Radial-Orbit Instability: How Radial Orbits Build a Bar

What the instability is and why it happens

The radial-orbit instability is a purely gravitational, collisionless instability: no gas, no collisions, no rotation are required. It afflicts stellar systems in which a large fraction of stars move on highly eccentric, nearly radial orbits — orbits that plunge deep toward the center and swing far out, carrying little angular momentum.

In a spherical potential each such orbit is, to a good approximation, a slowly precessing ellipse (a "rosette"). Individually these orbits are perfectly stable. The instability is collective: the combined self-gravity of many eccentric orbits amplifies any slight departure from perfect roundness. Because low-angular-momentum orbits precess very slowly, a faint non-spherical perturbation can grab them, align their long axes, and reinforce itself.

• Requires radial velocity anisotropy, quantified by the Fridman–Polyachenko–Shukhman parameter ξ ≡ 2T_r/T_t (isotropic ⇒ ξ = 1).
• Breaks spherical symmetry spontaneously — the endpoint is a triaxial, usually prolate bar-like body.
• Operates in the mean field; it is not driven by two-body relaxation.

The mechanism: torque, precession, and orbit trapping

The clearest physical picture, developed by Lynden-Bell, Polyachenko, Palmer, Papaloizou and Merritt, is one of orbital resonance and self-reinforcing alignment. Suppose a tiny bar-shaped overdensity appears by chance. An elongated orbit whose long axis lies near the bar feels a gravitational torque.

• As the orbit's apsis approaches the bar, the torque speeds up its precession; after it passes, the torque removes angular momentum and it slows down.
• The net effect is that orbits spend longer lingering near the bar's long axis, piling up mass there.
• That extra mass deepens the bar potential, torquing still more orbits into alignment — a positive feedback loop.

Formally the growing mode is a slow, near-stationary l = 2 (m = 2) pattern tied to the precession frequencies of low-J orbits, not to the fast radial oscillation. When enough orbits are trapped near this slowly rotating figure, their loop orbits convert into box orbits that respect a bar's symmetry, locking in the triaxial shape. The instability saturates once the reservoir of alignable eccentric orbits is exhausted.

Key numbers and a worked threshold

The single most-quoted number is the stability boundary in the anisotropy parameter ξ = 2T_r/T_t, where T_r and T_t are the radial and (both) tangential contributions to the kinetic-energy tensor. Polyachenko & Shukhman (1981) argued for an approximately universal threshold:

• ξ_crit ≈ 1.7 ± 0.25 — systems with ξ below this are stable, above it are ROI-unstable.
• The exact value is model-dependent: normal-mode analyses give ~1.4–1.6 for some profiles, ~1.9 for Plummer spheres, and up to ~2.3–2.5 for certain f(Q) models.

An equivalent local statement uses the anisotropy β ≡ 1 − σ_t²/(2σ_r²); the boundary corresponds to σ_r/σ_t ≈ 1.5. For Osipkov–Merritt models, where f = f(Q) with Q = E + L²/2r_a², the system goes unstable once the anisotropy radius drops below r_a ≈ 0.3 r_h (r_h the half-mass radius). Worked example: a cold collapse starting from virial ratio 2T/|W| ≲ 0.2 generically overshoots into strong radial anisotropy and ends prolate with axis ratios reaching ~2.5:1 — right up against the ~3:1 limit set by the bending instability. Growth takes only a few crossing times, t_cross ≈ 1/√(Gρ).

How it is detected and where it appears

ROI is a theorist's instability — you cannot watch a galaxy go unstable in real time — so evidence is indirect and numerical. Three lines of evidence stand out:

• N-body simulations: Since Hénon, Barnes, Merritt, Aguilar and many others, controlled collisionless simulations show radially anisotropic spheres buckling into bars exactly when ξ crosses ~1.7. This is the cleanest demonstration.
• Cold-collapse experiments: A cloud of stars released from rest collapses, develops strong radial anisotropy on the way in, and relaxes to a triaxial elliptical-like object rather than a sphere.
• Observed shapes: The prevalence of triaxiality in elliptical galaxies and prolate/boxy structure in bulges and dark-matter halos is widely attributed, at least in part, to ROI acting during dissipationless collapse and mergers.

Observationally one looks for triaxial kinematics: minor-axis rotation, isophotal twists, and radially rising velocity dispersion profiles that imply β > 0 in the outskirts of ellipticals and in stellar halos measured with integral-field spectroscopy.

How ROI differs from its close cousins

The radial-orbit instability is easy to confuse with the classical bar instability of a cold rotating disk, but the physics is distinct:

• ROI needs no rotation. Its energy source is anisotropy (excess radial motion), and the resulting bar is essentially non-rotating or very slowly tumbling. Disk bars, by contrast, arise in cold, dynamically rotating disks with low Toomre Q and spin rapidly.
• The disk-bar order parameter is velocity dispersion / Q; the ROI order parameter is velocity anisotropy / ξ.
• The bending (fire-hose) instability is the opposite of ROI in effect: it thickens and rounds out systems that become too elongated (axis ratio ≳ 3:1), so it caps how far ROI can drive prolateness.

ROI also has a triaxial counterpart — an instability that persists in already-barred systems and can drive further shape change — and it is sometimes discussed alongside the Lynden-Bell / Kalnajs bar-formation mechanisms, in which the mutual gravity of aligned eccentric orbits generates a slowly rotating bar. Even MOND versions have been studied, where modified gravity shifts the threshold.

Significance, open questions, and famous cases

ROI matters because it is one of the only mechanisms that lets a hot, non-rotating, collisionless system break spherical symmetry on its own. That single fact underpins how we think elliptical galaxies, spheroidal bulges, and dark-matter halos acquire their triaxial shapes without needing tides or mergers to do all the work.

• Universality of ξ_crit: Is the ~1.7 threshold truly independent of the density profile? Studies find scatter from ~1.4 to ~2.5, so the "universal" value is really a rough guide, not a law.
• The shape mismatch: Cold collapse driven by ROI predicts strongly prolate ~2.5:1 systems, yet observed ellipticals peak near axis ratio c/a ≈ 0.8 (rounder). The competition between ROI and the bending instability does not fully close this gap — an open problem.
• Role of a central mass or halo: A dense center, a black hole, or an embedding halo can partially suppress ROI by scattering the very radial orbits it feeds on.

Historically, ROI was foreshadowed by Antonov (1973), sharpened into the ξ-criterion by Fridman, Polyachenko & Shukhman (1981), and given its resonant-torque interpretation by Palmer & Papaloizou (1987) and Weinberg (1991). It remains a standard test problem for any new collisionless N-body code.

Radial-orbit instability compared with related symmetry-breaking instabilities in stellar systems

Instability	Host system	Driver / order parameter	Outcome
Radial-orbit instability	Spherical/near-spherical, non-rotating	Radial anisotropy, ξ = 2Tr/Tt ≳ 1.7	Triaxial (prolate) bar
Bar instability (disk)	Cold rotating disk	Low Toomre Q, low velocity dispersion	Rotating stellar bar
Bending (fire-hose) instability	Very flattened / elongated system	Axis ratio ≳ 3:1	Vertical thickening, limits elongation
Jeans instability	Gas or stars, any geometry	Self-gravity vs pressure, λ > λ_J	Gravitational collapse / fragmentation
Two-stream / counter-rotating	Counter-rotating stellar streams	Relative streaming of populations	Warps, m=1/m=2 modes