Poinsot Construction

By 1834 the free rigid body had been solved for seventy years — Euler's equations integrate, in the asymmetric case, into elliptic functions — and Louis Poinsot was openly unsatisfied. "In all these solutions," he complained of the analytic mechanics of Euler and Lagrange, "we see only calculations, without any clear picture of the motion." A tumbling body was a triple of time series $\omega_i(t)$ ; nobody could watch it. His Théorie nouvelle de la rotation des corps replaced the time series with a single mechanical image so vivid it has never been improved: attach to the body its inertia ellipsoid, and the entire torque-free motion is that ellipsoid rolling, without slipping, on an invisible fixed plane. Everything is forced by the two conservation laws. Conserved kinetic energy pins the tip of the (scaled) angular velocity to the ellipsoid's surface; conserved angular momentum makes the tangent plane at that point perpendicular to the fixed vector $\mathbf{L}$ and holds it at constant distance — a plane nailed to space, the invariable plane. And because the contact point lies on the instantaneous rotation axis, it is momentarily at rest: rolling without slipping, derived rather than assumed. The construction turns hard questions into things you can see. Why are spins about the extreme axes stable but the intermediate axis catastrophic? Look at the curves the contact traces. Why does a symmetric top precess at a fixed rate? The ellipsoid becomes a cone rolling on a cone. This article builds the construction from scratch, proves the rolling theorem, and runs the picture on real numbers — a tumbling satellite and the wobble of the Earth itself.

Key takeaways

For torque-free rigid-body motion the angular momentum $\mathbf{L}$ and kinetic energy $T$ are both conserved, which constrains the angular velocity $\boldsymbol{\omega}$ to lie on the intersection of two surfaces in $\boldsymbol{\omega}$ -space.
Poinsot's construction reinterprets this as a purely geometric picture: the inertia ellipsoid, rigidly attached to the body, rolls without slipping on a fixed plane — the invariable plane — perpendicular to $\mathbf{L}$ at constant distance $d=\sqrt{2T}/L$ .
The contact point is where the scaled spin vector $\boldsymbol{\rho}=\boldsymbol{\omega}/\sqrt{2T}$ pierces the ellipsoid; it lies on the instantaneous rotation axis, so its velocity is zero — exactly the rolling-without-slipping condition.
The contact traces the polhode on the ellipsoid (closed, fixed in the body) and the herpolhode on the invariable plane (generally never closing) — the detailed geometry of both curves is its own article.
Polhodes ring the axes of largest and smallest moment (stable spins) and cross in a separatrix at the intermediate axis — the geometric face of the intermediate-axis theorem.
For a symmetric top the ellipsoid degenerates to a cone rolling on a cone: circular polhode, circular herpolhode, and steady precession of $\boldsymbol{\omega}$ about $\mathbf{L}$ at rate $L/I_\perp$ .
Key idea: Torque-free motion = inertia ellipsoid rolling without slipping on the fixed invariable plane (perpendicular to $\mathbf{L}$ , at fixed distance set by $T$ ); contact point = instantaneous rotation axis.

Where this lives today

The construction earns its keep wherever a spin state must be understood rather than merely propagated. The most consequential recent case is Gravity Probe B: its science gyroscopes — the roundest objects ever manufactured — still had part-per-million inertia asymmetries, and their slow polhode wander modulated the readout in ways that had to be modeled with exactly this machinery before the relativity signal could be extracted. Space-situational-awareness groups fit the optical light-curves of defunct satellites and spent stages with torque-free Poinsot solutions to recover their spin states, then track the slow polhode migration toward the major axis as internal damping drains energy — the dissipative version of the rolling picture. Attitude engineers still speak its language daily: "polhode motion" is the standard term for the nutational component of any spacecraft's spin, and detumbling strategies are planned on the energy–momentum ( $\ell$ -dial) diagram of Worked Example 1. In the classroom and the simulator the construction has become the canonical visualization layer over Euler's equations: every serious rigid-body dynamics package renders the energy ellipsoid, momentum sphere, and rolling contact precisely because three coupled nonlinear ODEs communicate nothing and one rolling surface communicates everything.

