Epicyclic Approximation
The orbit of an ordinary disk star — the Sun, say — is nearly, but not quite, a circle: it drifts round the Galaxy at a few hundred kilometres per second while executing a small in-and-out wobble in radius and a gentler bob up and down through the plane. Integrating the full nonlinear orbit is possible but opaque; the epicyclic approximation is the far more illuminating move. Expand the effective potential about the circular orbit of the same angular momentum — the guiding centre — and keep only terms linear in the displacement. The planar path then collapses into a uniform circular drift plus a tiny harmonic oscillation, exactly the deferent-plus-epicycle picture Ptolemy used to save the appearances, but now falling straight out of Newtonian dynamics in an axisymmetric potential. The payoff is a single new frequency. Radial excursions are simple harmonic at the epicyclic frequency κ, fixed entirely by the local slope of the rotation curve. Because angular momentum is conserved, the coupled azimuthal motion is locked to the radial swing, so in the rotating frame the star traces a small ellipse, elongated azimuthally by 2Ω/κ and circulated retrograde. Perpendicular motion decouples by midplane symmetry into a second, faster oscillator at frequency ν, whose square reads off the local mass density through Poisson's equation. Three numbers — Ω, κ, ν — and their observational avatars A and B organize nearly all of near-circular galactic dynamics.
- The epicyclic approximation linearizes a near-circular disk orbit about its guiding centre, splitting the motion into a uniform circular drift plus small radial and vertical harmonic oscillations.
- Radial displacements oscillate at the epicyclic frequency κ, set by the local rotation-curve slope through κ² = R dΩ²/dR + 4Ω²; for V ∝ R^β this is simply κ/Ω = √(2(1+β)).
- Kepler (β = −1/2) gives κ = Ω (closed ellipses), a flat curve (β = 0) gives κ = √2 Ω (non-closing rosettes), solid-body (β = 1) gives κ = 2Ω.
- In the guiding-centre frame the star traces a small retrograde ellipse whose azimuthal semi-axis exceeds the radial one by 2Ω/κ.
- The Oort constants encode the same physics for observers: Ω = A − B, dVc/dR = −(A + B), κ² = −4BΩ; near the Sun A ≈ 15.3, B ≈ −11.9 km/s/kpc give κ/Ω ≈ 1.32.
- Vertical motion decouples and oscillates at ν² = 4πGρ₀ + 2(A² − B²) ≈ 4πGρ₀ — the Oort limit that weighs the local disk.
- Key idea: expand the effective potential of an axisymmetric disk about a circular orbit — radial displacements then oscillate at κ² = R dΩ²/dR + 4Ω², the azimuthal motion is fixed by angular-momentum conservation, and the two combine into a small retrograde epicyclic ellipse drifting on the guiding-centre circle.
Expand the effective potential of an axisymmetric disk about a circular orbit: radial displacements oscillate at κ² = R dΩ²/dR + 4Ω², the azimuthal motion follows from angular-momentum conservation, and the two combine into a small retrograde epicyclic ellipse that drifts on the guiding-centre circle.
Where this lives today
The epicyclic machinery is the working language of Milky Way dynamics in the Gaia era. Gaia's astrometry of over a billion stars has measured the Oort constants to percent precision and mapped how A and B vary across the disk; the residuals from axisymmetry reveal the non-circular streaming of the bar and spiral arms. The velocity ellipsoid obeys the epicyclic prediction σ_φ/σ_R = κ/2Ω, which through the Jeans equations yields the asymmetric drift, the lag of an old, hot population behind the circular speed. Spiral structure lives on the same frequencies: the inner and outer Lindblad resonances, where Ω_p = Ω ± κ/m, mark where a spiral can launch or absorb angular momentum, and are now imaged as ridges in the Gaia velocity field. Most striking is the vertical "phase spiral" — the snail-shell pattern Gaia found in the z–v_z plane — the epicyclic vertical oscillator caught mid-relaxation after a passing satellite rang the disk, its winding rate set by the very ν(R) the Oort limit measures. And the Oort limit remains a live probe: comparing the dynamical ρ₀ with the census of stars, remnants, and gas bounds any smooth dark matter near the Sun to a fraction of ρ₀.
