for the non-equilibrium abundances and cooling of in the top panel of Figure 1. In Paper I, we selected all clusters
H and He species (assuming self-shielding and a redshift-dependent with an 𝐹b ≥ 0.75 as proto-GC systems (excluding systems with a
radiation background; Pontzen et al. 2008; Haardt & Madau 2012), 𝜎★ < 20 km s−1 ): this region has been indicated by the gray-shaded
whereas the cooling from the fine structure lines of metals is calcu- area in the Figure. In the top panel, we see a new group of interesting
lated in photoionization equilibrium from the same radiation back- clusters, all being very close to the galactic centre of the main galaxy
ground (assuming no self-shielding; see Capelo et al. 2018 for a (𝑟 < 1.5 kpc), with a relatively high baryonic fraction (𝐹b > 0.35),

MNRAS 000, 1–12 (2023)

4 F. van Donkelaar et al.
Stellar Metallicity [Fe/H] 14
2.5 2.0 1.5 1.0 0.5 0.0 (r)
Hernquist profile
13

12

log( ) [M kpc 3]
11
101
10
r [kpc]

100 8
104 < Mb < 105
105 < Mb < 106
106 < Mb < 107 2 1 0
107 < Mb < 108 log(R) [kpc]
10 2 10 1 100 Figure 2. Stellar density profile of the central region of our simulated galaxy
Fb at 𝑧 = 4.4 (blue, solid line). The black, dashed line shows the Hernquist profile
1.0 with a scale radius 𝑎 = 78 pc and a total stellar mass 𝑀b,★ = 1010.49 M  .

product. Their baryonic mass fraction can change during their evo-
0.8 lution, especially when they fall into the high-density central region
of the galaxy. This is for example similar to what happened to “The
Imposter” in Paper I, which lost all of its DM while spiralling into
0.6 the MW analog.
F

0.4 3.1.1 Dynamical friction time-scale

In an attempt to estimate the time taken by a given possible NSCP to
decay to the centre of the galaxy, we first approximate the inner parts
0.2 of the stellar density profile with the Hernquist (1990) analytical
model. This can be done, as the central stellar region of the main
0.4 0.6 0.8 1.0 galaxy halo follows quite well the Hernquist profile, as shown in
Fb Figure 2. The solid, blue line shows the central stellar density profile
when assuming a spherically averaged distribution of the galaxy
Figure 1. Top panel: the object’s distance away from the galactic centre of
the main galaxy halo plotted against its baryonic fraction, 𝐹b , for all halos between 0 and 2 kpc, whereas the black, dashed line shows the
identified with a baryonic mass range 104 –108 M  at 𝑧 = 4.4. The sizes Hernquist profile with a scale radius of 78 pc and a total enclosed
of the markers indicate the mass range of the clusters. Bottom panel: the stellar mass of 𝑀b,★ = 1010.49 M .
stellar fraction, 𝐹★ , of the 56 clusters within the red circle of the top panel By taking into account the effects of dynamical friction in a fic-
is plotted against the baryonic fraction. The gray-shaded area indicates the titious stellar Hernquist bulge with a scale radius 𝑎 and total mass
region where 𝐹b ≥ 0.75 (one of the criteria used to categorize a cluster 𝑀b,★, the corresponding time-scale for a stellar cluster of total stellar
as a possible proto-GC; see Paper I). The colour bar represents the stellar mass 𝑀★ computed at half the virial radius in circular motion inside
metallicity of the clusters. such a bulge to decay from an initial radial distance 𝑟 i to a final one
𝑟 f is given by (Souza Lima et al. 2017; Tamburello et al. 2017)
and with a similar stellar metallicity (around solar values).1 We select √︄ ∫ 𝜒i 3/2
all bound clusters within this region, indicated by the red circle in 1 𝑎 3 𝑀b,★ 𝜒 ( 𝜒 + 3)
the top panel of the Figure, as possible NSCPs. 𝑡 DF = 𝑑𝜒, (4)
4𝜉 ln Λ 𝐺 𝑀b,★ 𝑀★ 𝜒f 𝜒+1
In the bottom panel of Figure 1, the stellar mass fraction is plotted
against the baryonic mass fraction for these 56 possible NSCPs. From with 𝜉 being a correction factor of order unity, ln Λ the Coulomb
this Figure, we can conclude that, even though this group of stellar logarithm, 𝐺 the gravitational constant, and 𝜒 = 𝑟/𝑎. From Figure 1,
clusters are in a similar region of the main galaxy and have a similar we can see that the selected clusters are located between ∼0.35 and
metallicity, they have a wide variety of stellar and baryonic mass 1.5 kpc from the galactic centre. Therefore, we can make the assump-
fractions. tion that 𝜒𝑖  1 (i.e. 𝑟 𝑖  𝑎), so that, additionally imposing 𝜒f = 0,
We have not used a minimum baryonic mass fraction as part of the integral of Equation (4) can be approximated as 2𝜒5/2 /5 and we
our selection criteria. This is because the NSCPs are not the final can write (Tamburello et al. 2017)

√︄
1 In this work, we compute the abundance ratios (e.g. [Fe/H] and [O/Fe]) 𝑀b,★ 𝜒5/2 𝑎3
𝑡 DF = . (5)
normalising them to the solar values provided by Asplund et al. (2009). 10𝜉 ln Λ𝑀★ 𝐺 𝑀b,★

MNRAS 000, 1–12 (2023)

 Hybrid nuclear star cluster formation 5
8 0.0
7
6 0.1

Iron-to-Hydrogen ratio [Fe/H]

