Final-Parsec Evolution and Coalescence

A massive black-hole pair formed by a galaxy merger must shrink from parsec scales to the regime where gravitational radiation completes the coalescence. Dynamical friction forms the bound pair; three-body scattering, nonspherical stellar orbits, and sometimes gas must then keep extracting energy and angular momentum. The “final-parsec problem” is not a universal extra force. It is a question about whether the host continuously supplies matter to the binary’s shrinking interaction region quickly enough to reach the regime dominated by gravitational radiation within cosmic time.

Helpful background. Black-Hole Geometry supplies relativistic inner scales; Relativistic Matter and Compact Objects supplies matter models; Black-Hole Astrophysics supplies nuclear and accretion context; Two-Body Dynamics supplies orbital energy and radiation reaction; Gravitational-Wave Astronomy supplies population and detector inference.

From a hard binary to a depleted loss cone

Let $M=M_1+M_2$, $\mu=M_1M_2/M$, and $q=M_2/M_1\le1$. A conventional hard-binary scale is

$$a_h=\frac{G\mu}{4\sigma^2},$$

where $\sigma$ characterizes the surrounding stellar velocity dispersion. Below $a_h$, an encounter with the binary can eject a star by the gravitational slingshot. The binary’s binding-energy magnitude,

$$B=\frac{G\mu M}{2a},$$

then grows at the star’s expense. In the ideal full-loss-cone limit, scattering experiments motivate

$$s_\star \equiv \frac{d}{dt}\left(\frac{1}{a}\right) \simeq H\frac{G\rho}{\sigma},$$

with dimensionless $H$ of order ten but dependent on $q$, $e$, and the stellar distribution. The corresponding shrinkage is $\dot a_\star=-s_\star a^2$ and $\dot B_\star=G\mu M s_\star/2$. These equations make the energy check explicit: the energy gained by ejected stars must match the increase in binary binding energy.

The supply is finite. Once stars on low-angular-momentum orbits have been expelled, a spherical collisionless nucleus cannot refill them. Two-body relaxation can diffuse new stars into the loss cone, but its time can exceed a galaxy age in a massive, low-density core. The hardening rate then falls below the full-loss-cone value and the binary may stall. Merritt and Milosavljević 2005, §§ 4.1–4.4 develops the fixed-background result, depletion, collisional refilling, and the distinct role of nonaxisymmetric nuclei.

Geometry changes the supply problem

Spherical symmetry conserves each stellar orbit’s angular-momentum magnitude in the smooth potential. Axisymmetry preserves one component and enlarges the accessible region, but that region can still drain. A triaxial potential admits centrophilic families whose angular momenta vary collisionlessly; even moderate triaxiality can keep feeding the binary.

This distinction is numerically delicate. A direct $N$-body model has far fewer, much heavier particles than a real galaxy, so artificial two-body relaxation and excessive Brownian motion can refill the loss cone. A measured hardening rate that depends on $N$, softening, or the black holes’ wandering cannot be extrapolated as purely collisionless. The useful convergence question is not merely whether $a(t)$ looks smooth. One must separate collisional diffusion from collisionless torques and demonstrate a large-$N$ limiting rate.

Monte Carlo and orbit-based calculations designed to suppress artificial relaxation find stalling in spherical systems and continued, though sub-full-loss-cone, shrinkage in triaxial systems; their fiducial triaxial models reach coalescence on timescales no longer than about a gigayear Vasiliev, Antonini, and Merritt 2015, §§ 2.1, 4.1–4.4, and 6.1. This is evidence for a conditional stellar solution, not proof that every merger remnant stays triaxial, dense, and undisturbed for the required interval.

Gas and the transition to gravitational radiation

Gas can add dynamical friction, a massive circumnuclear disk, streams, and circumbinary torques. Its sign and duration depend on cooling, fragmentation, star formation, magnetic stress, feedback, cavity structure, and external supply. A gas-rich merger is therefore not automatically an efficient inspiral. The separate circumbinary-disk treatment explains why the net angular-momentum flux must be measured rather than assigned.

A compact bridge model is

$$\frac{da}{dt} = -s_\star(a,t)a^2 +\dot a_{\rm gas}(a,t) -\frac{K_{\rm GW}}{a^3}\mathcal F(e),$$

where

$$K_{\rm GW} = \frac{64}{5}\frac{G^3\mu M^2}{c^5}, \qquad \mathcal F(e) = \frac{1+\frac{73}{24}e^2+\frac{37}{96}e^4} {(1-e^2)^{7/2}}.$$

Here $c$ has been restored to display the gravitational-wave scale. This is not closed until an eccentricity equation accompanies it. Stellar scattering and gas can raise or lower $e$, whereas radiation reaction circularizes; the angular-momentum budget must reproduce that evolution. Peters’ orbit-averaged energy and angular-momentum fluxes provide the gravitational-radiation term Peters 1964, § V, equations (5.6)–(5.7).

For a circular, gas-free model with constant $s_\star$, equality of the instantaneous stellar and radiation rates occurs at

$$a_{\rm cross} = \left(\frac{K_{\rm GW}}{s_\star}\right)^{1/5}.$$

The fifth root makes the transition separation relatively insensitive to a single normalization, but the time spent reaching it remains sensitive to loss-cone depletion. Eccentricity increases $\mathcal F(e)$ sharply and moves the transition outward.

Results and unresolved regimes

The evidence cutoff is 2026-08-18. The open question is population-wide: which host histories keep binaries moving, and with what delay distribution? Every claim must specify the parameter regime and boundary conditions—host shape and nuclear density, gas fraction and supply, loss-cone refill mechanism, $q$, $e$, and the stellar, gaseous, and radiation-reaction clocks.