The physical picture
Think of a star as a bead sliding in a rotating bowl whose profile is the effective potential Φ_eff(R) = Φ(R) + L_z²/2R². A circular orbit sits precisely at the bottom, where gravity inward balances the centrifugal term outward. Nudge the bead in radius and it rolls back and forth about that minimum; the steeper the bowl curves, the stiffer the restoring force and the faster the oscillation. That frequency is κ = √(d²Φ_eff/dR²), and because the bowl's shape is set by how the circular speed varies with radius, κ is nothing more than a local diagnostic of the rotation curve. The star does not oscillate in radius alone: conservation of angular momentum means that moving inward it speeds up in azimuth and races ahead of its guiding centre, and moving outward it lags. The two motions weave into a closed loop in the rotating frame — the epicycle.
The retrograde sense is the surprising part. Because a flat rotation curve gives κ = √2 Ω, less than 2Ω, the star completes slightly less than one radial oscillation per half-turn around the Galaxy, so the orbit never quite closes: it precesses into the familiar rosette. The ratio κ/Ω is a clean fingerprint of the local rotation-curve slope, running from 1 in a Keplerian falloff, through √2 for a flat curve, to 2 for solid-body rotation. The second figure makes that dependence the punchline, with the Milky Way's solar neighbourhood pinned onto the curve.
The core equations
Every result below descends from one object: the effective potential a star of fixed L_z feels in the plane, bundling gravity with the centrifugal barrier so the radial problem becomes one-dimensional motion in a fixed well.
where: Φ(R,z) — axisymmetric gravitational potential; L_z — conserved specific angular momentum about the symmetry axis; R, z — cylindrical galactocentric radius and height. The centrifugal term L_z²/2R² rises steeply toward small R, digging a minimum wherever it balances the inward pull of Φ. The bottom of that well is the circular orbit, the guiding centre R_g. Setting the radial force to zero balances gravity against the centrifugal term and fixes its angular momentum.
where: R_g — guiding-centre radius, the minimum of Φ_eff; Ω(R_g) — circular angular velocity there, with Ω² = (1/R) ∂_R Φ. A star launched at R_g with the matching L_z stays on the circle forever; anything else oscillates about it. The curvature of the well at its minimum is the squared radial frequency; the second derivative, with L_z eliminated by the circular-orbit condition, collapses into the local Ω and its gradient.
where: κ — the radial (epicyclic) oscillation frequency; Ω(R) — circular angular velocity vs radius. κ² is never negative for a stable disk, and for a power-law curve reduces to the clean ratio plotted above. The same quantity has two more faces — a fractional correction to 2Ω, and a form in Oort's vorticity constant B, which is what an observer measures.
where: B — Oort constant of local vorticity, defined below. The form −4BΩ turns two locally measurable numbers into the radial frequency without integrating an orbit. Solving the coupled equations to first order gives the orbit in the rotating frame: a small ellipse, its coordinates 90° out of phase, elongated azimuthally.
where: x = R − R_g — radial displacement from the guiding centre; y — azimuthal displacement R_g(φ − Ωt); X — radial amplitude; ψ — phase constant. The minus sign and 90° phase lag make the circulation retrograde. Oort's constants repackage the rotation curve's value and slope, built to be read off nearby-star kinematics without knowing the distance to the Galactic centre.
where: A — Oort constant of azimuthal shear, in km s⁻¹ kpc⁻¹; B — Oort constant of vorticity; V_c = RΩ — circular rotation speed. A measures how neighbouring orbits slip past each other; B the local rotation of the flow. Adding and subtracting these recovers the angular velocity, the rotation-curve slope, and — the payoff — the epicyclic frequency, all from A and B alone.
where: Ω = V_c/R. A shearing flat disk has A = −B > 0; a rigidly rotating one has A = 0. Motion out of the plane decouples at leading order by midplane symmetry; its frequency squared is the vertical curvature of the potential, which Poisson's equation ties to the local density.
where: ν — vertical oscillation frequency; ρ₀ — total local midplane mass density; G — gravitational constant. Because A² − B² is small near the Sun, ν² ≈ 4πGρ₀: the vertical wobble weighs the disk. (Note the sign — the in-plane term enters as +2(A² − B²), so a Keplerian field with ρ₀ = 0 correctly gives ν² = Ω² > 0.) The vertical motion is simple harmonic, an independent pendulum superposed on the in-plane epicycle.
where: Z — vertical amplitude; φ_z — vertical phase. With (5) this is the complete leading-order orbit: a guiding-centre drift, a retrograde radial-azimuthal ellipse, and an independent vertical bob.