5
0.2
4
N

3
0.3
2
1 0.4
0 10 1 100 101
tDF [Gyr] 0.5
Figure 3. The full dynamical friction time-scale distribution for all the iden-
tified possible NSCPs. The dashed vertical line indicates 12 Gyr, which is the 0.6
time between the last snapshot in the simulation, at 𝑧 = 4.4, and 𝑧 ∼ 0. Just
three systems out of 56 have a 𝑡DF > 12 Gyr.
0.25
The Coulomb logarithm, ln Λ, can be approximated as (Binney &
Tremaine 2008)
Oxygen-to-Iron Ratio [O/Fe] 0.20
𝑏 max 𝑏 max 𝑣 2typ
Λ≈ ≈ , (6)
𝑏 90 𝐺 𝑀★
where 𝑣 typ is the typical relative velocity, 𝑏 max is the maximum
0.15
impact parameter, and 𝑏 90 is the 90◦ deflection radius. Taking the
mean stellar mass, distance from the galactic centre, and typical
velocity of the clusters for 𝑀★, 𝑏 max , and 𝑣 typ , respectively, we a 0.10
value for the Coulomb logarithm of approximately 8.8. Lastly, we
set 𝜉 = 1 (as done in Souza Lima et al. 2017 and Tamburello et al.
2017). tDF < 12 Gyr
0.05 Fthin > 0.5
The resulting 𝑡 DF for the stellar clusters is shown in Figure 3. It Mean
shows that 53 of the 56 selected stellar clusters in Figure 1 will decay
104 105 106 107
to the centre within 12 Gyr (indicated by the dashed, vertical line), Stellar Mass [M ]
which is the time between the last snapshot in the simulation, at
𝑧 = 4.4, and 𝑧 ∼ 0. Moreover, the influence of the disc, bar, or other Figure 4. Abundance ratios (top panel: [Fe/H]; bottom panel: [O/Fe]) as
galactic structures have not been included in this model (see, e.g. a function of stellar mass. The blue dots indicate our new stellar cluster
Bar et al. 2022). Owing to the galactic bar, for example, a massive sample of 53 systems, defined by applying the selection criterion described
object with a similar inclination and distance away from the centre in Section 3.1 (𝑟 < 1.5 kpc and 𝐹b > 0.35) and additionally imposing
as our NSCPs can have its inspiral time-scale decreased, as shown 𝑡DF < 12 Gyr (Section 3.1.1). The dots with a black outline represent the
by Bortolas et al. (2020, 2022). They concluded that for in-plane stellar clusters with 𝐹thin > 0.5 (see Section 3.2.1). The gray horizontal lines
indicate the mean values.
perturbers there is a clear tendency to decrease the dynamical friction
time-scale, whereas for perturbers on arbitrary inclinations the effect
is stochastic, with both increases and decreases of the dynamical masses of all 53 NSCPs (hereafter, we exclude the three systems with
friction time-scale similarly possible. Note that our NSCPs are found 𝑡 DF > 12 Gyr), ∼107.5 M , is of the same order of magnitude as the
in a symmetric ring with a small inclination of ∼14◦ with respect to observed mass of the MW’s NSC (∼107.4 M , e.g. Schödel et al.
the galactic plane (see Section 3.4). Therefore, one would expect that 2014; Feldmeier-Krause et al. 2017), which reinforces the possibility
the decay time-scale would decrease when adding these structures to of these clusters being NSCPs. Nevertheless, one should note that 28
the model. We can thus conclude that most of the selected clusters per cent of this total stellar mass comes from one massive cluster.
from Figure 1 are consistent with the chosen definition of NSCPs. This cluster has, discarding the mass, average properties in compar-
ison to the other clusters and is therefore still included as a possible
NSCP. Furthermore, all the NSCPs are spherical clumps (as shown
3.2 Properties of the NSCPs
by the examples in Figure 5) and follow the radial density profiles of
We carry on by studying the properties of the selected stellar clusters the King (1972) model.
within 1.5 kpc of the centre of the main galaxy. Together with the dy- In the top panel of Figure 4, the stellar metallicity, [Fe/H], is
namical friction time-scale estimation, these properties will allow us plotted against the stellar mass for the selected sample of NSCPs.
to investigate the hypothesis that the clusters are indeed predecessors The metallicity of the possible NSCPs is spread between −0.47 and
of the NSC. Figure 4 shows how some properties of these selected −0.11, with a mean of −0.32. The clusters have a lower mean metal-
clusters relate to their stellar mass (𝑀★). The total sum of the stellar licity than what one would expect from today’s MW’s NSC, which

MNRAS 000, 1–12 (2023)

6 F. van Donkelaar et al.
M =1.97e+06 M M =1.89e+05 M M =5.48e+05 M
109
0.075 0.075 0.075
0.050 0.050 0.050
0.025 0.025 0.025 108

2
/M kpc
y/kpc

y/kpc

y/kpc
0.000 0.000 0.000
0.025 0.025 0.025 107
0.050 0.050 0.050
0.075 0.075 0.075
0.05 0.00 0.05 0.05 0.00 0.05 0.05 0.00 0.05 106
x/kpc x/kpc x/kpc

Figure 5. Stellar surface density maps of three selected NSCPs at 𝑧 = 4.4 centred at their centre of mass. Combining the surface density maps with the knowledge
that they follow the radial density profiles of the King model, we can assume they resemble spherical clumps. The total stellar mass of the clusters at this redshift
is given by the title of the plot. The gas mass of the shown clusters is between 105.1 . 𝑀gas [M  ] . 105.8 .

Mean age of the cluster [Myr] NSCPs is ∼0.15, whereas the mean [O/Fe] for the thin-disc stars in
50 100 150 200 250 300
the simulation is ∼0.13. Hence, from the [O/Fe] ratio in our sample,
we can deduce that most of the stars within the NSCPs, especially
the ones with a low [O/Fe] ratio, could have originated out of the thin
Fthin < 0.5
Fthin > 0.5 disc (see Section 3.2.1 for further reasoning).
All the NSCPs are still star-forming at 𝑧 = 4.4, which could
1 indicate that we are looking at a hybrid formation scenario of the
MW’s NSC if still star-forming clusters will inspiral in and bring the
star-forming gas to the galactic centre. For all clusters at 𝑧 = 4.4, the
z [kpc]

