Runaway Stellar Collisions and Dense Clusters

A young, compact star cluster can concentrate its most massive members faster than they evolve, drive its core to very high density, and feed repeated stellar mergers into one growing object. The mechanism is genuinely different from monolithic gas collapse: its clock is collisional stellar dynamics, and its product must survive stellar winds and nuclear evolution before it can become a black hole. The decisive question is therefore a race among relaxation, core collapse, collisions, binary heating, massive-star evolution, and mass loss.

Helpful background. Black-Hole Geometry and Mechanics supplies the compact remnant endpoint; Relativistic Matter and Compact Objects supplies stellar evolution and collision physics; Black-Hole Astrophysics supplies cluster and accretion context; and Relativistic Cosmology supplies the early-galaxy environment.

The cluster clock

For a cluster with $N$ stars, total mass $M_{\rm cl}$, and three-dimensional half-mass radius $r_h$, a useful half-mass relaxation estimate is

$$t_{rh}\simeq \frac{0.138N}{\ln\Lambda} \left(\frac{r_h^3}{GM_{\rm cl}}\right)^{1/2}, \qquad \ln\Lambda\sim\ln(0.4N).$$

Two-body encounters exchange energy while leaving the cluster approximately collisionless on each crossing. Massive stars lose orbital energy to the lighter background and segregate on roughly $t_{\rm seg}\sim(\langle m\rangle/m_{\rm massive})t_{rh}$. For a broad mass spectrum, the segregated subsystem can drive core collapse in $t_{cc}\sim0.1$–$0.2t_{rh}$, much faster than the equal-mass value. A stellar collision runaway requires $t_{cc}$ to be shorter than the roughly $3$–$5$ Myr life of the most massive stars.

Direct $N$-body experiments by Portegies Zwart and McMillan found the first collision near core collapse, recurrent participation by one target, and suppression of the runaway when the initial relaxation time exceeded roughly $25$ Myr. Their fitted outcome was a collision product of order $10^{-3}M_{\rm cl}$, not the entire cluster (Portegies Zwart & McMillan 2002, §§2–4, Eqs. 1–6 and Fig. 3). The numerical coefficient is model-dependent, but the timescale ordering is the channel’s durable organizing principle.

Why one collision becomes a runaway

For two stars with relative speed at infinity $v_\infty$, the collision cross-section is

$$\sigma_{12}=\pi(R_1+R_2)^2 \left[1+\frac{2G(m_1+m_2)}{v_\infty^2(R_1+R_2)}\right].$$

The bracket is gravitational focusing. In a young cluster whose velocity dispersion is well below stellar escape speeds, focusing dominates: becoming more massive and larger increases a merger product’s encounter cross-section, while mass segregation keeps it in the densest region. Dynamically formed hard binaries can catalyze collisions through resonant three-body encounters. Primordial binaries, however, can spread interactions across many pairs or heat and expand the core. Thus “binaries help” and “binaries stop collapse” are both possible statements in different binary-fraction and hardness regimes.

The cross-section follows directly from a grazing orbit. Put $R_{12}=R_1+R_2$. Conservation of relative orbital energy gives $v_p^2=v_\infty^2+2G(m_1+m_2)/R_{12}$ at grazing pericenter, and angular momentum gives $b_{\max}v_\infty=R_{12}v_p$. The area $\pi b_{\max}^2$ is the expression above. It assumes an isolated two-body approach up to contact, without gas drag or tidal dissipation before the encounter; a direct binary-mediated collision needs a few-body calculation.

At local stellar number density $n_2$, the collision time for a target is $t_{\rm coll}\simeq[n_2\langle\sigma_{12}v_\infty\rangle]^{-1}$. Thus a short relaxation time is only the first gate: after the core contracts, collisions must recur quickly enough to beat the target’s remaining lifetime. A large cross-section in a sparse cluster does not produce a runaway.

The merger product is not automatically a seed. Sticky-sphere prescriptions conserve mass too generously when hydrodynamic ejecta are important. Rejuvenation changes its remaining lifetime; rotation alters mixing; and metal-dependent line-driven winds can remove much of the assembled envelope. Low metallicity favors retention, but uncertainty in very-massive-star evolution can shift the final black-hole mass by orders of magnitude. Pair instability may disrupt some helium-core masses altogether.