The goal is to reduce the exact coupled nonlinear equations of planar motion to a single decoupled harmonic oscillator in the radial displacement, and to show that its frequency squared — the quantity we name κ² — collapses from a messy second derivative of the potential into the compact local combination R dΩ²/dR + 4Ω². The centrifugal barrier is absorbed into an effective potential, the guiding centre is defined as its minimum, and the circular-orbit condition trades the awkward L_z²/R⁴ terms for Ω, leaving a result written in observable rotation-curve quantities.
Step 1. Write the exact planar equations of motion. In cylindrical coordinates the radial equation is R̈ − Rφ̇² = −∂Φ/∂R, and the azimuthal equation integrates once to the conserved specific angular momentum, d(R²φ̇)/dt = 0, i.e. L_z = R²φ̇. Why: axisymmetry makes φ ignorable, so L_z is a constant of the motion and the two-degree-of-freedom problem reduces to one radial degree of freedom.
Step 2. Eliminate φ̇ = L_z/R² from the radial equation. The centrifugal acceleration Rφ̇² = L_z²/R³ becomes a gradient, so R̈ = −∂Φ/∂R + L_z²/R³ = −∂Φ_eff/∂R with Φ_eff = Φ + L_z²/2R². Why: this is the whole trick — the rotating problem is now one-dimensional radial motion of a bead in a fixed potential well Φ_eff.
Step 3. Define the guiding centre and expand. The circular orbit sits at the minimum R_g where ∂_R Φ_eff = 0, equivalently ∂_R Φ = L_z²/R³ = RΩ², the balance of gravity and centrifugal force. Writing R = R_g + x and Taylor-expanding,
Why: the linear term in the force vanishes at the minimum by construction, so the leading restoring force is proportional to x — simple harmonic motion.
Step 4. Evaluate the curvature. Differentiating twice, ∂²Φ_eff/∂R² = ∂²Φ/∂R² + 3L_z²/R⁴. On the circular orbit L_z²/R⁴ = Ω², and ∂²Φ/∂R² = d(RΩ²)/dR = Ω² + R dΩ²/dR. Summing,
Why: replacing the raw potential derivatives by Ω and its gradient is what turns κ into a directly measurable property of the rotation curve rather than of the unknown Φ. *Verified: reproduces Binney and Tremaine eq. (3.79); V_c = const gives R dΩ²/dR = −2Ω² and κ = √2 Ω, the Kepler case Ω ∝ R⁻³/² gives −3Ω² and κ = Ω. In general κ/Ω = √(2(1+β)).*
Go deeper · Derivation B · 2 minThe retrograde epicyclic ellipse and its axis ratio
The goal is to show that the radial oscillation does not happen in isolation: conservation of angular momentum forces a companion azimuthal motion, and the two together trace a closed ellipse in the frame rotating with the guiding centre. We will find the ellipse is azimuthally elongated by exactly 2Ω/κ, the two coordinates ninety degrees out of phase, and the circulation retrograde relative to galactic rotation — the most counter-intuitive feature of near-circular orbits, and the reason disk orbits precess into rosettes rather than closing.
Step 1. Take the radial solution from the previous derivation. With ẍ = −κ²x the radial displacement is x(t) = X cos(κt + ψ), a pure harmonic swing of amplitude X about the guiding centre. Why: this is the input oscillator; the azimuthal motion is entirely slaved to it through angular-momentum conservation.
Step 2. Linearize the azimuthal drift. In the rotating frame let y = R_g(φ − Ωt) be the azimuthal displacement. Expanding L_z = R²φ̇ = const to first order in x and using L_z = R_g²Ω gives
Why: moving inward (x < 0) conserves L_z only by speeding the star ahead of the guiding centre, moving outward makes it lag — the factor 2Ω is the linearized response of φ̇ to a change in radius.