SF episodes are between 0.7 and 1.2 Gyr long, with a mean specific
0 SF rate (SFR) between 0.8 and 1.2 Gyr−1 . The fact that the NSCPs
are still star-forming again shows that the NSCPs are different from
the proto-GCs discussed in Paper I, as we expect a burst of SF for
1 GCs. At 𝑧 = 4.4, all clusters experience their highest SFR since birth,
1 with SFRs between 10−3.9 and 10−1.6 M  yr −1 .
1
x [kp 0 0
c] ]
1 1 y [kpc 3.2.1 Thin-disc stars
As shown in Figure 4, approximately half of the NSCPs have the
Figure 6. The location of the 53 NSCPs in the 𝑥-𝑦-𝑧-space, with the stellar fraction of “born thin-disc” stars higher than 0.5. The “born thin-
galactic angular momentum pointing in the +𝑧 direction. The black ‘×’ indi- disc” stars are defined by applying the DBSCAN (see Ester et al.
cates the centre of the main galaxy halo. The colour bar represents the mean 1996) clustering algorithm to our simulation at different time steps.2
age of the stars within the clusters. This way, sequences of mutually spatially connected particles are
identified with a process comparable to that of a scatter kernel in-
terpolation in SPH. The ID of the stars born in a thin disc is saved
is solar to super solar, and at the lower end of what is expected from and from this we can calculate the fraction of “born thin-disc” stars,
galaxies with a stellar mass above 109 M (−0.5 . [Fe/H] . 0.5; 𝐹thin . The NSCPs with 𝐹thin > 0.5, indicated with the black outline
e.g. Kacharov et al. 2018; Schödel et al. 2020). However, this is rea- in Figure 4, have a higher mean stellar metallicity ([Fe/H] ∼ −0.22 )
sonable at 𝑧 = 4.4, as our clusters are still star-forming at this redshift and a lower [O/Fe] ratio than the mean of the whole sample, which is
and thus the mean stellar metallicity of the stars in the NSCPs can expected from thin-disc stars (see, e.g. Franchini et al. 2021; Reddy
still be raised. Furthermore, assuming a hybrid formation scenario et al. 2006; Bertran de Lis et al. 2016). Combining the birth envi-
for the MW’s NSC, stars formed through central SF would also raise ronment determined by DBSCAN and the discrepancy between the
the mean metallicity over time. Comparing this to the metallicity of metallicity properties of the two populations of NSCPs in Figure 4,
proto-GCs in Paper I (−1.8 . [Fe/H] . −0.8), we can assume that we can confidently say that the stars identified using DBSCAN are
the metallicity of these selected stellar clusters is too high to be a indeed “born thin-disc” stars. We find that all NSCPs consist a min-
proto-GC and closer to what one would expect from the MW’s NSC. imum of 3 per cent out of “born thin-disc” stars and 48 per cent of
The bottom panel of Figure 4 shows the [O/Fe] ratio of the possible the clusters the “born thin-disc” stars make up more than half of their
NSCPs, all within 0.02 . [O/Fe] . 0.26. In the MW, thick-disc stars total stars at 𝑧 = 4.4.
have a larger oxygen abundance than thin-disc stars with the same The different properties of these two types of NSCPs could suggest
[Fe/H], as shown by, e.g. Franchini et al. (2021). For [Fe/H] = −0.5,
the mean abundance ratios are [O/Fe] = 0.36 ± 0.19 and 0.24 ± 0.07
for the thick- and thin-disc stars, respectively (see Reddy et al. 2006; 2 See Tamfal et al. (2022) for an explicit approach of identifying stars as
Bertran de Lis et al. 2016). The mean [O/Fe] ratio for the possible thin-disc stars.

MNRAS 000, 1–12 (2023)

 Hybrid nuclear star cluster formation 7
Mb [M ]
103 104 105 106 107 108

106
Gas Mass [M ]

105

tDF < 12 Gyr

104 Mgas = M
104 105 106 107
Stellar Mass [M ]
Figure 7. Gas mass as a function of stellar mass. The 53 dots indicate the Figure 8. Gas surface density map of the main galaxy at 𝑧 = 5.0, with the
stellar cluster sample defined by applying the selection criterion described velocity vectors of the gas overlaid. The galaxy is centred face-on on the disc
in Section 3.1 (𝑟 < 1.5 kpc and 𝐹b > 0.35) and additionally imposing stars from the main galaxy halo.
𝑡DF < 12 Gyr (Section 3.1.1). The colour bar represents the total baryonic
mass, 𝑀b , of the stellar clusters. The gray, dashed line displays the 1:1 ratio
between the stellar and gas mass. clusters will contribute to the formation of the NSC in our simulated
galaxy. Nevertheless, we have also shown that in-situ SF will be
needed to acquire a stellar metallicity that is consistent with that of
that the NSCPs with 𝐹thin > 0.5 formed under different conditions. the MW’s NSC.
This is further investigated in Figure 6, where the NSCPs are plot- For the in-situ SF formation scenario, gas will need to fall into
ted in the 𝑥-𝑦-𝑧-space with the stellar galactic angular momentum the nucleus and then transform into stars (e.g Loose et al. 1982;
pointing in the +𝑧 direction at 𝑧 = 4.4. The NSCPs form an ellip- Milosavljević 2004). From the stellar-gas mass relation shown in
tical ring around the galactic centre of the main galaxy halo. From Figure 7, one can conclude that the clusters with a high baryonic
the Figure, it is clear that the two populations of NSCPs, high- and mass still include gas. The NSCPs have a 𝐹★ between 0.19 and
low-𝐹thin , can be found in different regions in this elliptical ring 0.94, with a mean stellar mass fraction of 0.56. Because the NSCs
and there is barely any mixing between the two groups. The colour of massive galaxies show significant contributions from young and
bar in Figure 6 shows the mean age of the stars within the NSCPs, enriched populations (e.g. Kacharov et al. 2018; Pinna et al. 2021;
from which we can deduce that the high-𝐹thin NSCPs have an older Fahrion et al. 2021, 2022), we would assume that the gas from these
mean stellar age. We find a mean stellar age for high-𝐹thin NSCPs of clusters that is not yet converted into stars when the cluster reaches
250.2 ± 45.2 Myr. For low-𝐹thin NSCPs, we find a mean stellar age the galactic centre would become part of the gas reservoir used for
of 102.1 ± 45.1 Myr. Notably, the “born thin-disc” stars are the oldest the central SF. Combining this with the dynamical friction time-scale
stars within the stellar clusters. The mean age of the NSCP stars is from Figure 3, we can infer that high-mass, gas-rich stellar clusters
therefore correlated with where the cluster can be found within the will contribute to the gas reservoir in the centre of the galaxy, which
elliptical ring at 𝑧 = 4.4. Tantalizingly, there is no clear relation be- will contribute to in-situ SF of the NSC.
tween the mean age of the “born thin-disc” stars within the NSCPs
Using the previous assumption, we could obtain an estimate on
and the birth radius of the oldest star within the cluster, even though
the maximum amount of stellar mass that can assemble from gas
some clusters include stars born far outside the disc.
coming from infalling stellar clusters. When only considering the
As expected, the SFHs of the two NSCP groups, high- and stellar clusters that have a dynamical friction time of 𝑡DF ≤ 1 Gyr to
low-𝐹thin , are also different. The 𝐹thin > 0.5 NSCPs start form- still be gas-rich when reaching the galactic centre of the galaxy, we
ing stars earlier, consistent with that fact that these clusters are find a total of gas mass that can be used for SF in the central region
older. However, they also show an approximately constant SFR of to be 107.3 M . Combining this to the SFR of the clusters discussed
∼0.4 × 10−2 M yr −1 from 𝑧 . 6, with an increase in SFR around in Section 3.2, we find that a total gas mass between ∼106.1 and
𝑧 ∼ 4.5. On the contrary, the 𝐹thin ≤ 0.5 NSCPs have on average a 107.3 M  will be transformed into stars through in-situ SF. Assuming
lower SFR and show this increase in SF from 𝑧 = 5. that NSCPs are the main source of baryonic mass for the NSC, we
can obtain an estimate on the contribution of in-situ SF within an
MW-like galaxy for the formation of an NSC. We find that in-situ SF
3.3 The hybrid formation scenario
from gas of accreted clusters contributes to a maximum 30 per cent
As discussed in Section 1, both GC accretion and in-situ SF could of the final NSC cluster mass.
contribute to the build-up of NSCs over a Hubble time (see, e.g. Furthermore, in a barred galaxy system, it is expected that the
Guillard et al. 2016). The NSC of the MW is a prominent example, bar will efficiently funnel the gas toward to centre, where it could
as it shows both young stars likely formed in-situ and old, metal-poor settle onto an elliptical ring (see Section 3.4; see, e.g. Contopoulos
components that possibly were accreted from GCs (e.g. Antonini & & Grosbol 1989; Binney et al. 1991; Knapen 1999; Regan & Teuben
Perets 2012; Feldmeier-Krause et al. 2015, 2020; Arca Sedda et al. 2003; Li et al. 2017; Sormani et al. 2022; Levy et al. 2022). The
2020). We have shown so far that the process of infalling stellar simulation displays bar formation (as discussed in Tamfal et al. 2022):

