Galaxy Merger · Restricted N-Body

Two rigid galaxy potentials on a Keplerian relative orbit, each populated with 104 non-interacting test particles. Tidal bridges and tails grow at constant cost — the 1972 IBM/360 calculation that produced the morphology of the Antennae and the Mice, in a browser tab at 60 fps.

Toomre & Toomre · 1972 · ApJ 178, 623 · 1972ApJ...178..623T
MODEL CLASS  RESTRICTED N-BODY  rigid potentials and massless test particles — no self-gravity
Antennae
configuration
19
pericenter q · kpc
1.00
eccentricity e
60
i1 · °
60
i2 · °
10000
test particles
Restricted N-Body · Rigid Hernquist Potentials · Chandrasekhar Friction · 3 Gyr

Two galaxies, two disks of 5,000 test particles, encounter through coalescence

Each galaxy is a rigid Hernquist halo plus a Miyamoto-Nagai disk; test particles see both potentials but not each other. The relative orbit is two-body Kepler plus a Chandrasekhar dynamical-friction prescription, so bound passages decay; below 1 kpc the centers lock to their common COM and the particles phase-mix into the remnant. The scrubber covers 3 Gyr in 5 Myr frames.

Computing trajectories…
leapfrog · 2×5000 particles · 0.5 Myr step · 3 Gyr

View

State · live

t0 Myr
separation100 kpc
d/dt−210 km/s
peri count0
coalesced
tail A length0 kpc
tail B length0 kpc
Disk A in orange, disk B in cyan. Particles inside their original disk are dim; particles unbound from their parent galaxy redden as escape velocity is exceeded.

Separation r(t)

0 → 3 Gyr · q markers
0t · Gyr3
t = 0 Myr · r = 100 kpc
Home · Entry III · Galaxy Merger

1 · The restricted N-body framework

Toomre & Toomre's 1972 demonstration that bridges and tails are tidal relics of close encounters was a restricted calculation: each galaxy was treated as a single point mass, and each disk was a swarm of non-self-gravitating test particles on initially circular orbits. The decoupling is the entire trick — the disk does not pull on itself, the disk does not pull on the companion, and the two galaxy centers fall along an exact two-body trajectory. Test-particle equations of motion run in parallel; cost is linear in particle count rather than the N² of a true N-body code. The morphology that emerges — the long ratty tail, the broad facing-side bridge, the prograde-retrograde asymmetry — is set by the relative orbit and the disk inclinations alone, and it survives essentially unchanged when self-gravity is reintroduced in full simulations.

This entry generalizes the 1972 setup in three ways: (i) the point-mass galaxies are replaced by analytic Hernquist 1990 halos plus thin Miyamoto-Nagai 1975 disks, so the disk rotation curve is realistic at small radii; (ii) integration is done with a kick-drift-kick leapfrog at 0.5 Myr step, symplectic in the canonical variables of each test particle, with the galaxy centers integrated in lockstep so the orbit is consistent at every drift; (iii) the relative orbit is selectable across parabolic, hyperbolic, and eccentric-bound regimes; (iv) a Chandrasekhar 1943 dynamical-friction prescription drags the center orbit through each rigid halo's density field (lnΛ = 2 fixed, Maxwellian F(X), σ ≈ Vc/√2), so bound passages decay and the centers coalesce; below 1 kpc separation the two potentials are locked to their common center of mass and the test particles phase-mix in the combined potential — the closest a restricted calculation can come to a remnant. What that remnant can and cannot be trusted for is in the validity panel below.

Equations of motion

 Φi(r)  =  −G Mhalo,i / (|r| + ah,i)  −  G Mdisk,i / √( R² + (ad,i + √(z² + bd,i²))² )     (Hernquist + Miyamoto-Nagai)
 ẍp  =  −∇Φ1(xp − x1(t))  −  ∇Φ2(xp − x2(t))     (test particle, no back-reaction)
 ẍ1 = −G M2 (x1−x2) / (|x1−x2|² + ε²)3/2  +  adf    (and symmetric for x2)     (galaxy centers, softened ε = 0.5 kpc)
 |adf|  =  4π G² Msat ρhost(r) lnΛ F(X) / vrel² ,    F(X) = erf(X) − (2X/√π) e−X² ,   X = vrel/(√2 σ)     (Chandrasekhar 1943; lnΛ = 2, reduced-mass weighted, COM-preserving)

2 · Configurations available

Five presets, four of which reproduce iconic interacting pairs at the qualitative level the 1972 paper claimed. The fifth (retrograde-retrograde) is the control: same orbit, same masses, opposite disk spins, no tails. Click a preset to load it, then change disk inclinations or pericenter and click Recompute.

