Toomre Stability Criterion

A galactic disk is a fight among three protagonists, and the Toomre criterion is the scorecard that says who wins. Self-gravity is the aggressor: any local overdensity pulls its neighbours inward and tries to grow. Two defenders resist. The first is random motion — pressure in a gas, radial velocity dispersion $\sigma_R$ in a swarm of stars — which fights compression hardest on small scales, exactly as in the classical Jeans analysis. The second defender is new, and it is what makes a disk different from a static cloud: rotation. Displace a parcel of matter radially inward and the Coriolis force deflects it, so instead of falling it swings around a closed epicyclic loop at frequency $\kappa$ . That restoring force is most effective over large regions. Stability therefore demands that no wavelength slip past both defenders at once. Alar Toomre in 1964 linearised the equations of motion for a razor-thin disk, wrote the dispersion relation for tightly wound axisymmetric density waves, and asked whether any wavenumber could grow. The answer collapses to a single dimensionless number, $Q$ . When $Q \ge 1$ the short-wavelength reach of pressure and the long-wavelength reach of rotation overlap, leaving no gap, and every axisymmetric mode is stable; when $Q < 1$ a gap opens and an intermediate band of wavelengths grows exponentially into rings and clumps. The same result was found independently by Safronov (1960) for gaseous protoplanetary disks. The criterion is deliberately narrow — axisymmetric modes, tight-winding (WKB) approximation — so it misses the trailing spirals that swing amplification produces even at $Q$ modestly above one. That is why real disks hover at $Q \sim 1$ – $2$ : instability heats the disk, raising $Q$ , until it self-regulates back to the edge of stability. It is the single most-used diagnostic in galactic dynamics.

Key takeaways

A differentially rotating, self-gravitating disk is locally stable to axisymmetric perturbations when the Toomre parameter is $Q = \dfrac{\sigma_R\,\kappa}{3.36\,G\,\Sigma} \ge 1$ for stars, or $Q = \dfrac{c_s\,\kappa}{\pi\,G\,\Sigma} \ge 1$ for gas.
It falls out of the tight-winding dispersion relation $\omega^2 = \kappa^2 - 2\pi G\Sigma\,|k| + k^2 c_s^2$ , in which rotation ( $\kappa^2$ ) stabilises long wavelengths, pressure ( $k^2c_s^2$ ) stabilises short ones, and self-gravity ( $-2\pi G\Sigma|k|$ ) attacks the middle.
The reference scale is the critical wavelength $\lambda_{\mathrm{crit}} = 4\pi^2 G\Sigma/\kappa^2$ — the longest wavelength a cold disk cannot resist; the fastest-growing mode near marginal stability sits at half of it, $\lambda_m = 2\pi^2 G\Sigma/\kappa^2$ .
The epicyclic frequency is $\kappa^2 = R\,\mathrm{d}\Omega^2/\mathrm{d}R + 4\Omega^2$ , giving $\kappa = 2\Omega$ (solid body), $\sqrt2\,\Omega$ (flat rotation curve), and $\Omega$ (Keplerian).
The criterion is strictly local and axisymmetric — a disk with $Q \gtrsim 1$ is still unstable to non-axisymmetric spiral modes via swing amplification, so $Q \sim 1$ – $2$ is the self-regulated state of galactic and protoplanetary disks.
Key idea: rotation stabilises long wavelengths through the epicyclic (Coriolis) restoring force $\kappa$ , and dispersion stabilises short wavelengths through pressure; the disk is stable only if these two protected windows overlap, which happens precisely when $Q \gtrsim 1$ .

The core insight

Rotation stabilises long wavelengths through the epicyclic/Coriolis restoring force $\kappa$ , and dispersion stabilises short wavelengths through pressure; the disk is stable only if these two windows overlap, which happens precisely when $Q \gtrsim 1$ .

Where this lives today

Toomre's $Q$ is a working number, not a museum piece. In the Milky Way, Gaia's astrometry now measures $\sigma_R$ , $\kappa$ and the surface density star-by-star, mapping $Q(R)$ across the disk and testing self-regulation directly; the age–velocity-dispersion relation of nearby stars is read as the heating that keeps $Q$ near unity. In external galaxies, resolved HI and CO surveys (THINGS, HERACLES) plot $Q_{\mathrm{gas}}(R)$ and the two-component $Q_{\mathrm{eff}}$ against where star formation switches on, refining the classic Kennicutt threshold. At high redshift, integral-field spectrographs (SINS, KMOS³D) find turbulent gas-rich disks with $Q \sim 1$ fragmenting into giant clumps of $10^8$ – $10^9\ M_\odot$ — Toomre instability caught building galaxies. In the protoplanetary regime, ALMA images young disks such as Elias 2-27 whose spiral arms and rings point to gravitational instability, and the disk-fragmentation channel for giant-planet formation turns on whether the cold outer disk reaches $Q < 1$ while cooling fast enough ( $t_{\mathrm{cool}} \lesssim \Omega^{-1}$ ). The open frontier is what sets the self-regulated value of $Q$ , and how the local axisymmetric criterion connects to the global non-axisymmetric modes that actually sculpt spiral galaxies.