MNRAS 000, 1–12 (2023)

8 F. van Donkelaar et al.
20
Centre
15
SFR [M yr 1]

10

20
Ring
15
SFR [M yr 1]

10

0.4 0.6 0.8 1.0 1.2

Time[Gyr]

Figure 9. The evolution of the SFR within the galactic centre and within the Figure 10. Gas surface density map of the main galaxy at 𝑧 = 4.4. The
NSR. We defined the galactic centre as a sphere of 100 pc around the centre dots represent all identified clusters within 𝑅 < 1.5 kpc. The stellar galactic
of mass of the main galaxy and the NSR to be the region between 200 pc and angular momentum is pointing in the +𝑦 direction.
1 kpc around the galactic centre of the simulated galaxy in the 𝑦-𝑧-plane and
with a maximum height of 500 pc.
stellar clusters and what one would identify as an NSR. From Section
3.1.1, we determined that most of the NSCPs will have fallen to the
this could therefore be another mechanism that will contribute to the galactic centre by 𝑧 = 0. However, this estimation did not account
gas reservoir in the centre needed for in-situ SF. Figure 8 displays the for mutual interactions between the stellar clusters. Consequently, at
gas surface density3 at 𝑧 = 5.0 with the velocity vectors of the gas 𝑧 = 0 there could still be a few of the stellar clusters in the spherical
overlaid. We can identify a couple of gas streams coming from the ring surrounding the NSC.
outer part of the galaxy towards the centre. However, further research The ring visible in Figure 10 has a larger radius and scale-height
is needed to determine the importance of this effect. than those of the present-day NSD of the MW (R ≈ 220 pc and h
The argument for a hybrid formation scenario for the NSC in an ≈ 50 pc; Launhardt et al. 2002; Nishiyama et al. 2013; Nogueras-Lara
MW-like galaxy is further supported by the evolution of the SFR et al. 2020a; Gallego-Cano et al. 2020). Nevertheless, the radius of the
within the galactic centre of the main galaxy, shown in the top panel NSR could decrease over time due to dynamical friction strengthened
of Figure 9, as the SFR is already happening within the central by the bar-induced resonances (see Bortolas et al. 2022). Addition-
region of the galaxy. Furthermore, because high-mass NSCs have a ally, a ring forming at the inner ends of bar-torque-produced dust
high ellipticity (e.g. Seth et al. 2006; Spengler et al. 2017; Georgiev lanes, as shown in Kim et al. (2012), will shrink in size by 10 to
& Böker 2014), one could infer that such ellipticity is more easily 20 per cent as collisions of dense clumps inside the ring take away
explained by in-situ SF from inflowing gas (see, e.g. Spengler et al. angular momentum from the ring. Therefore, we can assume that, at
2017). This is because elongation could originate from rotational 𝑧 = 0, the ring shown in Figure 10 will most likely have a radius and
flattening, which requires a dynamically important coherent angular scale-height closer to the MW’s NSD’s values.
momentum on the star-forming baryons, which is easier to maintain The MW’s NSD local orbital inclination relative to the disc’s
through dissipative accretion in the galactic nucleus than from non- midplane is between 7◦ and 14◦ , with a mean of ∼10◦ (Gillessen
dissipative merger dynamics of pre-existing stellar clusters. et al. 2009). The ring displayed in Figure 10 is on a ∼14◦ inclination
As a result, we have three channels that contribute to the total stellar relative to the disc stars, which is thus in correspondence with the
mass of the NSC: (i) gas-rich stellar cluster accretion brings in stars observations of the MW. Furthermore, both in the MW and the
formed outside of the galactic nucleus and (ii) contributes gas fuel simulation, the stellar orbits are on average not circular in the nuclear
to the central reservoir for subsequent in-situ SF; concurrently, (iii) ring and disc. The mean ellipticity of the MW’s nuclear disc inferred
galactic-scale non-axisymmetric structures such as bar and spiral by Bartko et al. (2009) is 0.37 ±0.07, whereas the ellipticity of the
structures funnel towards the centre diffuse gas (not previously in ring formed by the NSCPs in the simulation is 0.23.
clusters), which can then provide fuel for in-situ SF. The gas surface density plot in Figure 10 also depicts a gaseous
ring surrounding the galactic centre of the main galaxy. This is ex-
pected from NSRs, since these rings are most likely produced by the
3.4 The nuclear stellar ring
radial infall of gas caused by angular momentum loss as a conse-
In addition to the NSC, it is known that the MW has also an NSD- quence of their nonlinear interactions with an underlying stellar bar
like structure (e.g. Schödel 2021; Schultheis et al. 2021). Figure 10 potential (e.g. Combes & Gerin 1985; Athanassoula 1992; Buta &
projects the stellar clusters found within 1.5 kpc from the galactic Combes 1996; Patsis et al. 2010; Kim & Stone 2012). Tantalizingly,
centre at 𝑧 = 4.4 on the gas surface density of the galaxy. This the orbits of the stellar clusters are not aligned with this gaseous
shows that there is a striking resemblance between the location of the ring, which has an inclination relative to the disc stars of ∼14◦ . Nu-
merical simulations of SF in the nuclear ring find that the stellar
clusters and gaseous ring overlap, and have a similar inclination rel-
3 The depth of the surface density plots is the depth of the simulated box. ative to the disc’s midplane (e.g. Seo & Kim 2013, 2014; Moon et al.