The physical picture

Take the tumbling body — wrench, satellite, asteroid — and glue to it an imaginary glassy surface: its inertia ellipsoid, the quadric whose radius in every direction encodes the moment of inertia about that direction (built and read in Inertia Ellipsoid; its semi-axes run along the principal axes with lengths $1/\sqrt{I_i}$ ). The ellipsoid is a body-fixed object; it tumbles exactly as the body does. Now the claim. Somewhere in space there is one fixed, invisible tabletop. However wildly the body appears to tumble, its glassy ellipsoid is at every instant touching that table, pressed against it at exactly one point, rolling on it without ever sliding. The point of contact is wherever the instantaneous spin axis currently pierces the ellipsoid — and since a point on the rotation axis is momentarily at rest, the contact never scrubs. The tumble is not wild at all: it is the most orderly motion imaginable, a rolling.

Figure

Poinsot's construction: the body-fixed inertia ellipsoid rolls without slipping on the invariable plane perpendicular to L

Each conservation law owns one half of the picture. Energy conservation says the contact point sits at a fixed height: the center of the ellipsoid hovers at constant distance $d=\sqrt{2T}/L$ above the plane, so the table never has to move up or down. Angular-momentum conservation says the tangent plane at the contact is always perpendicular to the same space-fixed vector $\mathbf{L}$ — the table never tilts. A surface constrained to touch a fixed plane, with a contact point that cannot slide, has exactly one thing left to do: roll. And the two scratches the rolling leaves — the polhode worn into the ellipsoid, the herpolhode scuffed onto the table — are the complete record of the motion, taken up curve by curve in Polhode and Herpolhode Curves. The picture also settles stability at a glance. The family of possible polhodes on any asymmetric ellipsoid consists of small rings around the long-axis and short-axis vertices and a crossing separatrix through the intermediate-axis vertex — center, center, saddle. That is the intermediate-axis theorem, read off a surface (the dynamical growth rate that goes with the saddle is computed in Intermediate-Axis (Tennis-Racket) Theorem).

Figure

The family of polhode curves on the inertia ellipsoid: rings around the extreme-axis vertices, a separatrix through the intermediate vertex

Building the construction

The whole edifice rests on two conserved quantities and one scaling. With zero torque the angular momentum is constant as a vector in the inertial frame — magnitude and direction — which is what will nail the plane to space:

\mathbf{L}=\mathsf{I}\,\boldsymbol{\omega}=\text{const},\qquad \dot{\mathbf{L}}=0 \qquad (1)

where: $\mathsf{I}$ — the inertia tensor, diagonal $\mathrm{diag}(I_1,I_2,I_3)$ in the body's principal frame; $\boldsymbol{\omega}$ — the angular velocity. In the body frame the components of $\mathbf{L}$ wander (that wandering is Euler's equations at work — see Euler's Equations of Motion), but the space-frame vector is frozen. The kinetic energy is the second invariant, and it comes with an identity that will do quiet work throughout: $T$ can be written as half the projection of $\mathbf{L}$ onto $\boldsymbol{\omega}$ ,

T=\tfrac{1}{2}\,\boldsymbol{\omega}^{\mathsf T}\mathsf{I}\,\boldsymbol{\omega}=\tfrac12\,\boldsymbol{\omega}\cdot\mathbf{L}=\text{const} \qquad (2)

where: the constancy follows in one line from the vanishing rotational power, $\dot T=\boldsymbol{\omega}\cdot\boldsymbol{\tau}=0$ . Geometrically, equation (2) says the component of $\boldsymbol{\omega}$ along the fixed direction $\hat{\mathbf{L}}$ is pinned at $\omega_\parallel=2T/L$ — the first hint that a fixed plane is hiding in the problem. Now the scaling that hands the problem to geometry. Measure the spin vector in units of $\sqrt{2T}$ :