The physical picture

Picture a small overdense patch in the disk. Its self-gravity pulls surrounding material toward it and, left alone, that material would keep falling in. Two guardians stand in the way. The first is pressure, or for stars the random velocity dispersion: squeeze a small region and its internal motions push back sharply, so pressure defends short wavelengths best. The second guardian is rotation. Move matter radially and the Coriolis force curls it sideways into an epicyclic orbit rather than letting it plunge — a defence that works best over large regions, i.e. long wavelengths. Collapse can only win in a gap between the two guardians, and the number $Q$ measures whether such a gap exists at all. The clean way to see this is the dispersion relation, which assigns every wavelength a value of $\omega^2$ : positive means the mode oscillates and is stable, negative means it grows exponentially. The first figure plots it. As $Q$ falls, the curve's minimum sinks; at $Q = 1$ it just kisses zero at a single wavelength, and below $Q = 1$ a whole band of intermediate wavelengths goes unstable.

Figure

The dispersion relation ω²/κ² = 1 − K + (Q²/4)K² versus scaled wavenumber K = k/k_crit for three disks. Q = 1.5 stays above zero everywhere (stable), Q = 1.0 just touches ω² = 0 at K = 2 (marginal), and Q = 0.7 dips negative across a band of wavelengths where perturbations grow exponentially (unstable)

Gravity's pull scales as $-2\pi G\Sigma|k|$ (linear in $k$ ) while pressure's restoring stiffness scales as $+k^2c_s^2$ (quadratic), so gravity wins over an intermediate band if it wins anywhere. Rotation supplies the flat $+\kappa^2$ floor that lifts the whole curve; make the floor high enough (large $\kappa$ ) or the pressure stiff enough (large $c_s$ ) and the minimum never reaches zero. And because instability, once triggered, scatters stars and stirs turbulence, it raises $\sigma_R$ or $c_s$ and pushes $Q$ back toward one — a thermostat that keeps observed disks near marginal stability. The second figure shows the mechanism behind the rotation floor: how a displaced star is turned back by the Coriolis force into a closed epicyclic loop.

Figure

Rotation stabilises long wavelengths. A star displaced radially outward is turned by the Coriolis force into a closed epicyclic loop of frequency κ rather than drifting away. The loop is elongated in the tangential direction with axis ratio 2Ω : κ. For a flat rotation curve κ = √2 Ω, and at the Sun Ω ≈ 27 and κ ≈ 38 km/s per kpc

The core equations

The subject is organised around one dimensionless ratio and the dispersion relation it comes from. We state the two forms of the stability parameter first, then the dispersion relation that produces them, the length scale it selects, the epicyclic frequency that supplies the rotation floor, and the two-component generalisation for disks of both stars and gas. The stellar Toomre parameter compares the stabilising product of dispersion and rotation against the destabilising self-gravity of the surface density.

Q_{\star} \equiv \frac{\sigma_R\,\kappa}{3.36\,G\,\Sigma} \ge 1 \qquad (1)

where: $\sigma_R$ — radial velocity dispersion of the stars; $\kappa$ — epicyclic frequency; $\Sigma$ — surface mass density of the disk; $G$ — Newton's gravitational constant; $3.36$ — numerical constant from the collisionless (kinetic) reduction factor of Lin, Shu & Kalnajs. Read it as a race: raise $\sigma_R$ or $\kappa$ and stability improves; raise $\Sigma$ and gravity gains. For a gaseous disk the sound speed replaces the dispersion and the constant changes from $3.36$ to $\pi$ , reflecting the different response of a coherent fluid versus a collisionless swarm.