MNRAS 000, 1–12 (2023)

 Hybrid nuclear star cluster formation 9
2021). This difference can be explained by the different formation the MW’s Galactic Centre were formed more than 5 Gyr ago. In
environments of the stellar clusters. The stellar clusters within the Section 3.3, we estimated the contribution of stellar mass that will
simulations of Seo & Kim (2013, 2014) were entirely formed within form after 𝑧 = 4.4 from the gas-rich infalling clusters to be between
the ring, whereas we show that most of the NSCPs in this simulation ∼106.1 and 107.2 M , thus between 5 and 50 per cent of the stellar
include stars that have migrated from the stellar disc towards the nu- mass contribution of the NSCPs will be in the form of in-situ SF.
clear region. Therefore, the inclination of the clusters relative to the Nevertheless, one could conclude from Figure 9 that the final mass
gaseous ring could be due to the different environmental effects both of the NSC will surpass the estimated mass of maximum 107.70 M ,
systems have experienced and differing initial angular momentum. when the in-situ SF at the centre from other gas is included. However,
The total gas mass of the NSR is 3.4 × 108 M , which is in line we assume that only part of the SF at the centre will contribute to the
with what is observed in NSRs (∼1 to 6 × 108 M ; Buta et al. 2000; NSC and that most of the stars will be part of the pseudoubulge (as
Benedict et al. 2002; Sheth et al. 2005; Schinnerer et al. 2006). The argued in Tamfal et al. 2022). Therefore, the estimated final mass is
bottom panel of Figure 9 plots the temporal evolution of the SFR in most likely still in the correct order of magnitude.
the ring. As expected from the literature, the SFR of the NSR is within From the estimated present-day stellar mass of 107.70 M , we can
a range of ∼1 to 20 M  yr−1 (Mazzuca et al. 2008; Comerón et al. also calculate roughly the final effective radius of the NSC at 𝑧 = 0 to
2010). We find an SFR in the nuclear ring that is slowly increasing be between 6.1 and 8.8 pc, using the size-mass relation for late-type
with time, with short bursts of SF with time intervals between the galaxies described by Georgiev et al. (2016). Published values for the
bursts of roughly ∼50 Myr. effective radius of the MW’s NSC fall within the range of 4.2 ± 0.4
to 7.2 ± 2.0 pc (e.g. Schödel et al. 2014; Fritz et al. 2016), hence the
estimated present-day radius of the simulated NSC in this paper is in
accordance with observations.
4 DISCUSSION
We explored the possibility of stellar clusters within 1.5 kpc of the
4.1 Radial migration of stars
main galaxy halo at 𝑧 > 4 to be NSCPs, and their link to the thin
disc and the NSR, through the use of the 𝑁-body, hydrodynamical, Most of the stars within the NSCPs were born at around ∼1 kpc
cosmological “zoom-in” simulation GigaEris (Tamfal et al. 2022). from the galactic centre. Nevertheless, there is a small group of stars
There is no guarantee that all the stellar clusters identified as NSCPs that were born further away. Before all the bound stars of the NSCPs
in this paper will become part of the NSC at present day, as the become part of the elliptical ring around the galactic centre, as shown
calculation in Section 3.1.1 is very rough. In the calculation, we in Figure 6 at 𝑧 = 4.4, the stars born far outside the galactic centre
assume a spherically averaged distribution of the galaxy following within these clusters must have migrated towards the centre of the
a Hernquist profile. This, for example, does not take the interaction galaxy.
between the stellar clusters into account or the influence of galactic We followed the stellar particles for a couple of NSCPs from
structures like the bar and spirals, meaning that in reality the time 𝑧 = 6.91 to 𝑧 = 4.4. We find that, on average, roughly ∼20 per cent of
frame for a cluster to reach the galactic centre could be very different. the stars were born outside the cluster and migrated “alone”4 towards
Nevertheless, as the goal of this paper is to show the possibility of the cluster. These stellar particles often formed in the thin disc. For
the MW’s NSC being formed using the hybrid scenario and the link most stars, the migration towards the centre occurs around 𝑧𝑣 ∼ 6.8.
with the thin-disc stars, making these assumptions for the dynamical In the galactic centre they will become part of the main NSCPs and
friction time-scale is reasonable. correspond on average to ∼0.05 per cent of the total final mass of the
Although we are unable to follow the evolution of the selected stel- NSCPs. The exact dynamics and kinematics of the radial migration
lar clusters to 𝑧 = 0 and, therefore, cannot make definitive statements at 𝑧 > 4 will be studied in a subsequent paper.
on whether these clusters are truly NSCPs, many of the discussed
properties are grossly consistent with the present-day NSC of the
MW. One clear difference from observations is the metallicity of the 4.2 Comparison to literature
stars in the NSCPs in our simulation, which on average is slightly
While our work is one of the first to discuss the formation of an
lower than the average metallicity of the stars in the MW’s NSC (e.g.
NSC using a high-resolution, cosmological “zoom-in” simulation,
Schödel et al. 2020). Nevertheless, as we argue for a hybrid forma-
our results are mostly consistent with those from previous numerical
tion scenario, the in-situ SF from the gas-rich infalling clusters will
simulations (e.g. Antonini 2013; Seo & Kim 2013, 2014; Gnedin
increase the average metallicity of the stars in this region.
et al. 2014; Li et al. 2015; Guillard et al. 2016; Tsatsi et al. 2017; Seo
In the case of late-type host galaxies such as our MW, both the GC
et al. 2019; Sormani et al. 2022). For example, Gnedin et al. (2014)
and in-situ SF scenario are expected to work in parallel, as supported
found that for the MW the most probable value for the mass of the
by their observed metallicity spreads and complex SFHs (e.g. Do
central cluster formed by the infall of GCs is 2–6 × 107 M , which
et al. 2015). However, the contribution of each mechanism to the
is similar to the total stellar mass of our NSCPs, 3.2 × 107 M .
predominant build-up process of the MW’s NSC is not clear. We
Furthermore, in our simulation we find an SFR in the nuclear ring
find that 60 per cent of the baryonic mass of the NSCPs is stellar.
that is slowly increasing with time, with short bursts of SF between ∼1
Using the 1.5 and 2 times solar metallicity models for the SFH of the
and 20 M yr−1 . The time intervals between the bursts are roughly
MW’s NSC described by Schödel et al. (2020), we can estimate that
∼50 Myr. This is consistent with Seo & Kim (2013), who found that
between 10 and 35 per cent of the total expected stellar mass will
the SFR in nuclear rings displays a single primary burst followed by
form after 𝑧 = 4.4. The total stellar mass at 𝑧 = 4.4 of the NSCPs
a few secondary bursts with a time interval between the bursts to be
in our simulation is 107.5 M . Thus, using the model described by
Schödel et al. (2020), we find that between 106.5 and 107.9 M
stellar mass will form after 𝑧 = 4.4, resulting in a total mass between 4 In the code, a stellar particle represents an entire stellar population, meaning
107.55 and 107.70 M at 𝑧 = 0. This is in line with Blum et al. that when one stellar particle migrates, it is in reality an ensemble of stars.
(2003), who found that approximately 80 per cent of the stars in The stars are thus never really alone.