\boldsymbol{\rho}=\frac{\boldsymbol{\omega}}{\sqrt{2T}}\qquad\Longrightarrow\qquad \boldsymbol{\rho}^{\mathsf T}\mathsf{I}\,\boldsymbol{\rho}=\frac{2T}{2T}=1 \qquad (3)

where: $\boldsymbol{\rho}$ — the scaled angular velocity. Equation (3) is exactly the defining equation of the inertia ellipsoid: at every instant of the motion, the tip of $\boldsymbol{\rho}$ lies on the body's ellipsoid. The tumbling has been traded for a moving point of contact on a rigid quadric — and the next derivation shows that contact is a rolling one.

Derivation AThe rolling theorem: a fixed plane, touched without slipping

We are proving the three statements that together constitute Poinsot's theorem: (i) the tangent plane to the inertia ellipsoid at the point $\boldsymbol{\rho}$ is perpendicular to the space-fixed $\mathbf{L}$ ; (ii) that plane sits at a constant distance from the body's center, so it is one unmoving plane for the whole motion; (iii) the material contact point has zero velocity, so the ellipsoid rolls without slipping. The strategy is elementary vector calculus on the quadric $F(\boldsymbol{\rho})=\boldsymbol{\rho}^{\mathsf T}\mathsf{I}\boldsymbol{\rho}=1$ — its gradient gives the normal, the two conservation laws give the fixedness — and the point to watch is which frame each statement lives in: the ellipsoid is body-fixed, but both defining properties of the plane are space-frame facts, which is precisely why the construction couples the two frames.

Step 1. Compute the outward normal of the ellipsoid at the contact. The normal to a level surface is its gradient:

\nabla_{\boldsymbol{\rho}}\big(\boldsymbol{\rho}^{\mathsf T}\mathsf{I}\,\boldsymbol{\rho}\big)=2\,\mathsf{I}\boldsymbol{\rho}=\frac{2\,\mathsf{I}\boldsymbol{\omega}}{\sqrt{2T}}=\frac{2}{\sqrt{2T}}\,\mathbf{L}

The tangent plane at $\boldsymbol{\rho}$ is therefore perpendicular to $\mathbf{L}$ — a direction that never changes. Why this is the crux: the normal of a quadric at $\boldsymbol{\rho}$ is $\mathsf{I}\boldsymbol{\rho}$ , and multiplying $\boldsymbol{\omega}$ by the inertia tensor is exactly how angular momentum is built. The ellipsoid was constructed so that its geometry would convert $\boldsymbol{\omega}$ -facts into $\mathbf{L}$ -facts.

Step 2. Compute the distance from the ellipsoid's center to that tangent plane — the projection of the contact position onto the plane's unit normal:

d=\boldsymbol{\rho}\cdot\hat{\mathbf{L}}=\frac{\boldsymbol{\omega}\cdot\mathbf{L}}{\sqrt{2T}\,L}=\frac{2T}{\sqrt{2T}\,L}=\frac{\sqrt{2T}}{L}

Both $T$ and $L$ are constants of the motion, so $d$ is one number for the whole flight. Why constancy matters: at each instant we have a tangent plane perpendicular to $\mathbf{L}$ ; constancy of $d$ upgrades the family to a single plane, fixed in space for all time — Poinsot named it the invariable plane.

Step 3. Show the contact point never slides. The velocity of the material point of the (extended) body sitting at position $\boldsymbol{\rho}$ from the center is $\boldsymbol{\omega}\times\boldsymbol{\rho}$ :

\mathbf{v}_{\text{contact}}=\boldsymbol{\omega}\times\boldsymbol{\rho}=\frac{1}{\sqrt{2T}}\;\boldsymbol{\omega}\times\boldsymbol{\omega}=\mathbf{0}

Why zero: the contact point was defined as the piercing point of the instantaneous rotation axis, and points on the rotation axis are the ones momentarily at rest — rolling without slipping is not an added assumption but a restatement of what "instantaneous axis" means.