Q_{\mathrm{gas}} \equiv \frac{c_s\,\kappa}{\pi\,G\,\Sigma} \ge 1 \qquad (2)

where: $c_s$ — sound speed (or effective turbulent speed) of the gas; $\pi \approx 3.14159$ — fluid constant. Because $\pi < 3.36$ , a gas disk of the same $\Sigma$ , $\kappa$ and characteristic speed sits at a lower $Q$ than a stellar disk — gas fragments first, which is exactly why stars form out of gas rather than the stellar disk collapsing. Both criteria are the marginal-stability condition of a single dispersion relation. For the tight-winding (WKB) axisymmetric mode of a razor-thin gas disk it reads:

\omega^2 = \kappa^2 - 2\pi G\Sigma\,|k| + k^2 c_s^2 \qquad (3)

where: $\omega$ — perturbation frequency, with $\omega^2 < 0$ signalling exponential growth; $k = 2\pi/\lambda$ — radial wavenumber; the three terms are the rotation floor ( $+\kappa^2$ ), self-gravity ( $-2\pi G\Sigma|k|$ ), and pressure ( $+k^2c_s^2$ ). The stellar version replaces $c_s^2$ with $\sigma_R^2$ and multiplies the gravity term by a kinetic reduction factor $\mathcal{F} < 1$ . Minimising $\omega^2$ over $k$ selects a wavenumber, and hence a length scale, which is the characteristic size of the structures a marginally unstable disk makes.

k_{\mathrm{crit}} = \frac{\kappa^2}{2\pi G\Sigma}, \qquad \lambda_{\mathrm{crit}} = \frac{2\pi}{k_{\mathrm{crit}}} = \frac{4\pi^2 G\Sigma}{\kappa^2} \qquad (4)

where: $\lambda_{\mathrm{crit}}$ — the critical wavelength, the longest wavelength a cold ( $c_s \to 0$ ) disk cannot stabilise; $k_{\mathrm{crit}}$ — the smallest unstable wavenumber of that cold disk. When pressure is restored, the fastest-growing mode near $Q = 1$ sits at $k_m = \pi G\Sigma/c_s^2 = 2\,k_{\mathrm{crit}}$ , i.e. at half the critical wavelength, $\lambda_m = 2\pi^2 G\Sigma/\kappa^2 = \lambda_{\mathrm{crit}}/2$ . This factor of two is a common trap and is derived carefully below. The rotation floor $\kappa$ is not a free parameter: it is fixed by how the circular orbital frequency $\Omega(R)$ varies with radius.

\kappa^2 = R\frac{\mathrm{d}\Omega^2}{\mathrm{d}R} + 4\Omega^2 = 2\Omega\left(2\Omega + R\frac{\mathrm{d}\Omega}{\mathrm{d}R}\right) \qquad (5)

where: $\Omega(R) = v_c/R$ — the circular angular frequency at radius $R$ . The three canonical cases are solid-body rotation ( $\kappa = 2\Omega$ ), a flat rotation curve ( $\kappa = \sqrt2\,\Omega$ ), and a Keplerian fall-off ( $\kappa = \Omega$ ); every real disk lies in $\Omega \le \kappa \le 2\Omega$ . Real disks are neither pure gas nor pure stars, so the operative quantity is a combined two-component parameter. The Romeo–Wiegert approximation weights the two Toomre parameters by their wavenumber response.

\frac{1}{Q_{\mathrm{eff}}} = \frac{2}{Q_\star}\frac{q_\star}{1+q_\star^2} + \frac{2}{Q_{\mathrm{gas}}}\frac{W\,q_g}{1+q_g^2} \qquad (6)

where: $Q_{\mathrm{eff}}$ — effective stability parameter, unstable when below one; $W = 2\sigma_R\sigma_g/(\sigma_R^2 + \sigma_g^2)$ — the coupling weight between the cold gas and the hotter stars; $q_\star, q_g$ — dimensionless wavenumbers for each component evaluated at the least-stable scale. A thin layer of cold gas can drive $Q_{\mathrm{eff}} < 1$ even when the stellar disk alone has $Q_\star > 1$ — the gas destabilises the stars.

Derivation ADeriving the Toomre criterion from the gas dispersion relation

The goal is to turn the three-term dispersion relation into the single number $Q_{\mathrm{gas}}$ by finding the most dangerous wavelength and demanding that even it does not grow. The strategy is standard: the disk is unstable if any mode has $\omega^2 < 0$ , so we locate the wavenumber that minimises $\omega^2(k)$ , insist that this global minimum be non-negative, and read off the condition. Because $\omega^2(k)$ is an upward-opening parabola in $|k|$ , its single minimum is exactly the mode that first goes unstable as the disk cools, making the calculation both simple and exact within WKB.

Step 1. Start from the tight-winding gas dispersion relation, $\omega^2(k) = \kappa^2 - 2\pi G\Sigma\,|k| + k^2 c_s^2$ , and treat $|k|$ as the variable. Why: the three terms encode rotation, self-gravity and pressure separately, and only their combination decides stability, so we must optimise over all $k$ .