MNRAS 000, 1–12 (2023)

10 F. van Donkelaar et al.
roughly ∼50–80 Myr. Interestingly, both Seo & Kim (2013) and Seo reservoir in the centre needed for in-situ SF. Conjointly, the galactic
et al. (2019) found that the SFR will decrease during the simulation, structures like the disc and the bar will funnel the gas towards to
whereas we see an increase in SFR over time. In Seo & Kim (2013), centre, which can also be used for in-situ SF.
the SFR decreases to small values after the second burst. This differ- • Assuming that all the gas from the high-mass infalling stellar
ence is probably due to the fact that they start with an infinitesimally clusters with a 𝑡 DF ≤ 1 Gyr will turn into stars within the NSCPs,
thin, rotating disc, which is unmagnetized and isothermal. We, on the we find that in-situ SF contributes to a maximum of 30 per cent of
other hand, used a high-resolution, cosmological “zoom-in” simula- the stellar mass of the NSC cluster, only including the baryonic mass
tion of an MW-sized galaxy cluster to investigate the NSR. It could added from the NSCPs, of an MW-like galaxy.
be that Seo & Kim (2013) showed what would happen after 𝑧 = 4.4, • We detect an NSR within the GigaEris simulation with a radius
as our simulation ends with a thin-disc-like structure. of ∼500 pc at 𝑧 = 4.4. This ring hosts the NSCPs at this redshift and
The nuclear ring detected in the simulation has a larger radius has periodic SF short bursts of ∼10 M yr−1 , with time intervals of
than that of the present-day NSD of the MW, but also larger than ∼50 Myr.
the simulations performed by for example Seo et al. (2019; less
than ∼0.6 kpc). Nevertheless, Li et al. (2017) used hydrodynamic
simulations with static stellar potentials to show that a nuclear ring ACKNOWLEDGEMENTS
forms only in models with a central object exceeding ∼1 per cent of
the total disc mass, and that the ring size increases almost linearly PRC, LM, and FvD acknowledge support from the Swiss National
with the mass of the central object. As the simulation already has a Science Foundation under the Grant 200020_207406. FvD would
dense central region, from which most of the stars will become part like to thank Katja Fahrion for initiating the idea that the clusters in
of the pseudoubulge, it opens the possibility that the presence of a the centre could be part of the NSC and useful discussion along the
massive compact star-forming region or pseudobulge would make a way.
ring large when it first forms. The formed NSR can even become
larger as it grows due to an addition of gas with larger angular
momentum from outer regions, as discussed in Seo et al. (2019). DATA AVAILABILITY
The data underlying this article will be shared on reasonable request
to the corresponding author.
5 CONCLUSIONS
Using a high-resolution, cosmological “zoom-in” simulation of an
MW-sized galaxy halo (GigaEris; Tamfal et al. 2022), we have shown REFERENCES
the possibility of stellar clusters within 1.5 kpc of the main galaxy
Agarwal M., Milosavljević M., 2011, ApJ, 729, 35
halo at 𝑧 > 4 to be NSCPs, and their connection to the thin disc.
Agertz O., et al., 2021, MNRAS, 503, 5826
Moreover, we explored the relationship between the NSC and the Antonini F., 2013, ApJ, 763, 62
NSD. Our main conclusions are as follows: Antonini F., Perets H. B., 2012, ApJ, 757, 27
• We define NSCPs as clusters within 1.5 kpc from the centre of Arca-Sedda M., Capuzzo-Dolcetta R., 2014, ApJ, 785, 51
Arca Sedda M., Gualandris A., Do T., Feldmeier-Krause A., Neumayer N.,
the main galaxy and a 𝑡 DF < 12 Gyr. The total stellar mass of the
Erkal D., 2020, ApJ, 901, L29
clusters together, ∼107.5 M , is in the same order of magnitude as
Asplund M., Grevesse N., Sauval A. J., Scott P., 2009, ARA&A, 47, 481
the observed mass of the MW’s NSC, ∼107.4 M . Athanassoula E., 1992, MNRAS, 259, 345
• NSCPs at 𝑧 = 4.4 have a relatively low stellar metallicity, Bar N., Danieli S., Blum K., 2022, ApJ, 932
−0.47 . [Fe/H] . −0.11, in comparison to the stars in the MW’s Bartko H., et al., 2009, ApJ, 697, 1741
NSC. Nevertheless, this can be increased by the in-situ SF later during Becklin E. E., Gatley I., Werner M. W., 1982, ApJ, 258, 135
the formation of the NSC. The detected stellar metallicity makes the Bekki K., 2007, Publ. Astron. Soc. Australia, 24, 77
selected star clusters in this paper very different from the proto-GCs Bekki K., 2010, MNRAS, 401, 2753
discussed in Paper I. Benedict G. F., Howell D. A., Jørgensen I., Kenney J. D. P., Smith B. J., 2002,
• The NSCPs have a low oxygen-to-iron ratio, 0.02 . [O/Fe] . AJ, 123, 1411
Beraldo e Silva L., Debattista V. P., Nidever D., Amarante J. A. S., Garver
0.26, which points to the fact that a fraction of the stars within
B., 2021, MNRAS, 502, 260
the clusters are born within the thin disc (e.g Reddy et al. 2006;
Bertran de Lis S., et al., 2016, A&A, 590, A74
Bertran de Lis et al. 2016; Franchini et al. 2021). Combining the Binney J., Tremaine S., 2008, Galactic Dynamics: Second Edition
birth environment of the stars with the oxygen-to-iron ratio, we have Binney J., Gerhard O. E., Stark A. A., Bally J., Uchida K. I., 1991, MNRAS,
determined that the NSCPs consist of a minimum of 3 per cent out 252, 210
of “born thin-disc” stars, and for 48 per cent of the clusters the “born Blum R. D., Ramírez S. V., Sellgren K., Olsen K., 2003, ApJ, 597, 323
thin-disc” stars make up more than half of the total stars at 𝑧 = 4.4. Böker T., 2010, in The Impact of HST on European Astronomy. p. 99
• The NSCPs form an elliptical ring around the galactic centre of (arXiv:0708.1093), doi:J51-67977
the main galaxy halo at 𝑧 = 4.4. The NSCPs with 𝐹thin > 0.5 can be Böker T., Laine S., van der Marel R. P., Sarzi M., Rix H.-W., Ho L. C., Shields
found in different regions than the low-𝐹thin clusters in this ring and J. C., 2002, AJ, 123, 1389
Bortolas E., Capelo P. R., Zana T., Mayer L., Bonetti M., Dotti M., Davies
have a higher iron-to-hydrogen ratio. Furthermore, the NSCPs with
M. B., Madau P., 2020, MNRAS, 498, 3601
𝐹thin > 0.5 have a higher mean age of stars in the cluster than the
Bortolas E., Bonetti M., Dotti M., Lupi A., Capelo P. R., Mayer L., Sesana
other clusters. A., 2022, MNRAS, 512, 3365
• The hybrid formation scenario for the NSC of an MW-like Buta R., Combes F., 1996, Fundamentals Cosmic Phys., 17, 95
galaxy is the most likely. This is the result of three channels that Buta R., Treuthardt P. M., Byrd G. G., Crocker D. A., 2000, AJ, 120, 1289
contribute to the total stellar mass of the NSC. Gas-rich stellar cluster Capelo P. R., Bovino S., Lupi A., Schleicher D. R. G., Grassi T., 2018,
accretion brings in stars formed outside of the NSC and adds to the gas MNRAS, 475, 3283

