Professional Documents
Culture Documents
Elena Gonz Alez Prieto Newlin C. Weatherford Giacomo Fragione Kyle Kremer Frederic A. Rasio
Elena Gonz Alez Prieto Newlin C. Weatherford Giacomo Fragione Kyle Kremer Frederic A. Rasio
Elena González Prieto,1 Newlin C. Weatherford,1 Giacomo Fragione,1 Kyle Kremer,2 and Frederic A. Rasio1
1 Center for Interdisciplinary Exploration & Research in Astrophysics (CIERA) and Department of Physics & Astronomy, Northwestern
University, Evanston, IL 60208, USA
2 TAPIR, California Institute of Technology, Pasadena, CA 91125, USA
ABSTRACT
Very massive stars (VMSs) formed via a sequence of stellar collisions in dense star clusters have
been proposed as the progenitors of massive black hole seeds. VMSs could indeed collapse to form
intermediate-mass black holes (IMBHs), which would then grow by accretion to become the super-
massive black holes observed at the centers of galaxies and powering high-redshift quasars. Previous
studies have investigated how different cluster initial conditions affect the formation of a VMS, in-
cluding mass segregation, stellar collisions, and binaries, among others. In this study, we investigate
the growth of VMSs with a new grid of Cluster Monte Carlo (CMC) star cluster simulations—the most
expansive to date. The simulations span a wide range of initial conditions, varying the number of
stars, cluster density, stellar initial mass function (IMF), and primordial binary fraction. We find a
gradual shift in the mass of the most massive collision product across the parameter space; in particu-
lar, denser clusters born with top-heavy IMFs provide strong collisional regimes that form VMSs with
masses easily exceeding 1000 M⊙ . Our results are used to derive a fitting formula that can predict the
typical mass of a VMS formed as a function of the star cluster properties. Additionally, we study the
stochasticity of this process and derive a statistical distribution for the mass of the VMS formed in
one of our models, recomputing the model 50 times with different initial random seeds.
sufficiently dense clusters, growth will start as a result of 2009), and pulsar acceleration measurements (Kızıltan
stellar collisions and mergers. As the mass of a merger et al. 2017).
product grows, so does its collision cross-section, result- The likelihood of a cluster forming a VMS depends
ing in even more collisions. This results in a positive on various physical properties, among which is the IMF,
feedback loop that can rapidly produce a VMS of hun- which remains poorly constrained to this day, especially
dreds to thousands of solar masses. at high stellar masses. Although many studies of GCs
To study VMS formation in star clusters, Gürkan et al. assume a canonical Kroupa IMF (Kroupa 2001), obser-
(2004) investigated the effects of mass segregation and vations suggest that it may not be universal (e.g., De
core collapse. This was accomplished through the imple- Marchi et al. 2007; Bartko et al. 2010; Haghi et al. 2017;
mentation of Monte Carlo simulations in systems where Wirth et al. 2022). Furthermore, several studies have
parameters such as the cluster density profile, stellar shown that the IMF strongly impacts the dynamical evo-
IMF, and initial star count were systematically varied. lution and survival of GCs (e.g., Chernoff & Weinberg
This study found that the mass of the collapsing core 1990; Chatterjee et al. 2017; Giersz et al. 2019; Weath-
was always close to ∼10−3 times that of the total clus- erford et al. 2021).
ter mass. Remarkably, this follows the observed corre- In particular, Weatherford et al. (2021) explored the
lation between central BH mass and total host mass in impact of the slope of the IMF (at the high-mass end)
many astrophysical environments (Ferrarese & Merritt on the compact object population. This study found
2000). Note, however, that Gürkan et al. (2004) did not that in addition to producing more BHs, clusters with a
account for the effects of stellar evolution. top-heavy IMF also produce substantially more binary
Freitag et al. (2006a,b) performed the first cluster sim- BH (BBH) mergers, especially those involving (or result-
ulations that included precise treatment of stellar colli- ing in) production of upper-mass-gap BHs (e.g., Spera
sions and followed the evolution of the cluster and for- & Mapelli 2017; Takahashi et al. 2018; Marchant et al.
mation of a collisional runaway. Particularly, Freitag 2019; Farmer et al. 2019) and IMBHs. The latter is due
et al. (2006b) studied runaway collisions in young star to three factors; top-heavy IMFs produce heavier stars
clusters by varying physical parameters such as cluster and therefore heavier BHs, but also lead to several times
mass, size, and initial concentration. These studies con- more stellar collisions—due to scaling of stellar radii and
firmed that when the core collapse timescale is shorter gravitational focusing with mass—and more hierarchical
than the main-sequence (MS) evolution timescale for the mergers.
most massive stars (t ≈ 3 Myr), the cluster will undergo Another physical parameter that influences the rate
a collisional runaway. However, these studies did not of dynamical interactions in star clusters is the primor-
incorporate the role of binaries in the runaway process, dial binary fraction (e.g., Heggie & Hut 2003; Chatterjee
which have an important role in the evolution of the et al. 2010; Fregeau & Rasio 2007). Since binaries have
cores of star clusters. a larger interaction cross-section than single stars, they
VMSs are often assumed to be progenitors of IMBHs offer a larger effective area for encounters to take place.