Step 2. Differentiate and set to zero, $\mathrm{d}\omega^2/\mathrm{d}k = -2\pi G\Sigma + 2k c_s^2 = 0$ , giving $k_{\min} = \pi G\Sigma/c_s^2$ . Why: the positive $k^2$ coefficient makes the one extremum a minimum, so this $k_{\min}$ is where instability first appears.

Step 3. Substitute $k_{\min}$ back in. The gravity term becomes $-2(\pi G\Sigma)^2/c_s^2$ and the pressure term $+(\pi G\Sigma)^2/c_s^2$ , combining to $\omega^2_{\min} = \kappa^2 - (\pi G\Sigma)^2/c_s^2$ . Why: evaluating at the minimum collapses two competing terms into a clean contest between rotation ( $\kappa^2$ ) and the gravity-pressure combination.

Step 4. Demand $\omega^2_{\min} \ge 0$ , i.e. $\kappa^2 c_s^2 \ge (\pi G\Sigma)^2$ . Positive square roots give $Q_{\mathrm{gas}} = c_s\kappa/(\pi G\Sigma) \ge 1$ . Why: if the deepest point of the curve is non-negative then every mode oscillates, so this one inequality certifies stability against all axisymmetric wavelengths at once. *Note: as $\kappa \to 0$ the rotation floor vanishes and $\omega^2_{\min} = -(\pi G\Sigma)^2/c_s^2 < 0$ , recovering unconditional 2D Jeans instability — every axisymmetric mode collapses when there is no rotation to defend long wavelengths.*

Go deeper · Derivation B · 1 minThe critical wavelength and the onset of ring fragmentation

Derivation BThe critical wavelength and the onset of ring fragmentation

The aim is to pin down the physical size of the structures a marginally unstable disk produces, and to resolve a factor-of-two that trips up many first readings: the critical wavelength bounding a cold disk's unstable range is twice the most-unstable wavelength that grows fastest near $Q = 1$ . We evaluate the two natural wavenumbers — the smallest unstable one for a pressureless disk, and the one minimising $\omega^2$ when pressure is present — and relate them at marginal stability, so the fragment spacing seen in simulations follows from $\Sigma$ and $\kappa$ alone.

Step 1. Take the cold-disk limit $c_s \to 0$ , where $\omega^2 = \kappa^2 - 2\pi G\Sigma|k|$ . This is unstable whenever $|k| > \kappa^2/(2\pi G\Sigma) \equiv k_{\mathrm{crit}}$ , so the cold disk is unstable to every wavelength shorter than $\lambda_{\mathrm{crit}} = 2\pi/k_{\mathrm{crit}} = 4\pi^2 G\Sigma/\kappa^2$ . Why: with no pressure to defend the short end, only rotation resists, and it fails below this ceiling — so $\lambda_{\mathrm{crit}}$ is the longest unstable wavelength, not the fastest-growing one.

Step 2. Restore pressure; the minimum of $\omega^2$ sits at $k_m = \pi G\Sigma/c_s^2$ . At $Q = 1$ , $c_s = \pi G\Sigma/\kappa$ , so $k_m = \kappa^2/(\pi G\Sigma) = 2\,k_{\mathrm{crit}}$ . Why: substituting the marginal sound speed ties the pressure-selected wavenumber to the rotation scale, delivering a clean factor of two — the fastest-growing mode has twice the wavenumber, half the wavelength.

Step 3. Convert to a wavelength, $\lambda_m = 2\pi/k_m = 2\pi^2 G\Sigma/\kappa^2 = \tfrac12\lambda_{\mathrm{crit}}$ . This is the fragmentation scale: a disk pushed just below $Q = 1$ breaks into rings, and later clumps, spaced by roughly $\lambda_m$ . Why: this is the length that appears in N-body and hydrodynamic simulations as the clump separation; quoting $\lambda_{\mathrm{crit}}$ overestimates it twofold.

Step 4. The scale is purely local — it depends only on $\Sigma$ and, through $\kappa$ , the rotation, not on any global property. Why: this locality is what makes $Q(R)$ a radius-by-radius diagnostic, predicting where a disk fragments without a global eigenvalue problem. *Note: the cold-disk bound $\lambda_{\mathrm{crit}} = 4\pi^2 G\Sigma/\kappa^2$ and the fastest-growing $\lambda_m = 2\pi^2 G\Sigma/\kappa^2$ differ by exactly a factor of two — a distinction worth keeping in mind, since the two values are quoted interchangeably in some sources.*

Go deeper · Derivation C · 1 minThe epicyclic frequency from radial oscillations about a circular orbit

Derivation CThe epicyclic frequency from radial oscillations about a circular orbit

The purpose is to derive the rotation floor $\kappa$ from first principles, showing it is the frequency of small radial oscillations of a star about its circular orbit and is fixed entirely by the rotation curve. The method reduces the 2D orbital motion to a 1D problem in an effective potential using conservation of angular momentum, then expands about the circular-orbit radius so the restoring force — and the oscillation frequency — appears as a second derivative re-expressed through the observable $\Omega(R)$ .