MNRAS 000, 1–12 (2023)

 Hybrid nuclear star cluster formation 11
Capuzzo-Dolcetta R., 1993, ApJ, 415, 616 Kretschmer M., Dekel A., Teyssier R., 2022, MNRAS, 510, 3266
Capuzzo-Dolcetta R., Miocchi P., 2008, ApJ, 681, 1136 Kroupa P., 2001, MNRAS, 322, 231
Carollo C. M., Stiavelli M., de Zeeuw P. T., Mack J., 1997, AJ, 114, 2366 Launhardt R., Zylka R., Mezger P. G., 2002, A&A, 384, 112
Clarke A. J., et al., 2019, MNRAS, 484, 3476 Levy R. C., et al., 2022, ApJ, 935, 19
Combes F., Gerin M., 1985, A&A, 150, 327 Li Z., Shen J., Kim W.-T., 2015, ApJ, 806, 150
Comerón S., Knapen J. H., Beckman J. E., Laurikainen E., Salo H., Martínez- Li Z., Sellwood J. A., Shen J., 2017, ApJ, 850, 67
Valpuesta I., Buta R. J., 2010, MNRAS, 402, 2462 Loose H. H., Kruegel E., Tutukov A., 1982, A&A, 105, 342
Contopoulos G., Grosbol P., 1989, A&ARv, 1, 261 Lotz J. M., Telford R., Ferguson H. C., Miller B. W., Stiavelli M., Mack J.,
Côté P., et al., 2006, ApJS, 165, 57 2001, ApJ, 552, 572
Do T., Kerzendorf W., Winsor N., Støstad M., Morris M. R., Lu J. R., Ghez Matthews L. D., et al., 1999, AJ, 118, 208
A. M., 2015, ApJ, 809, 143 Mayer L., Kazantzidis S., Escala A., 2008, Mem. Soc. Astron. Italiana, 79,
Ester M., Kriegel H.-P., Sander J., Xu X., et al., 1996, in kdd. pp 226–231 1284
Fahrion K., et al., 2019, A&A, 628, A92 Mazzuca L. M., Knapen J. H., Veilleux S., Regan M. W., 2008, ApJS, 174,
Fahrion K., et al., 2020, A&A, 634, A53 337
Fahrion K., et al., 2021, A&A, 650, A137 McLaughlin D. E., King A. R., Nayakshin S., 2006, ApJ, 650, L37
Fahrion K., et al., 2022, A&A, 667, A101 Menon H., Wesolowski L., Zheng G., Jetley P., Kale L., Quinn T., Governato
Feldmeier-Krause A., et al., 2015, A&A, 584, A2 F., 2015, Computational Astrophysics and Cosmology, 2, 1
Feldmeier-Krause A., Zhu L., Neumayer N., van de Ven G., de Zeeuw P. T., Milosavljević M., 2004, ApJ, 605, L13
Schödel R., 2017, MNRAS, 466, 4040 Moon S., Kim W.-T., Kim C.-G., Ostriker E. C., 2021, ApJ, 914, 9
Feldmeier-Krause A., et al., 2020, MNRAS, 494, 396 Neumayer N., Seth A., Böker T., 2020, A&ARv, 28, 4
Ferland G. J., Fabian A. C., Johnstone R. M., 1998, in American Astronomical Nishiyama S., Nagata T., Tamura M., Kandori R., Hatano H., Sato S., Sugitani
Society Meeting Abstracts. p. 38.13 K., 2008, ApJ, 680, 1174
Ferland G. J., et al., 2013, Rev. Mex. Astron. Astrofis., 49, 137 Nishiyama S., et al., 2013, ApJ, 769, L28
Figer D. F., et al., 2002, ApJ, 581, 258 Nogueras-Lara F., et al., 2018, A&A, 610, A83
Fragkoudi F., Athanassoula E., Bosma A., 2016, MNRAS, 462, L41 Nogueras-Lara F., et al., 2019, A&A, 631, A20
Franchini M., et al., 2021, AJ, 161, 9 Nogueras-Lara F., et al., 2020a, Nature Astronomy, 4, 377
Fritz T. K., et al., 2016, ApJ, 821, 44 Nogueras-Lara F., Schödel R., Neumayer N., Gallego-Cano E., Shahzamanian
Fux R., 1999, A&A, 345, 787 B., Gallego-Calvente A. T., Najarro F., 2020b, A&A, 641, A141
Gadotti D. A., et al., 2019, MNRAS, 482, 506 Nogueras-Lara F., Schödel R., Neumayer N., 2021, ApJ, 920, 97
Gallego-Cano E., Schödel R., Nogueras-Lara F., Dong H., Shahzamanian B., Nogueras-Lara F., Schultheis M., Najarro F., Sormani M. C., Gadotti D. A.,
Fritz T. K., Gallego-Calvente A. T., Neumayer N., 2020, A&A, 634, A71 Rich R. M., 2023, A&A, 671, L10
Georgiev I. Y., Böker T., 2014, MNRAS, 441, 3570 Patsis P. A., Kalapotharakos C., Grosbøl P., 2010, MNRAS, 409, L94
Georgiev I. Y., Böker T., Leigh N., Lützgendorf N., Neumayer N., 2016, Paudel S., Yoon S.-J., 2020, ApJ, 898, L47
MNRAS, 457, 2122 Pierce-Price D., et al., 2000, ApJ, 545, L121
Gill S. P. D., Knebe A., Gibson B. K., 2004, MNRAS, 351, 399 Pinna F., et al., 2021, ApJ, 921, 8
Gillessen S., Eisenhauer F., Trippe S., Alexander T., Genzel R., Martins F., Pizzella A., Corsini E. M., Morelli L., Sarzi M., Scarlata C., Stiavelli M.,
Ott T., 2009, ApJ, 692, 1075 Bertola F., 2002, ApJ, 573, 131
Gnedin O. Y., Ostriker J. P., Tremaine S., 2014, ApJ, 785, 71 Planck Collaboration et al., 2016, A&A, 594, A13
Guedes J., Callegari S., Madau P., Mayer L., 2011, ApJ, 742, 76 Pontzen A., et al., 2008, MNRAS, 390, 1349
Guillard N., Emsellem E., Renaud F., 2016, MNRAS, 461, 3620 Raiteri C. M., Villata M., Navarro J. F., 1996, A&A, 315, 105
Haardt F., Madau P., 2012, ApJ, 746, 125 Reddy B. E., Lambert D. L., Allende Prieto C., 2006, MNRAS, 367, 1329
Hahn O., Abel T., 2011, MNRAS, 415, 2101 Regan M. W., Teuben P., 2003, ApJ, 582, 723
Hartmann M., Debattista V. P., Seth A., Cappellari M., Quinn T. R., 2011, Renaud F., Agertz O., Read J. I., Ryde N., Andersson E. P., Bensby T., Rey
MNRAS, 418, 2697 M. P., Feuillet D. K., 2021, MNRAS, 503, 5846
Haywood M., 2008, MNRAS, 388, 1175 Ridley M. G. L., Sormani M. C., Treß R. G., Magorrian J., Klessen R. S.,
Helmi A., Babusiaux C., Koppelman H. H., Massari D., Veljanoski J., Brown 2017, MNRAS, 469, 2251
A. G. A., 2018, Nature, 563, 85 Roškar R., Debattista V. P., Quinn T. R., Stinson G. S., Wadsley J., 2008, ApJ,
Hernquist L., 1990, ApJ, 356, 359 684, L79
Jetley P., Gioachin F., Mendes C., Kale L. V., Quinn T., 2008, in 2008 IEEE Schinnerer E., Böker T., Emsellem E., Lisenfeld U., 2006, ApJ, 649, 181
International Symposium on Parallel and Distributed Processing. pp 1– Schödel R., 2021, in Tsuboi M., Oka T., eds, Astronomical Society of the
12, doi:10.1109/IPDPS.2008.4536319 Pacific Conference Series Vol. 528, New Horizons in Galactic Center
Jetley P., Wesolowski L., Gioachin F., Kalé L. V., Quinn T. R., 2010, in SC Astronomy and Beyond. p. 353
’10: Proceedings of the 2010 ACM/IEEE International Conference for Schödel R., Najarro F., Muzic K., Eckart A., 2010, A&A, 511, A18
High Performance Computing, Networking, Storage and Analysis. pp Schödel R., Feldmeier A., Kunneriath D., Stolovy S., Neumayer N., Amaro-
1–11, doi:10.1109/SC.2010.49 Seoane P., Nishiyama S., 2014, A&A, 566, A47
Kacharov N., Neumayer N., Seth A. C., Cappellari M., McDermid R., Walcher Schödel R., Nogueras-Lara F., Gallego-Cano E., Shahzamanian B., Gallego-
C. J., Böker T., 2018, MNRAS, 480, 1973 Calvente A. T., Gardini A., 2020, A&A, 641, A102
Kim S. S., Lee A., 2012, in Journal of Physics Conference Series. p. 012049, Schönrich R., Binney J., 2009, MNRAS, 396, 203
doi:10.1088/1742-6596/372/1/012049 Schultheis M., et al., 2021, A&A, 650, A191
Kim W.-T., Stone J. M., 2012, ApJ, 751, 124 Sellwood J. A., Binney J. J., 2002, MNRAS, 336, 785
Kim S. S., Saitoh T. R., Jeon M., Figer D. F., Merritt D., Wada K., 2011, ApJ, Seo W.-Y., Kim W.-T., 2013, ApJ, 769, 100
735, L11 Seo W.-Y., Kim W.-T., 2014, ApJ, 792, 47
Kim W.-T., Seo W.-Y., Kim Y., 2012, ApJ, 758, 14 Seo W.-Y., Kim W.-T., Kwak S., Hsieh P.-Y., Han C., Hopkins P. F., 2019,
King I. R., 1972, ApJ, 174, L123 ApJ, 872, 5
Knapen J. H., 1999, in Beckman J. E., Mahoney T. J., eds, Astronomical Soci- Seth A. C., Dalcanton J. J., Hodge P. W., Debattista V. P., 2006, AJ, 132, 2539
ety of the Pacific Conference Series Vol. 187, The Evolution of Galaxies Shen S., Wadsley J., Stinson G., 2010, MNRAS, 407, 1581
on Cosmological Timescales. pp 72–87 (arXiv:astro-ph/9907290) Shen S., Madau P., Guedes J., Mayer L., Prochaska J. X., Wadsley J., 2013,
Knollmann S. R., Knebe A., 2009, ApJS, 182, 608 ApJ, 765, 89