(e.g., Ebisuzaki et al. 2001; Portegies Zwart et al. 2004; As shown in previous studies by González et al. (2021)
Gürkan et al. 2006; Giersz et al. 2015; Mapelli 2016). In and González Prieto et al. (2022), increasing the binary
an early study, Ebisuzaki et al. (2001) introduced the fraction for high-mass stars (M > 15M⊙ ) to 100%, more
collisional runaway formation scenario for IMBHs and in line with observed binary fractions in the Galactic
discussed the possibility that these IMBHs will even- field (e.g., Sana et al. 2012a; Moe & Di Stefano 2017),
tually sink to the Galactic center and be the seeds for dramatically increases the number of massive stellar col-
super-massive BHs (SMBHs). The possibility that mas- lisions and thus results in more massive BHs.
sive collision products could avoid the pair-instability In this paper, we re-examine the formation of VMSs
regime and directly collapse into a massive BH (e.g., while self-consistently modeling stellar and binary evo-
Spera et al. 2019; Di Carlo et al. 2019, 2020) has recently lution. We cover systematically the parameter space,
been confirmed via hydrodynamic simulations of stel- extending boundaries in cluster size, density, and mass.
lar collisions and the product’s ensuing evolution (Costa Furthermore, we fully explore the stochasticity of this
et al. 2022; Ballone et al. 2023). The possible presence process and derive statistical distributions for the
of IMBHs at the centers of GCs has also been studied masses of the VMSs. In Section 2, we describe the
for many years (see Greene et al. 2020, for a review). physical prescriptions and parameters varied in this
Tentative evidence of massive BHs at the cores of GCs study. In Section 3, we analyze the formation of VMSs
includes velocity dispersion signatures in nearby GCs in our models, while Section 4 presents a simple equa-
(e.g., Noyola et al. 2010; Jalali et al. 2012; Baumgardt tion to estimate the most massive star formed through
2017), accretion signatures from radio observations (e.g., collisions in a cluster. We discuss the resulting BH pop-
Maccarone 2004; Paduano et al. 2024), hypervelocity ulation in Section 5 and present a statistical study of
stars (e.g., Edelmann et al. 2005; Gualandris & Porte- our models in Section 6. Finally, in Section 7, we discuss
gies Zwart 2007), observations of ultra-luminous X-ray the implications and caveats of this study.
sources (e.g., Colbert & Mushotzky 1999; Farrell et al.
Growth of Massive Stars in Star Clusters 3
(t1 , t2 ) are the stellar ages of the two collision compo- Breivik et al. (2020) for a discussion of these rejuve-
nents. frejuv is a coefficient that determines the level of nation prescriptions as well as the choice for frejuv .
rejuvenation experienced by the collision product. We
adopt frejuv = 1 by default and refer the reader to
Table 1 continued
Growth of Massive Stars in Star Clusters 5
Table 1 (continued)
1 2 3 4 5 6 7 8 9 10 11
Model N rv α3 ffb,high Mclus M⋆,max1 M⋆,max2 M⋆,max3 Mmodel,max Ncoll × fmassive
Table 1 continued
6 González Prieto et al.
Table 1 (continued)
1 2 3 4 5 6 7 8 9 10 11
Model N rv α3 ffb,high Mclus M⋆,max1 M⋆,max2 M⋆,max3 Mmodel,max Ncoll × fmassive
Note— Columns 2–6 list the initial physical parameters of our clusters, including the initial number of objects, virial radius, absolute value of
the high-mass stellar IMF slope (α3 ), high-mass binary fractio, and cluster mass. Columns 7–9 list the mass of the most massive star formed
in each realization of the same model. The dagger indicates masses that result from stellar IMF alone (i.e., those that do not experience any
collisional growth). Column 10 lists the mass of the most massive star as predicted by the fitting formula described in Section 4, with the
error bars indicating the 95% confidence interval. Finally, column 11 records the number of binary–single and binary–binary interactions
involving the merger of at least two stars that are both more massive than 15 M⊙ , normalized by the initial number of massive stars sampled
in the cluster. This gives a rough sense of the number of collisions per massive star in each model.
0
1.
2.
3.