Host and channel	Evidence or result	Required independent check	Scope, limitation, or unresolved boundary
Spherical, gas-poor, collisionless nucleus	The initially populated loss cone drains and stellar hardening stalls	Separate physical diffusion from particle-number relaxation and close the stellar energy budget	Real nuclei may retain weak asymmetry, perturbers, or time dependence
Triaxial stellar merger remnant	Centrophilic orbits can sustain sub-full-loss-cone hardening to the radiation regime	Converge with $N$, shape evolution, orbit family, and binary wandering measured independently	The duration and prevalence of central triaxiality are not fixed
Gas-rich nucleus	Gas may shorten, lengthen, or intermittently pause the residence time	Close mass, energy, and angular momentum across nested radii while varying feedback and boundaries	Thin radiative MHD coverage across $q$ and $e$ remains sparse
Cosmological binary population	A mixture of prompt and delayed channels is plausible	Forward-model host selection and the complete delay distribution into survey and detector likelihoods	No current population uniquely identifies a universal hardening law

The credible competing positions are therefore: efficient coalescence is generic in sufficiently nonspherical or gas-supplied remnants; long stalling survives in spherical depleted nuclei; or delays are channel-dependent and broad enough that neither slogan describes the population. A calculation discriminates among them only if it integrates the stellar, gas, and gravitational-wave terms with consistent $a$ and $e$, verifies energy and angular-momentum closure, and rejects tracks longer than the available cosmic time.

The interpretation could change with a selection-corrected pulsar-timing or space-based gravitational-wave population constraint that resolves the environmental turnover or delay distribution; a new nuclear-shape and density survey tied to merger age; or a converged multiphysics merger calculation that spans the relevant mass ratios and host boundary conditions.

Interpretation and scope

“Solved” can mean three different things: one simulated binary crossed the gap; one class of triaxial models does so in under a gigayear; or the cosmic population coalesces at the rate inferred from data. These claims require different evidence. The first is an existence result, the second a conditional mechanism, and the third a hierarchical population inference.

The environmental mechanisms considered here must bring a hard massive binary to radiation-dominated inspiral; the calculation identifies when they fail. General two-body dynamics and detector likelihoods require separate treatments.

Common pitfalls

Using the full-loss-cone rate after depletion. $HG\rho/\sigma$ is a benchmark, not a guaranteed late-time rate.

Treating finite-$N$ relaxation as galaxy physics. Apparent convergence of $a(t)$ is insufficient unless the collisional and collisionless supplies are separated.

Adding timescales instead of rates. Stellar, gas, and radiation terms act simultaneously and may have opposite signs; evolve the coupled equations.

Test a bridge to coalescence

Take an equal-mass circular binary with $M_1=M_2=10^8M_\odot$, a constant stellar hardening rate $s_\star=50\,{\rm pc^{-1}\,Gyr^{-1}}$, no gas, and initial separation $a_0=1\,{\rm pc}$. Using $K_{\rm GW}=8.60\times10^{-7}\,{\rm pc^4\,Gyr^{-1}}$, find $a_{\rm cross}$. First estimate the time by retaining only stars above the crossing and only gravitational radiation below it. Then write the simultaneous-rate integral. Does this idealized track meet a $1\,{\rm Gyr}$ bound?

Solution

Rate equality gives

$$a_{\rm cross} = \left(\frac{8.60\times10^{-7}}{50}\right)^{1/5}{\rm pc} =0.0280\,{\rm pc}.$$

Because $d(1/a)/dt=s_\star$,

$$t_\star = \frac{a_{\rm cross}^{-1}-a_0^{-1}}{s_\star} = \frac{35.7-1.0}{50}\ {\rm Gyr} =0.694\ {\rm Gyr}.$$

For $\dot a=-K_{\rm GW}/a^3$, integration from $a_{\rm cross}$ to zero gives

$$t_{\rm GW} = \frac{a_{\rm cross}^4}{4K_{\rm GW}} =0.179\ {\rm Gyr}.$$

The sum is about $0.873\,{\rm Gyr}$. This is an upper bound for the stated constant-rate model: at every separation it discards one additional inward contribution. When both mechanisms act simultaneously, the time is instead

$$t_{\rm both} =\int_0^{a_0}\frac{a^3\,da}{s_\star a^5+K_{\rm GW}} =\frac{1}{s_\star a_{\rm cross}} \int_0^{a_0/a_{\rm cross}}\frac{x^3}{1+x^5}\,dx \simeq0.744\,{\rm Gyr}.$$

The integrand is regular at zero; the final strong-field merger lies outside the orbit-averaged formula but contributes negligibly on this galactic timescale. Both estimates meet the bound. Neither is an astrophysical population conclusion: depletion that lowers $s_\star$, gas expansion, or an evolving host can lengthen the history, while nonzero eccentricity can shorten it.

References

• Merritt, David, and Miloš Milosavljević. “Massive Black Hole Binary Evolution.” Living Reviews in Relativity 8 (2005): 8. doi:10.12942/lrr-2005-8.
• Peters, P. C. “Gravitational Radiation and the Motion of Two Point Masses.” Physical Review 136 (1964): B1224–B1232. doi:10.1103/PhysRev.136.B1224.
• Vasiliev, Eugene, Fabio Antonini, and David Merritt. “The Final-Parsec Problem in the Collisionless Limit.” The Astrophysical Journal 810 (2015): 49. doi:10.1088/0004-637X/810/1/49.