Presetq (kpc)eMA (1011)MB (1011)iA, ωAiB, ωBComment
Antennae (NGC 4038/9)191.001.01.060°, 30°60°, 60°parabolic; Toomre 1972 best-fit geometry
Mice (NGC 4676)161.000.70.715°, 0°−10°, 90°parabolic; Barnes 2004 disposition
M51 (NGC 5194/5)221.001.20.420°, 170°30°, 0°single-passage (Salo & Laurikainen 2000)
Bound coalescing (toy)80.501.01.045°, 0°45°, 90°elliptic, T ≈ 420 Myr; friction decays the orbit through repeated passages → coalescence + spheroidal remnant
Retrograde control191.001.01.0120°, 30°120°, 60°parabolic, both disks retrograde — no tails

Orbital geometry & disk orientation

19 kpc
closest approach of the two galaxy centers
1.00
e<1 bound, e=1 parabolic, e>1 hyperbolic flyby
60°
angle of disk A normal vs. orbital L̂; <90° prograde
60°
angle of disk B normal vs. orbital L̂; <90° prograde
30°
argument of line of nodes for disk A
60°
argument of line of nodes for disk B
1.0 ·10¹¹ M
total mass of galaxy A (halo + disk)
1.0 ·10¹¹ M
total mass of galaxy B (halo + disk)
parameter changes do not take effect until the leapfrog re-runs

3 · Boundary of validity  mandatory

Where rigid-potential restricted N-body is trustworthy — and where it isn't

Two things are removed from the equations: (i) self-gravity of the disks and (ii) all gas physics — pressure forces, shocks, cooling, star formation, and feedback. A third, dynamical friction, is no longer absent but is included only as the Chandrasekhar 1943 analytic prescription on rigid Hernquist profiles with a fixed lnΛ = 2 — a calibrated stand-in for the emergent wake, not the wake itself. The failure modes do not appear at the same time in the merger sequence.

What stays right

The morphology and the timing of tidal tail and bridge formation at first pericenter. Tail length grows linearly with time after first pericenter at ≈ vtail ≈ vesc(q), reproducing Hibbard & Mihos 1995 and Barnes & Hibbard 2009 to within ~10%.

valid · 0 → ~400 Myr post-peri
Prograde / retrograde asymmetry

Toomre & Toomre's central result that prograde disks form long thin tails and retrograde disks barely respond is purely kinematic. It is reproduced exactly by the restricted calculation — the retrograde control preset is the clean demonstration.

valid · independent of N-body refinement
Orbit decay: prescribed, not emergent

The orbital separation now shrinks between passages via the Chandrasekhar 1943 friction term, reproducing the qualitative Karl+ 2010 Antennae sequence (second pericenter a few hundred Myr after the first as friction lowers the apocenter). But lnΛ is a fixed constant and the host density rigid, so the rate of decay carries a factor-of-~2 uncertainty — coalescence times are indicative, not predictive (Boylan-Kolchin+ 2008).

prescribed · tmerge uncertain by ~2×
Disk self-gravity → bar/spiral feedback

Real disks respond to a perturbation by raising bars and spiral arms that amplify the tidal response. The restricted calculation gives only the linear tidal kick. Tail mass and width are accurate; inner disk structure (bars, ovals, m=2 spirals) is not reproduced. M51-like grand-design arms in the secondary are partly tidal and partly self-gravitating; see Dobbs+ 2010 for the breakdown.

fails · |r| < ~5 kpc inside each disk
No gas, no starbursts, no AGN fueling

Mergers funnel gas inward and trigger 10–100 Myr starbursts (Mihos & Hernquist 1996) and obscured AGN (Hopkins+ 2008). None of that is in this calculation. The tool produces a stellar-tracer prediction, not a luminosity or color prediction.

fails · entire baryon physics layer
Final coalescence and remnant

Coalescence now occurs: below 1 kpc the centers are locked to their COM and the disks phase-mix into a spheroidal swarm in the combined potential. But the real remnant is shaped by violent relaxation in a time-varying self-consistent potential (Barnes 1992, Hopkins+ 2009) — here the potentials stay frozen, so the de Vaucouleurs profile, kinematic misalignments, and boxy/disky isophotes are not reproduced. The remnant shown is a morphological cartoon: right shape class, wrong profile.

partial · remnant profile & kinematics unreliable

Bottom line. The restricted N-body model is a kinematic predictor of the first 300–500 Myr after pericenter: tail length, bridge geometry, sense of asymmetry, dependence on disk inclination. With the friction prescription it now also gives a plausible decay-and-coalescence sequence and a spheroidal end state — but merger timescales are indicative only, and remnant profile, kinematics, and every baryonic observable remain out of scope. The crossover to full SPH/GADGET/FIRE is well-mapped in the literature cited below.

References