=
=
3
3
α
α
—
—
— fb,high = 1.0
— fb,high = 0.05
16 × 105
Stellar Mass [M ]
8 × 105 103
4 × 105
102
0.5 1 2
rv [pc]
Figure 1. The mean mass of the most massive star formed in a cluster based on its initial conditions. Specifically, the
horizontal axis specifies the cluster’s initial virial radius rv , and the vertical axis specifies its initial number of objects N (both
singles and binaries). Within each box in the rv –N grid, we present a 3x3 sub-grid to distinguish models with different high-mass
stellar IMF slope α3 = (1.6, 2.3, 3.0), from left to right, and high-mass binary fraction fb,high = (0.05, 0.25, 1.0), from bottom to
top. The size and color of the circles reflect the mass of the most massive star formed at each set of initial conditions (averaged
over all three realizations).
formation of a VMS via a sequence of stellar collisions. ucts of binary coalescences, which increase as the pri-
The trend becomes apparent as the slope of the IMF be- mordial fraction of binaries increases. When combined
comes shallower, which is equivalent to a leftward move- with a slightly denser cluster, this mechanism facilitates
ment within each 3x3 box in Figure 1. Most notably, in the formation of a VMS.
models with α3 = (1.6, 2.3) and fb,high = 1, the forma- Finally, models with initial virial radii of 0.5 pc repre-
tion of M⋆,max primarily occurs through a few collisions sent the densest clusters in the grid. Within this subset,
(typically between 2 or 3) that involve massive stars. In we observe a greater diversity of pathways leading to the
contrast with the formation channel described for the formation of massive stars. In models with a top-light
2 pc models, the stars involved in these collisions are IMF (α3 = 3.0), the progenitors of massive stars ex-
slightly more massive than those initially sampled from perience multiple collisions both during their MS stage
the IMF. These unusually-massive stars tend to be prod- and their giant phase. These collisions typically result
8 González Prieto et al.
1.0
0.8
0.6
CDF
0.4
rv = 0.5 α3 = 1.6 fb,high = 1.0
0.2 rv = 1 α3 = 2.3 fb,high = 0.25
rv = 2 α3 = 3.0 fb,high = 0.05
0.0
100 101 102 103 104 100 101 102 103 104 100 101 102 103 104
Total Number of Collisions
Figure 3. The cumulative distribution of the total number of collisions that contributed to the formation of the most massive
star in each of the models. Each panel shows the distribution across all model realizations, differentiated by virial radius (left
panel), high-mass stellar IMF slope (center), and high-mass binary fraction (right).
central density forms new binaries through three-body runaway process halts once the core re-expands. As a re-
binary formation, which in turn heats the core (Cohn sult, we are not in a regime (unlike previous studies, e.g.,
et al. 1989). Furthermore, the collisions that formed the Freitag et al. 2006a,b) where an extreme collisional run-
massive object have extracted some of the (negative) away scenario causes the entire cluster core to collapse
potential energy of the cluster. Thus, once the series of and form a VMS with a mass ∼10−3 times the cluster
collisions stops, the cluster begins to re-expand. In cases mass. Instead, we observe smaller-scale runaways that
where a very massive star does not form, like model a4 enable the cluster to re-expand and continue its evolu-
with a canonical stellar IMF (purple curve), it is a com- tion. This can be primarily attributed to the delay in the
bination of mass loss due to supernovae and three-body core collapse timescale because of updates in stellar evo-
binary formation that drives the core re-expansion. lution prescriptions and the role of binaries in heating
To demonstrate how the initial core collapse con- up the cluster core. Consequently, none of our models
tributes to the formation of a VMS, we present the col- reach core collapse during the initial 10 Myr. The future
lisional history of M⋆,max in run a1 in the lower panel fate of such clusters (beyond the 10 Myr modeled in this
of Figure 4. Each distinct colored curve within the di- paper) falls outside the scope of our current study, but
agram represents a separate branch in the evolutionary will be investigated in a subsequent publication.
process. At about 2.3 Myr, three separate massive stars
form, each weighing approximately 700 M⊙ . This occurs 4. FITTING FORMULAE
precisely when the core undergoes its initial contraction. Carefully mapping the different evolutionary out-
The stars formed in the yellow and magenta pathways comes of clusters across a broad physical spectrum
merge, giving rise to a star of roughly 1000 M⊙ . While requires an extensive, high-resolution grid of simula-
the core begins to re-expand, the merger product under- tions. However, this task is rather computationally
goes gradual mass loss due to stellar winds. At approx- impractical, so we develop a simple fitting formula that
imately 4.5 Myr, this star collides with a star of mass can be used to estimate M⋆,max for a cluster, based
400 M⊙ , resulting in a collision product with a mass of upon the model grid explored in this paper. We begin
approximately 1100 M⊙ . with a simple power-law formula for the dependence of
It is important to note that at any given time in a clus- the maximum stellar mass on N , rv , α3 , and fb,high :
ter, more than one VMS might be present, which is not
depicted in the figures in this paper. Our study focuses M⋆,max
N
η
rv
−β
on assessing the extent of runaway phenomena occur- =A· (α3 )−ω (fb,high )γ . (3)
M⊙ 105 pc
ring within the initial 10 Myr of the cluster’s lifetime.