MNRAS 000, 1–12 (2023)

12 F. van Donkelaar et al.
Sheth K., Vogel S. N., Regan M. W., Thornley M. D., Teuben P. J., 2005, ApJ,
632, 217
Sormani M. C., Barnes A. T., 2019, MNRAS, 484, 1213
Sormani M. C., et al., 2022, MNRAS, 512, 1857
Sotillo-Ramos D., et al., 2022, MNRAS, 516, 5404
Souza Lima R., Mayer L., Capelo P. R., Bellovary J. M., 2017, ApJ, 838, 13
Spengler C., et al., 2017, ApJ, 849, 55
Stinson G., Seth A., Katz N., Wadsley J., Governato F., Quinn T., 2006,
MNRAS, 373, 1074
Tamburello V., Capelo P. R., Mayer L., Bellovary J. M., Wadsley J. W., 2017,
MNRAS, 464, 2952
Tamfal T., Mayer L., Quinn T. R., Babul A., Madau P., Capelo P. R., Shen S.,
Staub M., 2022, ApJ, 928, 106
Thielemann F. K., Nomoto K., Yokoi K., 1986, A&A, 158, 17
Tremaine S. D., Ostriker J. P., Spitzer L. J., 1975, ApJ, 196, 407
Tress R. G., Sormani M. C., Glover S. C. O., Klessen R. S., Battersby C. D.,
Clark P. C., Hatchfield H. P., Smith R. J., 2020, MNRAS, 499, 4455
Tsatsi A., Mastrobuono-Battisti A., van de Ven G., Perets H. B., Bianchini P.,
Neumayer N., 2017, MNRAS, 464, 3720
van Donkelaar F., Mayer L., Capelo P. R., Tamfal T., Quinn T. R., Madau P.,
2022a, arXiv e-prints, p. arXiv:2210.04915
van Donkelaar F., Agertz O., Renaud F., 2022b, MNRAS, 512, 3806
Walcher C. J., et al., 2005, ApJ, 618, 237
Woosley S. E., Weaver T. A., 1995, ApJS, 101, 181
Wyse R. F. G., 2001, in Funes J. G., Corsini E. M., eds, Astronomical Society
of the Pacific Conference Series Vol. 230, Galaxy Disks and Disk Galax-
ies. pp 71–80 (arXiv:astro-ph/0012270), doi:10.48550/arXiv.astro-
ph/0012270

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

MNRAS 000, 1–12 (2023)