Step 3. Integrate the azimuthal equation. Substituting x = X cos(κt + ψ) and integrating,
Why: the integration divides by κ and introduces the sine, so y lags x by 90° and carries the amplitude (2Ω/κ)X — the ellipse's azimuthal semi-axis.
Step 4. Read off the geometry. Eliminating time, (x/X)² + (yκ/2ΩX)² = 1: an ellipse with radial semi-axis X and azimuthal semi-axis (2Ω/κ)X, so the axis ratio is 2Ω/κ ≥ 1. The signs of (5) mean that as the star reaches maximum outward x it is moving backward in y, so it circulates retrograde. Why: since κ < 2Ω for any realistic declining or flat rotation curve, the ellipse is always azimuthally elongated, and its non-integer κ/Ω is why the orbit fails to close. *Verified: axis ratio 2Ω/κ = √2 for a flat curve, 1 for solid-body; the retrograde sense and closed ellipse in the rotating frame match Binney and Tremaine §3.2.3.*
Go deeper · Derivation C · 2 minThe vertical frequency and the Oort limit
The goal is to establish that motion perpendicular to the disk decouples from the in-plane epicycle at leading order, that it is harmonic with a frequency ν set by the vertical curvature of the potential, and — crucially — that Poisson's equation converts that curvature into a direct measurement of the local mass density. This is the dynamical route to weighing the disk: watching how fast stars bob through the midplane infers ρ₀ without counting a single star, exactly how Oort first sensed unseen matter near the Sun. We track the in-plane Laplacian term carefully, since its sign is a classic trap.
Step 1. Establish the decoupling. The disk potential is symmetric about the midplane, Φ(R,z) = Φ(R,−z), so its expansion in z contains only even powers and no cross term ∝ xz at first order. Hence ∂²Φ/∂R∂z = 0 at z = 0 and the vertical equation is an independent oscillator z̈ = −ν²z with ν² = ∂²Φ/∂z²|₀. Why: midplane symmetry is what lets us treat the vertical motion separately — without it, radial and vertical oscillations would mix.
Step 2. Write Poisson's equation at the midplane. In cylindrical coordinates,
Why: the vertical curvature we want is one term of the Laplacian; the other term is fixed by the rotation curve, so Poisson's equation links ν² to ρ₀.
Step 3. Evaluate the in-plane term. Using ∂Φ/∂R = V_c²/R the radial part becomes (1/R) d(V_c²)/dR = 2Ω dV_c/dR. Inserting the Oort constants V_c/R = A − B and dV_c/dR = −(A + B),
Why: this is the trap — the in-plane term is negative, −2(A² − B²), so when it is moved to the other side of Poisson's equation it raises ν², not lowers it.
Step 4. Solve for the vertical frequency. Substituting into Poisson's equation, 4πGρ₀ = −2(A² − B²) + ν², so
Why: since A² − B² is small near the Sun the correction is a few percent and ν² ≈ 4πGρ₀ — the Oort limit — but the sign is fixed by the point-mass check, where ρ₀ = 0 must give ν² = Ω² > 0. *Verified: for a Kepler point mass A = 3Ω/4, B = −Ω/4 give 2(A² − B²) = Ω², and with ρ₀ = 0 the formula yields ν² = Ω², matching ∂²(−GM/√(R²+z²))/∂z²|₀ = GM/R³ = Ω². The plus sign is required; a minus sign gives the unphysical ν² = −Ω².*
Limiting cases
The single controlling ratio is κ/Ω = √(2(1+β)) for a power-law curve V ∝ R^β — so reading κ/Ω is reading the local rotation-curve slope, and vice versa.