This takes into account whether or not the massive stars To determine the values of the coefficient A and the
formed in the cluster have enough time to sink to the power-law exponents, we perform a Bayesian inference
cluster center due to dynamical friction and merge. technique known as nested sampling (see Skilling 2006,
In general, although VMSs form during the initial core for a review of the method), computing the parameters
collapse (core contraction) in our cluster simulations, the constraining the model using the nestle package (Shaw
10 González Prieto et al.
α3 = 1.6
103
tion timescale for the massive stars, promoting earlier
and more frequent massive collisions.
Stellar Mass [M ]
imum stellar mass as a function of virial radius of the González et al. 2021; Weatherford et al. 2021; González
CMC data plotted in Figure 5. Currently, our fitting for- Prieto et al. 2022).
mulae does not take into account the assumed IMF, so
the prediction for the maximum mass is allowed to go 6. CMC STATISTICAL ANALYSIS
below the IMF upper limit.
Given the inherently statistical nature of the Monte
It is important to note that this fitting formula is in-
Carlo algorithm, we now investigate whether simulations
trinsic to our assumed stellar evolution prescriptions. As
with identical macroscopic initial conditions but differ-
such, it may not hold for clusters with different stellar
ent random seeds (which set the exact initial stellar posi-
treatments and physical assumptions (e.g., an Elson pro-
tions and velocities) yield a statistical mass distribution
file), or for clusters that deviate significantly from the
for M⋆,max . In Figure 7, we present the distribution of
parameter space covered by our model suite (e.g., very
stellar masses from a set of 50 simulations with an initial
low-mass or very high-mass clusters). Moreover, in the
population of 16 × 105 objects, a virial radius of 1 pc,
denser clusters, the mechanism through which a mas-
α3 = 1.6, and a high-mass binary fraction of 1. For qual-
sive star forms is stochastic, so we expect considerable
itative comparison, we also overplot the first three runs
variance in mass within those regimes.
in orange. These specific initial conditions were chosen
5. MASSIVE BH FORMATION because they represent one of the densest regions in our
current grid, often resulting in the formation of a VMS
An essential question stemming from this research with a mass of a few hundred M⊙ . While these models
is the fate of the massive stars in these clusters, and are to some extent stochastic in nature, an examination
whether the formation of the runaway object has an im- of Figure 7 reveals that the stellar masses roughly fol-
pact on the compact object population. In Figure 6, low a Gaussian distribution centered at 440 M⊙ , with a
we show the spectrum of BH masses formed across the spread of 50 M⊙ (the Gaussian is added for illustrative
first 10 Myr of our cluster models. For all values of N, purposes, using the mean and standard deviation of the
as rv decreases (moving left in each row), there’s a no- data). While the distribution appears to cover a broad
table increase in the number of BHs formed within or be- range of stellar masses, it is much narrower than the
yond the upper–mass-gap (assumed here to be between spread of masses shown across the entire grid explored
40–120M⊙ , but boundaries are uncertain; e.g., Spera & in this study (see Figure 1).
Mapelli 2017; Takahashi et al. 2018; Farmer et al. 2019). In all 50 realizations, the formation of the massive
This agrees with the findings of González Prieto et al. object results from a series of stellar collisions occur-
(2022) who showed that denser clusters exhibit higher ring during binary–single and binary–binary interac-
rates of stellar collisions, facilitating the formation of tions. Figure 8 illustrates the cumulative distribution
BHs in the upper-mass-gap and IMBH regimes. of the total number of collisions that contributed to the
When increasing the initial number of objects while formation of M⋆,max in each of the 50 cluster models. To
maintaining a constant rv (moving downward in each emphasize the collisions that significantly contribute to
column), we observe that the total number of massive mass buildup, we show in orange those collisions where
BHs formed increases for dense models. This is due to the colliding star is more massive than 15 M⊙ . As de-
the increased “mass budget” as the number of initial picted, for most runs, the massive star forms after a few
objects grows, which allows more stars that were not (≈3–5) stellar collisions. In numerous cases, more than
previously massive BH progenitors to merge and pop- one star merges during a binary-mediated interaction.