Step 4. Assemble. The body-fixed ellipsoid maintains tangency with one space-fixed plane (Steps 1–2) at a contact point that is instantaneously at rest (Step 3). That is the complete kinematic characterization of rolling without slipping: the free rigid body is its inertia ellipsoid rolling on the invariable plane, with the center gliding parallel to the plane at constant height $d$ . Why the range of $d$ makes sense: dimensionally, $d=\sqrt{2T}/L$ carries the units of $\rho$ — inverse square root of moment of inertia — matching the ellipsoid's semi-axes $1/\sqrt{I_i}$ ; and $d$ correctly ranges between $1/\sqrt{I_3}$ and $1/\sqrt{I_1}$ , the shortest and longest semi-axes, as the inequality of the next section requires.

Which curve you are on: the single dial l

Rolling needs a track, and the track is selected by the launch conditions through one number. Squaring the angular momentum in principal components and comparing with the energy shows that the motion fixes the ratio

\ell\equiv\frac{L^2}{2T}=\frac{\sum_i I_i^2\omega_i^2}{\sum_i I_i\omega_i^2},\qquad I_1\le\ell\le I_3 \qquad (4)

where: $\ell$ — an "effective moment of inertia" of the motion, with units of $I$ . The bounds follow because $\ell$ is a weighted average of the $I_i$ with non-negative weights $I_i\omega_i^2$ : no free motion can be more momentum-heavy than pure spin about $I_3$ ( $\ell=I_3$ , $d$ at its minimum) or more energy-heavy than pure spin about $I_1$ ( $\ell=I_1$ , $d$ at its maximum). Every value in between selects one polhode track on the ellipsoid, and the watershed value $\ell=I_2$ selects the separatrix.

Go deeper · Derivation B · 2 minOne dial classifies every polhode

Derivation BOne dial classifies every polhode

The goal is to show that the value of $\ell=L^2/2T$ relative to the middle moment $I_2$ determines the global character of the rolling track: rings around $\hat{\mathbf{e}}_1$ for small $\ell$ , rings around $\hat{\mathbf{e}}_3$ for large $\ell$ , and the separatrix at exactly $\ell=I_2$ . The strategy is to combine the two conserved quadrics into a single homogeneous equation and read off which axis the resulting cone encloses. This complements the local saddle-versus-center analysis in Intermediate-Axis (Tennis-Racket) Theorem — same physics, but organized globally by launch conditions rather than locally by axis.

Step 1. Subtract $\ell$ times the energy relation from the momentum relation, term by term:

\sum_i I_i\,(I_i-\ell)\,\omega_i^2=0

This single equation, together with the energy normalization, defines the polhode: it is the intersection of the energy ellipsoid with a cone through the origin whose shape is controlled by $\ell$ . Why a cone: the equation is homogeneous of degree two — if $\boldsymbol{\omega}$ satisfies it so does every multiple — so its solution set is a cone from the origin, and the polhode is the ring where that cone punches through the ellipsoid.

Step 2. Read the signs of the three coefficients $I_i(I_i-\ell)$ . For $I_1<\ell<I_2$ : the first coefficient is negative, the other two positive — the cone opens around the $\hat{\mathbf{e}}_1$ axis, and the polhode is a pair of small closed rings encircling the two $\hat{\mathbf{e}}_1$ vertices. For $I_2<\ell<I_3$ : only the third coefficient is negative and the rings encircle the $\hat{\mathbf{e}}_3$ vertices instead. Why "encircling" follows from the signs: with one negative coefficient the cone's axis is the corresponding principal axis; slicing a one-axis cone with the surrounding ellipsoid can only produce loops around that axis.