Step 1. In an axisymmetric potential $\Phi(R)$ the angular momentum $L_z = R^2\dot\phi$ is conserved, so the radial motion is that of a 1D particle in $\Phi_{\mathrm{eff}}(R) = \Phi(R) + L_z^2/(2R^2)$ . Why: eliminating the azimuthal motion in favour of conserved $L_z$ turns a 2D orbit into a 1D oscillation, the setting in which a well-defined frequency exists.

Step 2. A circular orbit sits at the guiding radius $R_g$ where $\mathrm{d}\Phi_{\mathrm{eff}}/\mathrm{d}R = 0$ , and small excursions oscillate with $\kappa^2 = \left.\mathrm{d}^2\Phi_{\mathrm{eff}}/\mathrm{d}R^2\right|_{R_g}$ . Why: near a potential minimum any system does simple harmonic motion whose squared frequency is the second derivative, so $\kappa$ is the stiffness of the radial well.

Step 3. Carrying out the derivatives with $\Omega^2(R) = (1/R)\,\mathrm{d}\Phi/\mathrm{d}R$ gives $\kappa^2 = R\,\mathrm{d}\Omega^2/\mathrm{d}R + 4\Omega^2$ . The canonical curves follow: solid body ( $\Omega$ constant) gives $\kappa = 2\Omega$ ; flat ( $\Omega \propto R^{-1}$ ) gives $\kappa^2 = 2\Omega^2$ , i.e. $\kappa = \sqrt2\,\Omega$ ; Keplerian ( $\Omega \propto R^{-3/2}$ ) gives $\kappa = \Omega$ . Why: these let you read $\kappa$ straight off a measured rotation curve, and show $\kappa$ always lies between $\Omega$ and $2\Omega$ , keeping the rotation floor comparable to the orbital frequency across a real disk. *Note: the three canonical values $\kappa = \{2, \sqrt2, 1\}\,\Omega$ for solid-body, flat, and Keplerian curves follow directly from the formula above; at the Sun, $\Omega_\odot \approx 27$ and $\kappa_\odot \approx 38$ km s⁻¹ kpc⁻¹, in line with $\kappa = \sqrt2\,\Omega$ for the nearly-flat Milky Way rotation curve.*

Limiting cases

Reading the dispersion relation $\omega^2 = \kappa^2 - 2\pi G\Sigma|k| + k^2 c_s^2$ in each regime shows which agent controls stability where, and recovers the familiar sub-cases as limits.

Case	Result	Meaning
Long wavelength ( $k \to 0$ )	$\omega^2 \to \kappa^2 > 0$	Rotation dominates; the epicyclic floor stabilises every sufficiently large-scale mode regardless of $Q$ .
Short wavelength ( $k \to \infty$ )	$\omega^2 \to k^2 c_s^2 > 0$	Pressure dominates; small scales are always stable sound waves, the disk analogue of small-scale Jeans stability.
Non-rotating ( $\kappa \to 0$ )	$\omega^2 = k^2 c_s^2 - 2\pi G\Sigma\|k\|$	Recovers 2D Jeans instability: all $k < 2\pi G\Sigma/c_s^2$ collapse. Without rotation a cold disk has no large-scale defence at all.
Cold disk ( $c_s, \sigma_R \to 0$ )	unstable for all $\lambda < \lambda_{\mathrm{crit}}$	Only rotation resists; everything shorter than $\lambda_{\mathrm{crit}} = 4\pi^2 G\Sigma/\kappa^2$ grows. $Q \to 0$ .
Marginal stability ( $Q = 1$ )	$\omega^2_{\min} = 0$ at $k_m = \kappa^2/(\pi G\Sigma)$	The curve just touches zero at a single wavenumber; a slightly colder disk fragments at $\lambda_m = \lambda_{\mathrm{crit}}/2$ .