ulate the massive BH region. This is also apparent in To account for this, we also plot in black the number
the decrease in BHs with masses in the range 4–10 M⊙ . of interactions that resulted in collisions. It is evident
Furthermore, within each panel, a higher binary frac- that the number of interactions follows a narrow distri-
tion often correlates with the formation of more massive bution with a mean of approximately 2.7 and a stan-
BHs, alongside a lower α3 value. This is also consistent dard deviation of 1.68. Thus, across all 50 simulations,
with the general trends we observe in Figure 1. there is a high level of consistency in the number of in-
Crucially, these BH spectra solely represent the initial teractions that M⋆,max undergoes. This marks the first
10 Myr of the cluster’s evolution and do not constitute a time we have been able to characterize the realization-
complete sample of the full BH population. In fact, due to-realization variability of CMC models with high reso-
to the rejuvenation prescriptions outlined in Section 2 lution, even in the densest and most stochastic regimes.
and lower-mass stars, there are BH progenitors left in The consistent agreement in the low number of colli-
most clusters at that time. We defer more detailed anal- sions required for the formation of a VMS emphasizes
yses of the long-term population and retention of com- the importance of studies using precise hydrodynamic
pact objects to future studies. Nonetheless, the pres- simulations to understand and model the properties of
ence of such a diverse population of massive compact collision products (Costa et al. 2022; Ballone et al. 2023).
objects in these clusters hints at the possibility they
could significantly contribute to numerous gravitational-
wave events (Rodriguez et al. 2019; Kremer et al. 2020b;
12 González Prieto et al.
α3 = 1.6 α3 = 2.3 α3 = 3.0
400000 fb,high = 0.05
104 800000
1600000
3200000
103
102
101
fb,high = 0.25
104
Mstar,max
103
102
101
fb,high = 1.0
104
103
102
101
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
rv [pc]
Figure 5. The maximum stellar mass as a function of initial cluster parameters, as given by the fitting formula Equation (3).
The shaded region represents the 95% credibility region. The mean of the CMC data is over-plotted, with the error bars indicating
the maximum and median values obtained in our models. Additional models that were not utilized in the parameter estimation
process but are used to demonstrate the performance of the predictive model are shown in triangles. Supplementary versions of
this Figure with the axes instead showing α3 and fb,high are available as a Figure Set in the online journal.
7. DISCUSSION AND CONCLUSIONS Table 2), the maximum mass reached depends strongly
7.1. Summary on the high-mass slope of the stellar IMF, the initial
cluster density, and the high-mass primordial binary
This paper presents findings from an extensive grid fraction.
of CMC simulations tracking the dynamical evolution of 2. We have derived a fitting formula that can be used
star clusters during the first 10 Myr of their lifetime to estimate the mass of the VMS as a function of ini-
across a spectrum of initial conditions. We particularly tial cluster conditions. While this equation depends on
focus on the formation process of VMSs, which are likely specific assumptions regarding stellar evolution and col-
progenitors of IMBHs. The results from this study can lision prescriptions, it serves as a useful tool to evaluate
be condensed into three principal findings: the potential for a collisional runaway before performing
1. Clusters that start with sufficiently high densities computationally-expensive N -body simulations.
experience a phase of core contraction at early times. 3. At the end of our simulations, some of the VMSs
If this contraction precedes the first supernovae in the have collapsed to form a BH in the upper-mass-gap or an
cluster, it leads to the formation of a VMS through a col- IMBH. These massive BHs will sink to the center of the
lisional runaway instability. In order of importance (see cluster and participate in dynamical encounters that will
Growth of Massive Stars in Star Clusters 13
100 N = 400000; rv = 0.5 N = 400000; rv = 1 N = 400000; rv = 2
10−4
10−6
10−4
10−6
dN/d(log MBH)
10−4
10−6
10−4
10−6
Figure 6. Normalized BH mass distribution for all models listed in Table 1. Models with fb,high = 1 are depicted in black, those
with fb,high = 0.25 in magenta, and models with fb,high = 0.05 in green. The solid lines represent models with α3 = 1.6, dashed
lines denote models with α3 = 2.3, and dotted lines correspond to models with α3 = 3.0. The blue-shaded region indicates the
upper-mass-gap (defined here between 40–120 M⊙ ) while an arrow marks the start of the IMBH regime (M > 120 M⊙ ).
result in interesting signatures such as tidal disruption McMillan 2002; Devecchi & Volonteri 2009; Tagawa
events, and binary mergers. In particular, BBH mergers et al. 2020) or hierarchical stellar-mass BH mergers in
containing a more massive component are potentially dense clusters (e.g., Davies et al. 2011; Kroupa et al.
very important LIGO/LISA sources (Rodriguez et al. 2020; Atallah et al. 2023).