| Case | Result | Meaning |
|---|---|---|
| Keplerian (point mass) | V ∝ R^(−1/2) ⇒ κ = Ω | Orbits close after one turn — the epicycle is the Keplerian ellipse; the guiding-centre picture reproduces the exact two-body ellipse. |
| Flat rotation curve | V = const ⇒ κ = √2 Ω | The realistic case for disks at intermediate radii; orbits are non-closing rosettes with azimuthal-to-radial axis ratio √2. |
| Solid-body rotation | V ∝ R ⇒ κ = 2Ω | The central regions and harmonic cores; the epicycle degenerates to a closed ellipse centred on the galaxy. |
| Harmonic core | Φ = ½Ω₀²R² ⇒ κ = 2Ω₀ | As solid body; every orbit is a centred ellipse closing after one radial period. |
| Vertical, thin disk | ν² ≈ 4πGρ₀ | The A² − B² correction vanishes for a flat curve (A = −B); the wobble reads the local density directly. |
Worked examples
1 · The Sun's epicyclic orbit from the Oort constants (fully worked)
Take the Gaia-era solar-neighbourhood values A = 15.3 km s⁻¹ kpc⁻¹ and B = −11.9 km s⁻¹ kpc⁻¹, with Galactocentric radius R₀ = 8.2 kpc. Find Ω₀, V_c, κ, the ratio κ/Ω, the epicyclic axis ratio, and the orbital and radial periods.
Step 1. Angular velocity and speed. Ω₀ = A − B = 15.3 − (−11.9) = 27.2 km s⁻¹ kpc⁻¹, and V_c = Ω₀R₀ = 27.2 × 8.2 = 223 km s⁻¹, squarely in the observed range.
Step 2. Epicyclic frequency. κ² = −4BΩ = −4(−11.9)(27.2) = 1295 (km s⁻¹ kpc⁻¹)², so κ = 36.0 km s⁻¹ kpc⁻¹ and κ/Ω = 1.32, just below the flat-curve √2 ≈ 1.41, telling us the local curve is slightly declining (β ≈ −0.13).
Step 3. Epicycle shape. The azimuthal-to-radial axis ratio is 2Ω/κ = 2(27.2)/36.0 = 1.51: a star's excursion in the direction of rotation is half again as large as its in-and-out radial swing, and it circulates retrograde.
Step 4. Periods. With 1 km s⁻¹ kpc⁻¹ = 1.023 Gyr⁻¹, the orbital period is T_Ω = 2π/Ω₀ = 226 Myr, the Sun's "galactic year", and the radial period T_κ = 2π/κ = 171 Myr. Because these are incommensurate the orbit never closes. Ω₀ = 27.2, κ = 36.0 km s⁻¹ kpc⁻¹, κ/Ω = 1.32, axis ratio 1.51, T_Ω ≈ 226 Myr, T_κ ≈ 171 Myr.
2 · Weighing the local disk with the Oort limit (fully worked)
The local total mass density is measured dynamically as ρ₀ ≈ 0.10 M_⊙ pc⁻³. Predict the vertical oscillation frequency ν, the vertical period, and the interval between successive midplane crossings, and check the small in-plane correction.
Step 1. Leading term. With G = 4.301 × 10⁻³ pc M_⊙⁻¹ (km s⁻¹)², 4πGρ₀ = 4π(4.301 × 10⁻³)(0.10) = 5404 (km s⁻¹ kpc⁻¹)², giving ν ≈ √5404 = 73.5 km s⁻¹ kpc⁻¹ to leading order.
Step 2. In-plane correction. With A = 15.3, B = −11.9, 2(A² − B²) = 2(234.1 − 141.6) = 185 (km s⁻¹ kpc⁻¹)². Then ν² = 5404 + 185 = 5589, so ν = 74.8 km s⁻¹ kpc⁻¹ — a 2% correction, confirming ν² ≈ 4πGρ₀ is an excellent approximation.
Step 3. Vertical period. With 1 km s⁻¹ kpc⁻¹ = 1.023 Gyr⁻¹, ν = 76.5 Gyr⁻¹ and T_z = 2π/ν = 82 Myr. The star crosses the midplane twice per period, so crossings recur about every 41 Myr.
Step 4. Sanity check. The vertical period (≈82 Myr) is roughly half the epicyclic period (≈171 Myr) and a third of the orbital period (≈226 Myr): the Sun bobs vertically several times per orbit, the ordering that keeps the thin disk thin while it precesses. ν ≈ 74.8 km s⁻¹ kpc⁻¹, T_z ≈ 82 Myr, midplane crossings every ≈41 Myr.