Common trapA disk with Q > 1 is fully stable

Not so — $Q \ge 1$ certifies stability only against axisymmetric ( $m = 0$ ) perturbations, and only in the tight-winding WKB limit. A disk with $Q = 1.5$ can still be violently unstable to non-axisymmetric spiral modes: shearing waves are transiently amplified by swing amplification, which stays effective up to $Q \sim 2$ . So $Q > 1$ means "no ring-like fragmentation", not "globally stable" — the persistent spirals of real galaxies are the visible proof. Two smaller traps ride along. First, use the right constant: $3.36$ for stars, $\pi$ for gas, a difference that shifts $Q$ by about 15 percent. Second, $\lambda_{\mathrm{crit}} = 4\pi^2 G\Sigma/\kappa^2$ is the longest unstable wavelength of a cold disk, not the fragment spacing; the fastest-growing mode is at half that, $\lambda_m = 2\pi^2 G\Sigma/\kappa^2$ .

Worked examples

1 · Is the solar neighbourhood stable? (fully worked)

Estimate the stellar Toomre parameter at the Sun's radius using $\sigma_R \approx 38$ km s⁻¹ (old thin-disk stars), $\kappa \approx 37$ km s⁻¹ kpc⁻¹, and stellar surface density $\Sigma \approx 50\ M_\odot\,\mathrm{pc}^{-2}$ .

Step 1. Fix units so $G$ is convenient. In astronomers' units $G = 4.301 \times 10^{-3}\ \mathrm{pc}\,M_\odot^{-1}\,(\mathrm{km\,s^{-1}})^2$ . Convert $\kappa$ to per-parsec: $\kappa = 37\ \mathrm{km\,s^{-1}\,kpc^{-1}} = 0.037\ \mathrm{km\,s^{-1}\,pc^{-1}}$ , so that every quantity is expressed in pc, $M_\odot$ and km s⁻¹.

Step 2. Form the numerator, the stabilising product: $\sigma_R\,\kappa = 38 \times 0.037 = 1.406\ (\mathrm{km\,s^{-1}})^2\,\mathrm{pc}^{-1}$ .

Step 3. Form the denominator, the destabilising self-gravity term with the stellar constant $3.36$ : $3.36\,G\,\Sigma = 3.36 \times (4.301\times10^{-3}) \times 50 = 0.7226\ (\mathrm{km\,s^{-1}})^2\,\mathrm{pc}^{-1}$ .

Step 4. Divide: $Q_\star = 1.406/0.7226 = \mathbf{1.95}$ . The solar neighbourhood is stable to axisymmetric collapse, but only by a factor of two — sitting squarely in the self-regulated $Q \sim 1$ – $2$ band that all cold galactic disks occupy. It is stable enough to avoid runaway fragmentation, marginal enough to keep making spiral structure. Q_⋆ ≈ 2 — marginally stable, exactly as observed.

2 · The fragmentation scale of the Milky Way disk

Using the same $\Sigma = 50\ M_\odot\,\mathrm{pc}^{-2}$ and $\kappa = 0.037\ \mathrm{km\,s^{-1}\,pc^{-1}}$ , find the critical and most-unstable wavelengths.

Step 1. Compute $\kappa^2 = (0.037)^2 = 1.369\times10^{-3}\ (\mathrm{km\,s^{-1}})^2\,\mathrm{pc}^{-2}$ and $G\Sigma = 4.301\times10^{-3} \times 50 = 0.2151\ (\mathrm{km\,s^{-1}})^2\,\mathrm{pc}^{-1}$ .

Step 2. The critical (longest unstable) wavelength: $\lambda_{\mathrm{crit}} = 4\pi^2 G\Sigma/\kappa^2 = 39.48 \times 0.2151 / (1.369\times10^{-3}) = 6.2\times10^{3}\ \mathrm{pc} = \mathbf{6.2\ kpc}$ .

Step 3. The fastest-growing fragment spacing is half of this, $\lambda_m = \lambda_{\mathrm{crit}}/2 = \mathbf{3.1\ kpc}$ . These are enormous — a substantial fraction of the disk's radius — which is why a disk safely at $Q \approx 2$ does not shred into kiloparsec rings: the modes that would fragment it are the ones rotation most easily stabilises. λ_crit ≈ 6 kpc, fragment scale ≈ 3 kpc.

3 · A gas disk and the flat-rotation-curve shortcut

The outer Milky Way has a nearly flat rotation curve with $v_c \approx 220$ km s⁻¹ at $R \approx 8.2$ kpc, gas sound/turbulent speed $c_s \approx 8$ km s⁻¹, and gas surface density $\Sigma_g \approx 10\ M_\odot\,\mathrm{pc}^{-2}$ . Find $\kappa$ and then $Q_{\mathrm{gas}}$ .

Step 1. Circular frequency $\Omega = v_c/R = 220/8.2 = 26.8\ \mathrm{km\,s^{-1}\,kpc^{-1}}$ . Equation (5) gives the flat-curve shortcut $\kappa = \sqrt2\,\Omega = 37.9\ \mathrm{km\,s^{-1}\,kpc^{-1}} = 0.0379\ \mathrm{km\,s^{-1}\,pc^{-1}}$ — the value used above.