2019; Kremer et al. 2020b; González et al. 2021; Weath- The clusters modeled in this work could be similar to
erford et al. 2021; González Prieto et al. 2022). the massive star clusters that are believed to be proto-
Observations of very high-redshift quasars have GCs. These clusters are thought to be massive and
sparked debates concerning the formation mechanisms dense, with low metallicity. Our models can thus help
for supermassive BHs. Various proposed channels that constrain the initial properties of clusters that might be
can explain the rapid formation and growth of super- the birth place of the seeds for the very high-redshift
massive BHs include the direct collapse of massive Popu- quasars observed by many telescopes, including JWST.
lation III stars (e.g., Stacy et al. 2012; Hirano & Bromm
2017; Kimura et al. 2021) or massive clouds (e.g., Loeb 7.2. Caveats and Future Work
& Rasio 1994; Oh & Haiman 2002; Mayer et al. 2010), The study of the evolution of single massive stars
or the formation of BH seeds via repeated stellar merg- is currently an active area of research. As an added
ers (e.g., Quinlan & Shapiro 1990; Portegies Zwart & layer of complexity, most massive stars are observed in
14 González Prieto et al.
APPENDIX
0
0
00
00
00
0.2
0.4
0.6
0.8
1.0
0.4
0.8
1.2
1.6
2.0
1.2
1.6
2.0
2.4
2.8
0.1
0.2
0.3
0.4
0.5
40
80
12
16
20
Figure 9. Corner plot for the parameters of the nested sampling fit—Equation (3)—to the distribution of maximum stellar
mass from a collisional runaway.
Growth of Massive Stars in Star Clusters 17
0
1.
2.
3.
=
=
3
3
α
α
—
—
— fb,high = 1.0
— fb,high = 0.05
16 × 105
8 × 105
4 × 105
0.5 1 2
rv [pc]
Figure 10. As in Figure 1, the size of the black circles shows the average mass of the most massive VMS formed in a cluster
with the given initial conditions. In this version, we also overlay in purple the predicted mean mass derived from 1000 samples
obtained from sampling values predicted by Equation (3).
REFERENCES
Agrawal, P., Hurley, J., Stevenson, S., Szécsi, D., & Flynn, Chatterjee, S., Rodriguez, C. L., & Rasio, F. A. 2017, ApJ,
C. 2020, MNRAS, 497, 4549, 834, 68, doi: 10.3847/1538-4357/834/1/68
doi: 10.1093/mnras/staa2264 Chernoff, D. F., & Weinberg, M. D. 1990, ApJ, 351, 121,
Atallah, D., Trani, A. A., Kremer, K., et al. 2023, MNRAS, doi: 10.1086/168451
523, 4227, doi: 10.1093/mnras/stad1634 Cohn, H., Hut, P., & Wise, M. 1989, ApJ, 342, 814,
Ballone, A., Costa, G., Mapelli, M., et al. 2023, MNRAS,
doi: 10.1086/167638
519, 5191, doi: 10.1093/mnras/stac3752
Colbert, E. J. M., & Mushotzky, R. F. 1999, ApJ, 519, 89,
Bartko, H., Martins, F., Trippe, S., et al. 2010, ApJ, 708,
doi: 10.1086/307356
834, doi: 10.1088/0004-637X/708/1/834
Costa, G., Ballone, A., Mapelli, M., & Bressan, A. 2022,
Baumgardt, H. 2017, MNRAS, 464, 2174,
MNRAS, 516, 1072, doi: 10.1093/mnras/stac2222
doi: 10.1093/mnras/stw2488
Breivik, K., Coughlin, S., Zevin, M., et al. 2020, ApJ, 898, Davies, M. B., Miller, M. C., & Bellovary, J. M. 2011,
71, doi: 10.3847/1538-4357/ab9d85 ApJL, 740, L42, doi: 10.1088/2041-8205/740/2/L42
Chatterjee, S., Fregeau, J. M., Umbreit, S., & Rasio, F. A. De Marchi, G., Paresce, F., & Pulone, L. 2007, ApJL, 656,