3 · Oort's double-sine law for a nearby star
A star at distance d and Galactic longitude ℓ, on a circular orbit near the Sun's, has line-of-sight velocity v_r = (Ω(R) − Ω₀)R₀ sin ℓ. Linearizing with R − R₀ ≈ −d cos ℓ and A = −½R₀(dΩ/dR)₀ collapses this to v_r = A d sin 2ℓ — a velocity that peaks toward ℓ = 45°, 135°, 225°, 315° and vanishes toward the centre and anticentre, which is Oort's 1927 detection of Galactic rotation. At ℓ = 45° and d = 0.5 kpc, v_r = A d = (15.3)(0.5) = <strong>7.65 km s⁻¹</strong>.
Physics of the portrait
The epicyclic framework has two parents: Bertil Lindblad supplied the dynamics — the guiding centre, the radial oscillation, the resonances — while Jan Oort supplied the observational handle that turned it into a measuring instrument for the Galaxy. Oort's constants let a table of nearby-star velocities be read as the local rotation curve, and his vertical analysis first weighed the disk.
Origins · 1 minJan Oort and how this was discovered
Jan Hendrik Oort was the Dutch astronomer who, more than anyone, turned the Milky Way from a smudge of stars into a rotating dynamical system with measurable numbers. In 1927, building on Bertil Lindblad's just-published theory of galactic rotation, Oort showed that a differentially rotating disk makes nearby stars display a double-sine pattern in their radial velocities and proper motions as a function of Galactic longitude — and he found it in the data, distilling the local kinematics into two constants, A (shear) and B (vorticity). They pinned the Sun's rotation speed and the rotation-curve slope without any knowledge of the distance to the Galactic centre, and slot straight into the epicyclic frequency through κ² = −4BΩ. In 1932 he pushed the same reasoning perpendicular to the plane: measuring how the density and vertical velocities of disk stars fall off with height, he inferred the total gravitating density near the Sun — the "Oort limit" — and noted it exceeded the visible mass, an early hint of unseen matter. Oort went on to map the Galaxy's spiral structure in the 21-cm radio line, to explain the reservoir of long-period comets now called the Oort cloud, and to lead post-war European astronomy for a generation. The epicyclic approximation is his measuring stick made theory: three frequencies, two constants, one nearly circular orbit.
Going further
Texts. Binney and Tremaine, Galactic Dynamics, 2nd ed., §3.2.3 (the epicycle approximation) and §4.5 · Carroll and Ostlie, An Introduction to Modern Astrophysics, 2nd ed., §24.3 (differential rotation and the Oort constants) · Mo, van den Bosch and White, Galaxy Formation and Evolution, §11.1 (disk dynamics) · Oort, J. H. 1927, Bulletin of the Astronomical Institutes of the Netherlands, 3, 275.
Provenance. The equations follow Binney and Tremaine §3.2.3 and §4.5 and Carroll and Ostlie §24.3. A verifier should recompute: (1) κ² = 4Ω² + R dΩ²/dR from the second derivative of Φ_eff, and the general κ/Ω = √(2(1+β)) for V ∝ R^β; (2) the limits κ = Ω (Kepler), √2 Ω (flat), 2Ω (solid body); (3) the Oort relations Ω = A − B, dV_c/dR = −(A + B), κ² = −4BΩ; (4) the corrected vertical relation ν² = 4πGρ₀ + 2(A² − B²), checked against the Kepler point mass ρ₀ = 0 ⇒ ν² = Ω²; (5) the solar-neighbourhood numbers Ω₀ = 27.2, κ = 36.0, κ/Ω = 1.32, T_Ω ≈ 226 Myr, T_κ ≈ 171 Myr, ν ≈ 74.8 km s⁻¹ kpc⁻¹, T_z ≈ 82 Myr for A = 15.3, B = −11.9, R₀ = 8.2 kpc, ρ₀ = 0.10 M_⊙ pc⁻³. The Oort constants and local density are standard Gaia-era values.
Prerequisites. Orbits in Axisymmetric and Triaxial Potentials · Galactic Potential Models · Orbits in Spherical Potentials. Leads to. Action-Angle Variables · Jeans Equations · Distribution Function and Phase Space.
Ask the tutor: Ask me why the Sun's orbit is a retrograde rosette, or to derive the Oort limit and weigh the local disk from scratch.
Why is it called the "epicyclic" approximation?