Step 2. Numerator $c_s\,\kappa = 8 \times 0.0379 = 0.303$ ; denominator with the gas constant $\pi$ is $\pi\,G\,\Sigma_g = 3.1416 \times (4.301\times10^{-3}) \times 10 = 0.1351$ (both in $(\mathrm{km\,s^{-1}})^2\,\mathrm{pc}^{-1}$ ).

Step 3. Divide: $Q_{\mathrm{gas}} = 0.303/0.1351 = \mathbf{2.2}$ . The outer gas disk is stable, consistent with the drop-off of star formation beyond the optical radius where $Q_{\mathrm{gas}}$ climbs above one. The gas $Q$ is comparable to the stellar $Q$ despite far colder gas, because the gas surface density is five times smaller. Q_gas ≈ 2, stable; κ = √2 Ω short-cuts the calculation.

Physics of the portrait

The Toomre criterion carries one name because one 1964 paper turned a tangle of competing effects into a single number — and because its author spent the following decades insisting that the number was only the beginning of the story.

Origins · 1 minAlar Toomre and how this was discovered

Alar Toomre

b. 1937 · galactic dynamics and disk stability

Alar Toomre was born in 1937 in Tallinn, Estonia, fled the Soviet occupation as a child, and became an applied mathematician at MIT after taking his doctorate in fluid dynamics at Manchester. His 1964 paper "On the gravitational stability of a disk of stars" (ApJ 139, 1217) did for galactic disks what Jeans had done for clouds: it reduced local stability to one dimensionless parameter, the $Q$ that now bears his name, complete with the kinetic reduction factor and the constant $3.36$ that distinguish a collisionless star swarm from a fluid. Two years later, with William Julian, he showed that even $Q > 1$ disks amplify shearing waves — swing amplification — seeding the transient spiral arms a purely axisymmetric analysis cannot produce. This put him at odds with the Lin–Shu picture of long-lived quasi-stationary spirals; Toomre argued instead that spiral structure is recurrent and transient, continually regenerated. With his brother Jüri he produced the celebrated 1972 simulations of interacting galaxies, reproducing the tidal bridges and tails of systems like the Antennae from gravity alone and founding the modern study of galaxy mergers (the "Toomre sequence"). His hallmark was to force a qualitatively-described phenomenon to yield a clean, checkable number — while never mistaking that number for the whole truth.

History

1960

— Safronov derives the gravitational-instability condition for a gaseous disk in the context of planet formation

1964

— Toomre publishes the axisymmetric stability parameter Q for stellar disks

1965

— Goldreich and Lynden-Bell analyse gas-disk fragmentation and shearing instabilities

1966

— Julian and Toomre describe swing amplification of non-axisymmetric shearing waves

1989

— Kennicutt links galactic star-formation thresholds to the gas Toomre Q

2011

— Romeo and Wiegert give the two-component stars-plus-gas effective Q

Going further

Texts. Binney & Tremaine, Galactic Dynamics, 2nd ed., §6.2 (dispersion relation, gas and stellar Q, the reduction factor) · Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., Ch. 24–25 · Mo, van den Bosch & White, Galaxy Formation and Evolution, §11.1 (disk instabilities) · Toomre, A. 1964, ApJ, 139, 1217 · Safronov, V. S. 1960, Ann. Astrophys., 23, 979 · Goldreich & Lynden-Bell 1965, MNRAS, 130, 125. Key results to keep in hand: the minimisation $k_{\min} = \pi G\Sigma/c_s^2$ yields $Q_{\mathrm{gas}} = c_s\kappa/\pi G\Sigma$ ; the epicyclic identity is $\kappa^2 = R\,\mathrm{d}\Omega^2/\mathrm{d}R + 4\Omega^2$ with limits $\kappa = \{2, \sqrt2, 1\}\Omega$ ; the critical wavelength $\lambda_{\mathrm{crit}} = 4\pi^2 G\Sigma/\kappa^2$ is the longest unstable mode of a cold disk, while the fastest-growing mode at $Q = 1$ is $\lambda_m = 2\pi^2 G\Sigma/\kappa^2 = \lambda_{\mathrm{crit}}/2$ — a factor of two quoted inconsistently in some sources; and the solar-neighbourhood worked values are $Q_\star \approx 1.95$ and $\lambda_{\mathrm{crit}} \approx 6.2$ kpc.