2010, ApJ, 719, 915, doi: 10.1088/0004-637X/719/1/915 L65, doi: 10.1086/512856
18 González Prieto et al.
Devecchi, B., & Volonteri, M. 2009, ApJ, 694, 302, Heggie, D., & Hut, P. 2003, The Gravitational
doi: 10.1088/0004-637X/694/1/302 Million-Body Problem: A Multidisciplinary Approach to
Di Carlo, U. N., Giacobbo, N., Mapelli, M., et al. 2019, Star Cluster Dynamics
MNRAS, 487, 2947, doi: 10.1093/mnras/stz1453 Hirano, S., & Bromm, V. 2017, MNRAS, 470, 898,
Di Carlo, U. N., Mapelli, M., Bouffanais, Y., et al. 2020, doi: 10.1093/mnras/stx1220
MNRAS, 497, 1043, doi: 10.1093/mnras/staa1997 Jalali, B., Baumgardt, H., Kissler-Patig, M., et al. 2012,
Ebisuzaki, T., Makino, J., Tsuru, T. G., et al. 2001, ApJ, A&A, 538, A19, doi: 10.1051/0004-6361/201116923
562, L19, doi: 10.1086/338118 Joshi, K. J., Rasio, F. A., & Portegies Zwart, S. 2000, ApJ,
Edelmann, H., Napiwotzki, R., Heber, U., Christlieb, N., & 540, 969, doi: 10.1086/309350
Reimers, D. 2005, ApJL, 634, L181, doi: 10.1086/498940 Kimura, K., Hosokawa, T., & Sugimura, K. 2021, ApJ, 911,
Farmer, R., Renzo, M., de Mink, S. E., Marchant, P., & 52, doi: 10.3847/1538-4357/abe866
Justham, S. 2019, ApJ, 887, 53, Kızıltan, B., Baumgardt, H., & Loeb, A. 2017, Nature, 542,
doi: 10.3847/1538-4357/ab518b 203, doi: 10.1038/nature21361
Farrell, S. A., Webb, N. A., Barret, D., Godet, O., & Kremer, K., Ye, C. S., Rui, N. Z., et al. 2020a, ApJS, 247,
Rodrigues, J. M. 2009, Nature, 460, 73, 48, doi: 10.3847/1538-4365/ab7919
doi: 10.1038/nature08083 Kremer, K., Spera, M., Becker, D., et al. 2020b, ApJ, 903,
Feroz, F., Hobson, M. P., & Bridges, M. 2009, MNRAS, 45, doi: 10.3847/1538-4357/abb945
398, 1601, doi: 10.1111/j.1365-2966.2009.14548.x Kroupa, P. 2001, MNRAS, 322, 231,
doi: 10.1046/j.1365-8711.2001.04022.x
Ferrarese, L., & Merritt, D. 2000, ApJL, 539, L9,
Kroupa, P., Subr, L., Jerabkova, T., & Wang, L. 2020,
doi: 10.1086/312838
MNRAS, 498, 5652, doi: 10.1093/mnras/staa2276
Fregeau, J. M., Cheung, P., Portegies Zwart, S. F., &
Lee, H. M. 1987, ApJ, 319, 801, doi: 10.1086/165498
Rasio, F. A. 2004, MNRAS, 352, 1,
Lightman, A. P., & Shapiro, S. L. 1978, Rev. Mod. Phys.,
doi: 10.1111/j.1365-2966.2004.07914.x
50, 437, doi: 10.1103/RevModPhys.50.437
Fregeau, J. M., & Rasio, F. A. 2007, ApJ, 658, 1047,
Loeb, A., & Rasio, F. A. 1994, ApJ, 432, 52,
doi: 10.1086/511809
doi: 10.1086/174548
Freitag, M., Gürkan, M. A., & Rasio, F. A. 2006a, MNRAS,
Lombardi, James C., J., Rasio, F. A., & Shapiro, S. L.
368, 141, doi: 10.1111/j.1365-2966.2006.10096.x
1996, ApJ, 468, 797, doi: 10.1086/177736
Freitag, M., Rasio, F. A., & Baumgardt, H. 2006b,
Ma, X., Grudić, M. Y., Quataert, E., et al. 2020, MNRAS,
MNRAS, 368, 121, doi: 10.1111/j.1365-2966.2006.10095.x
493, 4315, doi: 10.1093/mnras/staa527
Giersz, M., Askar, A., Wang, L., et al. 2019, MNRAS, 487,
Maccarone, T. J. 2004, MNRAS, 351, 1049,
2412, doi: 10.1093/mnras/stz1460
doi: 10.1111/j.1365-2966.2004.07859.x
Giersz, M., Leigh, N., Hypki, A., Lützgendorf, N., & Askar,
Mapelli, M. 2016, MNRAS, 459, 3432,
A. 2015, MNRAS, 454, 3150, doi: 10.1093/mnras/stv2162
doi: 10.1093/mnras/stw869
González, E., Kremer, K., Chatterjee, S., et al. 2021, ApJL, Marchant, P., Renzo, M., Farmer, R., et al. 2019, ApJ, 882,
908, L29, doi: 10.3847/2041-8213/abdf5b 36, doi: 10.3847/1538-4357/ab3426
González Prieto, E., Kremer, K., Fragione, G., et al. 2022, Mayer, L., Kazantzidis, S., Escala, A., & Callegari, S. 2010,
ApJ, 940, 131, doi: 10.3847/1538-4357/ac9b0f Nature, 466, 1082, doi: 10.1038/nature09294
Greene, J. E., Strader, J., & Ho, L. C. 2020, ARA&A, 58, Moe, M., & Di Stefano, R. 2017, ApJS, 230, 15,
257, doi: 10.1146/annurev-astro-032620-021835 doi: 10.3847/1538-4365/aa6fb6
Grudić, M. Y., Hafen, Z., Rodriguez, C. L., et al. 2023, Mowla, L., Iyer, K. G., Desprez, G., et al. 2022, ApJL, 937,
MNRAS, 519, 1366, doi: 10.1093/mnras/stac3573 L35, doi: 10.3847/2041-8213/ac90ca
Gualandris, A., & Portegies Zwart, S. 2007, MNRAS, 376, Mukherjee, P., Parkinson, D., & Liddle, A. R. 2006, ApJL,
L29, doi: 10.1111/j.1745-3933.2007.00280.x 638, L51, doi: 10.1086/501068
Gürkan, M. A., Fregeau, J. M., & Rasio, F. A. 2006, ApJL, Noyola, E., Gebhardt, K., Kissler-Patig, M., et al. 2010,
640, L39, doi: 10.1086/503295 ApJL, 719, L60, doi: 10.1088/2041-8205/719/1/L60
Gürkan, M. A., Freitag, M., & Rasio, F. A. 2004, ApJ, 604, Oh, S. P., & Haiman, Z. 2002, ApJ, 569, 558,
632, doi: 10.1086/381968 doi: 10.1086/339393
Haghi, H., Khalaj, P., Hasani Zonoozi, A., & Kroupa, P. Paduano, A., Bahramian, A., Miller-Jones, J. C. A., et al.