The name is a deliberate nod to Ptolemy. In the Almagest, planetary motion was a large circle (the deferent) carrying a small circle (the epicycle) — a big drift plus a small loop. The approximation reproduces exactly that structure, but for a disk star and derived from Newtonian gravity rather than assumed: the guiding centre is the deferent, drifting at rate Ω, and the small radial-azimuthal ellipse is the epicycle. The difference is that here the loop is a genuine consequence of the potential and angular-momentum conservation, and its frequency κ carries real information about the rotation curve.
Why is the epicyclic motion retrograde?
Because angular momentum is conserved. When the star wanders inward from its guiding centre, R shrinks, and to keep L_z = R²φ̇ fixed its angular rate φ̇ must increase — it races ahead of the guiding centre. Half an oscillation later it is outward, R is larger, and it falls behind. Tracking this in the frame rotating with the guiding centre, the star runs forward at closest approach and backward at farthest, which traces the ellipse in the sense opposite to the Galaxy's rotation. So the retrograde sense is not an extra assumption; it is forced by the same conservation law that couples the radial and azimuthal motions in the first place.
Why don't disk orbits close, and what is a rosette?
An orbit closes only if κ/Ω is rational — if the star completes a whole number of radial oscillations in a whole number of azimuthal revolutions. For a Keplerian potential κ/Ω = 1 and the orbit closes into an ellipse; for a solid-body core κ/Ω = 2 and it also closes. But for the flat and gently declining curves of real disks, κ/Ω ≈ 1.3–1.4, an irrational ratio, so each radial cycle brings the star back to perigalacticon at a slightly different azimuth. Over many cycles the perigalacticon precesses steadily and the orbit fills an annular band with a flower-petal pattern — the rosette.
Why does vertical motion oscillate faster than the radial epicycle?
Because the disk is thin, the vertical restoring force is stiff. Near the midplane the density is high and concentrated, so the potential curves sharply in z, giving a large ν² = 4πGρ₀ + 2(A² − B²). Radially, by contrast, the effective potential is a shallow well whose curvature κ² reflects the gentle large-scale rotation curve. For the Sun ν ≈ 75 versus κ ≈ 36 km s⁻¹ kpc⁻¹, so the vertical period (≈82 Myr) is roughly half the epicyclic period (≈171 Myr): the star bobs through the plane several times during each radial swing, which is what keeps the thin disk thin.
What exactly do the Oort constants A and B measure?
They are the local value and slope of the rotation curve, repackaged for an observer who cannot easily measure the distance to the Galactic centre. A = ½(V_c/R − dV_c/dR) is the azimuthal shear — how fast neighbouring circular orbits slide past one another — and shows up as a sin 2ℓ modulation of nearby stars' radial velocities. B = −½(V_c/R + dV_c/dR) is the vorticity, the local rotation of the velocity field, measured through proper motions. Their combinations recover everything: Ω = A − B, the rotation-curve slope −(A + B), and the epicyclic frequency κ² = −4BΩ. A rigidly rotating disk has A = 0; a purely shearing flat disk has A = −B.
When does the epicyclic approximation break down?
Whenever the excursions stop being small. The scheme is a first-order expansion about a circular orbit, so it fails for large radial amplitude (halo stars on plunging orbits, or stars near the Galactic centre), for large vertical excursions where the harmonic ν²z² potential gives way to the true thick-disk profile, and near resonances where higher-order terms couple the oscillators and drive slow growth. It also assumes an axisymmetric, time-independent potential, so it cannot by itself capture the bar, spiral arms, or a passing satellite — although, tellingly, those perturbations are most cleanly described as forcings of the epicyclic oscillators, which is why it remains the right starting point even when it breaks.
How do the epicyclic amplitudes relate to actions?
Very simply, which is why the epicyclic limit is the gateway to action-angle variables. Each oscillator is one-dimensional and harmonic, and for a harmonic oscillator the action is the energy divided by the frequency: J_R = E_R/κ = ½κX² and J_z = E_z/ν = ½νZ². The amplitudes X and Z of the two epicyclic swings are therefore direct measures of the two actions, which are adiabatic invariants conserved under slow changes in the potential. This makes actions the natural labels for disk orbits and the coordinates in which disk distribution functions are most cleanly written — the subject the next articles take up.