Prerequisites. Collisionless Boltzmann Equation · Orbits in Axisymmetric and Triaxial Potentials · Jeans Equations · Distribution Function and Phase Space. Leads to. Galactic Potential Models · Gravitational N-Body Problem · Virial Theorem in Stellar Systems.

Ask the tutor: Ask me to walk through why a disk safely at $Q \approx 2$ still grows spiral arms, or to work out $Q$ and the fragmentation scale for any disk you can give surface density and rotation for.

Is Q a local number or a global one?

It is strictly local. $Q = Q(R)$ is evaluated point-by-point from the surface density, dispersion and rotation at that radius, and it certifies stability only against short-wavelength axisymmetric perturbations there, in the tight-winding approximation. That locality is a feature: it lets you draw a curve of $Q$ across a galaxy and mark exactly where the disk crosses into instability, without solving a global mode problem. But it is also a limitation — genuinely global, non-axisymmetric modes (bars, grand-design spirals) are not captured, and a disk can be locally stable everywhere yet globally unstable to a large-scale mode.

Why is there a factor of two between the critical and fragmentation wavelengths?

Because they answer two different questions. $\lambda_{\mathrm{crit}} = 4\pi^2 G\Sigma/\kappa^2$ is the longest wavelength that a pressureless (cold) disk cannot stabilise — the rotation-set ceiling below which everything is unstable when there is no pressure. The most-unstable wavelength $\lambda_m = 2\pi^2 G\Sigma/\kappa^2$ is where the growth rate is largest once pressure is present and the disk is near $Q = 1$ ; it is the spacing you actually see between rings and clumps. Working through the algebra, the pressure-selected wavenumber at marginal stability is exactly twice the cold-disk critical wavenumber, so $\lambda_m = \lambda_{\mathrm{crit}}/2$ . Quoting $\lambda_{\mathrm{crit}}$ for the fragment spacing overestimates it twofold.

What physically is the epicyclic frequency κ?

It is the frequency at which a star oscillates radially about its circular orbit. Displace a star inward and it does not simply fall to the centre: angular-momentum conservation, expressed through the Coriolis force, deflects it, and it traces a small closed loop — an epicycle — returning to its starting radius $2\pi/\kappa$ later. The value of $\kappa$ is set entirely by the local shape of the rotation curve, $\kappa^2 = R\,\mathrm{d}\Omega^2/\mathrm{d}R + 4\Omega^2$ , and always lies between $\Omega$ (Keplerian) and $2\Omega$ (solid body). In the Toomre analysis $\kappa$ is the "rotation floor" that lifts the dispersion relation, stabilising long wavelengths that pressure alone could never protect.

Why do real disks sit near Q ≈ 1 rather than being very stable or very unstable?

Because $Q$ self-regulates. If a disk cools below $Q = 1$ — say by accreting cold gas — it goes unstable, and the resulting gravitational scattering and transient spirals stir the disk, raising $\sigma_R$ (or the turbulent $c_s$ ) until $Q$ climbs back toward one. A very hot disk ( $Q \gg 1$ ) makes no structure and simply relaxes. The stable fixed point of this thermostat is marginal stability, which is why galactic and protoplanetary disks cluster at $Q \sim 1$ – $2$ rather than at either extreme. The heating leaves a fossil record — the age–velocity-dispersion relation of disk stars.

Does Q > 1 mean a galaxy cannot have spiral arms?

No — quite the opposite. $Q > 1$ only rules out axisymmetric ring-like fragmentation. Non-axisymmetric shearing waves are still transiently amplified by swing amplification, which remains effective up to $Q \sim 2$ , and this is exactly the mechanism that generates flocculent and grand-design spiral arms. So the observed prevalence of spirals in disks that are demonstrably stable to fragmentation ( $Q \approx 1.5$ – $2$ ) is not a contradiction of the Toomre criterion; it is a reminder of its scope. Toomre himself made this point forcefully, arguing that spiral structure is recurrent and transient, continually regenerated by swing amplification rather than frozen into a long-lived wave.

When should I use the gas constant π versus the stellar constant 3.36?

Use $\pi$ when the component behaves as a fluid — interstellar gas or a protoplanetary disk — where a sound speed $c_s$ sets the small-scale restoring force. Use $3.36$ for a collisionless stellar component, where the relevant quantity is the radial dispersion $\sigma_R$ and the response is spread over many epicyclic orbits, captured by the Lin–Shu–Kalnajs reduction factor. The two differ because a coherent fluid responds in lockstep while a star swarm smears its response over phase space, making stars marginally easier to stabilise. If a disk contains both, use the two-component $Q_{\mathrm{eff}}$ — a cold gas layer can destabilise an otherwise-stable stellar disk.