2017, ApJ, 839, 60, doi: 10.3847/1538-4357/aa6719 2024, ApJ, 961, 54, doi: 10.3847/1538-4357/ad0e68
Growth of Massive Stars in Star Clusters 19
Portegies Zwart, S. F., Baumgardt, H., Hut, P., Makino, J., Skilling, J. 2006, Bayesian Analysis, 1, 833 ,
& McMillan, S. L. W. 2004, Nature, 428, 724, doi: 10.1214/06-BA127
doi: 10.1038/nature02448
Spera, M., & Mapelli, M. 2017, MNRAS, 470, 4739,
Portegies Zwart, S. F., & McMillan, S. L. W. 2002, ApJ,
doi: 10.1093/mnras/stx1576
576, 899, doi: 10.1086/341798
Quinlan, G. D., & Shapiro, S. L. 1990, ApJ, 356, 483, Spera, M., Mapelli, M., Giacobbo, N., et al. 2019, MNRAS,
doi: 10.1086/168856 485, 889, doi: 10.1093/mnras/stz359
Rodriguez, C. L., Hafen, Z., Grudić, M. Y., et al. 2023, Stacy, A., Greif, T. H., & Bromm, V. 2012, MNRAS, 422,
MNRAS, 521, 124, doi: 10.1093/mnras/stad578
290, doi: 10.1111/j.1365-2966.2012.20605.x
Rodriguez, C. L., Zevin, M., Amaro-Seoane, P., et al. 2019,
PhRvD, 100, 043027, doi: 10.1103/PhysRevD.100.043027 Tagawa, H., Haiman, Z., & Kocsis, B. 2020, ApJ, 892, 36,
Rodriguez, C. L., Weatherford, N. C., Coughlin, S. C., doi: 10.3847/1538-4357/ab7922
et al. 2022, ApJS, 258, 22, doi: 10.3847/1538-4365/ac2edf Takahashi, K., Yoshida, T., & Umeda, H. 2018, ApJ, 857,
Rui, N. Z., Kremer, K., Weatherford, N. C., et al. 2021, 111, doi: 10.3847/1538-4357/aab95f
ApJ, 912, 102, doi: 10.3847/1538-4357/abed49
Vanzella, E., Claeyssens, A., Welch, B., et al. 2023, ApJ,
Sana, H., de Mink, S. E., de Koter, A., et al. 2012a,
Science, 337, 444, doi: 10.1126/science.1223344 945, 53, doi: 10.3847/1538-4357/acb59a
—. 2012b, Science, 337, 444, doi: 10.1126/science.1223344 Vishniac, E. T. 1978, ApJ, 223, 986, doi: 10.1086/156332
Sanders, R. H. 1970, ApJ, 162, 791, doi: 10.1086/150711
Weatherford, N. C., Fragione, G., Kremer, K., et al. 2021,
Shaw, J. R., Bridges, M., & Hobson, M. P. 2007, MNRAS,
ApJL, 907, L25, doi: 10.3847/2041-8213/abd79c
378, 1365, doi: 10.1111/j.1365-2966.2007.11871.x
Sills, A., Faber, J. A., Lombardi, James C., J., Rasio, F. A., Wirth, H., Kroupa, P., Haas, J., et al. 2022, MNRAS, 516,
& Warren, A. R. 2001, ApJ, 548, 323, 3342, doi: 10.1093/mnras/stac2424
doi: 10.1086/318689