Gas-rich and remnant variants

Early clusters may remain embedded in gas. Accretion bloats protostars and enlarges geometric cross-sections, while gas drag accelerates segregation; on the other hand, gas expulsion changes the potential and can expand the cluster. Fragmentation at extremely high density could create clusters far denser than nearby examples. Current synthesis allows stellar-collision, gas-assisted hybrid, and repeated remnant-black-hole routes, with plausible intermediate-mass products around $10^3$–$10^5M_\odot$ under selected conditions (Inayoshi, Visbal & Haiman 2020, §5.4, pp. 70–71).

The remnant route has a different clock. Massive stars first become black holes; those remnants segregate and form binaries whose hardening can lead to mergers. Recoil may eject merger products from low-escape-speed clusters, and repeated mergers require a retained population. A relativistic cluster collapse is yet another matter model, relevant only at far larger compactness. These variants should not be combined into a single efficiency parameter.

Results and limitations through 2026-08-18

The evidence cutoff is 2026-08-18. Later results are outside this assessment, and no observed cluster presently fixes all of the required birth conditions and remnant physics.

Evidence or process	Supported result	Scope, status, or limitation
Relaxation and segregation theory	Compact clusters can concentrate massive stars before global relaxation	Spherical equilibrium estimates omit substructure, gas, rotation, and time-dependent tides
Direct cluster calculations	Core collapse can trigger repeated collisions into one runaway object	Results depend on $N$, stellar radii, few-body integration, binary population, and collision treatment
Very-massive-star evolution	Winds, mixing, pair instability, and lifetime set the retained remnant	Extrapolations are especially uncertain at low metallicity and above observed stellar masses
Gas-embedded cluster models	Accretion and drag can accelerate or enlarge the collision sequence	Feedback and gas removal can instead expand the system; multiscale convergence is incomplete
Intermediate-mass black-hole candidates	A retained central object would be a plausible descendant	Dynamical degeneracies and accretion alternatives do not reconstruct a unique collisional origin

Credible competing positions are a pre-supernova stellar runaway, binary heating that delays collapse, wind-limited growth, gas-assisted mergers, repeated black-hole mergers, and relativistic collapse in an exceptionally compact system. The evidence does not establish which, if any, supplies a major seed population; the conclusion depends on the initial data class and stellar-evolution prescription.

The interpretation could change after a resolved census constrains young-cluster birth radii and binary fractions, merger-product calculations close the rotation and wind budget, or a direct collisional simulation converges while increasing particle number and varying stellar radii. A secure intermediate-mass black-hole dynamical detection with a reconstructable host history would also change the empirical status.

Formulation-choice diagnostic

A cluster has $M_{\rm cl}=10^5M_\odot$, $N=2\times10^5$, $r_h=0.50$ pc, and a broad mass spectrum for which $t_{cc}=0.15t_{rh}$. Use $G=4.50\times10^{-3}\,\mathrm{pc^3\,M_\odot^{-1}\,Myr^{-2}}$. Should the first viability test be a stellar collision cross-section or the cluster relaxation clock, and does collapse precede a 3 Myr massive-star deadline? How would shrinking $r_h$ to $0.30$ pc change the answer?

Solution

The relaxation clock comes first because the cross-section matters only after segregation has raised the core density. Here $\ln(0.4N)=\ln(80{,}000)=11.29$ and the dynamical factor is $[r_h^3/(GM_{\rm cl})]^{1/2}=0.0167$ Myr. Thus $t_{rh}\simeq40.8$ Myr and $t_{cc}\simeq6.1$ Myr, later than the deadline: the declared cluster fails. At fixed $M_{\rm cl}$ and $N$, times scale as $r_h^{3/2}$. Reducing the radius multiplies them by $(0.30/0.50)^{3/2}=0.465$, giving $t_{cc}\simeq2.85$ Myr. That second cluster is only marginally viable; winds, binaries, and collision losses still decide whether a seed survives.

References

• Inayoshi, Kohei, Eli Visbal, and Zoltán Haiman. “The Assembly of the First Massive Black Holes.” Annual Review of Astronomy and Astrophysics 58 (2020): 27–97. DOI; Open PDF.
• Portegies Zwart, Simon F., and Stephen L. W. McMillan. “The Runaway Growth of Intermediate-Mass Black Holes in Dense Star Clusters.” The Astrophysical Journal 576 (2002): 899–907. DOI; Open PDF.