Step 3. Set $\ell=I_2$ . The middle coefficient dies and the cone degenerates into a pair of planes,

I_1(I_1-I_2)\,\omega_1^2+I_3(I_3-I_2)\,\omega_3^2=0\;\Longrightarrow\;\omega_3=\pm\,\omega_1\sqrt{\frac{I_1(I_2-I_1)}{I_3(I_3-I_2)}}

Two planes through the intermediate axis cut the ellipsoid in two ellipses that intersect at the $\hat{\mathbf{e}}_2$ vertices — the separatrix crossing, the saddle of the whole phase portrait. Why the watershed is exactly $I_2$ : the sign pattern of the three coefficients can only change when $\ell$ passes a principal moment, and passing $I_1$ or $I_3$ exits the allowed range (4); $I_2$ is the only interior crossing.

Step 4. Conclude stability by inspection: launches with $\ell$ near the ends of its range roll on tiny rings (bounded wobble about an extreme axis — stable), launches with $\ell$ near $I_2$ roll on tracks that pass close to the separatrix and carry the contact point across the whole ellipsoid (the flip — unstable). The quantitative growth rate at the saddle belongs to the linearized analysis in the intermediate-axis article; the geometry here shows its inevitability. Two further checks on the sign analysis: the separatrix slope reproduces the linearized eigendirections of Euler's equations at the saddle; and for a symmetric body ( $I_1=I_2$ ) the watershed collapses against the edge of the allowed range, correctly leaving only stable circular tracks.

The symmetric top: cones rolling on cones

When two moments coincide ( $I_1=I_2=I_\perp$ ) the ellipsoid becomes a surface of revolution and the rolling picture simplifies into its most famous special case. The polhode becomes a circle about the symmetry axis; the vector $\boldsymbol{\omega}$ , at fixed angle to the symmetry axis, sweeps a body cone, while in space it sweeps a space cone about the fixed $\mathbf{L}$ ; the motion is the body cone rolling without slipping on the space cone. Their common line of tangency is the instantaneous axis. The space-frame precession rate of $\boldsymbol{\omega}$ (and of the symmetry axis) about $\mathbf{L}$ comes straight from the construction:

\Omega_{\text{prec}}=\frac{L}{I_\perp} \qquad (5)

where: $I_\perp$ — the doubly degenerate transverse moment. The companion body-frame rate, $\Omega_b=\omega_3(I_3-I_\perp)/I_\perp$ , is the speed at which the contact circles the polhode. The full development of the cones — including the oblate/prolate distinction of whether the body cone rolls outside or inside the space cone — is the business of Free Symmetric Top and Body-Cone/Space-Cone Precession; here the point is simply that it is the degenerate Poinsot picture, with the invariable plane's tangency circle playing the role of the rolling track.

Limiting cases

Case	Result	Meaning
spherical top, $I_1=I_2=I_3$	ellipsoid is a sphere; $\boldsymbol{\omega}\parallel\mathbf{L}$ always	contact point never moves: polhode and herpolhode are points, spin is steady about a fixed axis
symmetric top, $I_1=I_2\ne I_3$	ellipsoid of revolution; circular tracks	body cone rolls on space cone; precession at $L/I_\perp$ — eq. (5)
asymmetric top, $I_1<I_2<I_3$	generic rolling	closed non-circular polhodes; herpolhode generally never closes
pure principal-axis spin	$\ell=I_i$ exactly, $d=1/\sqrt{I_i}$	contact sits on a vertex; the "curves" are points; steady rotation
$\ell\to I_2$ (separatrix launch)	crossing tracks through the $\hat{\mathbf{e}}_2$ vertex	the flip trajectory of the intermediate-axis instability
$\ell\to I_1$ (max $d$ ) / $\ell\to I_3$ (min $d$ )	plane tangent at the longest / shortest semi-axis	energy-heavy vs momentum-heavy extremes of the allowed range (4)

Common trapThe ellipsoid is rolling on a real surface — no: nothing touches anything

The construction seduces you into imagining contact mechanics — friction gripping the plane, normal forces, the possibility of slipping if the spin is "too fast." None of that exists. The inertia ellipsoid is an abstract body-fixed quadric, the invariable plane is a bookkeeping surface defined by the conserved $\mathbf{L}$ and $T$ , and "rolling without slipping" is a derived kinematic identity ( $\boldsymbol{\omega}\times\boldsymbol{\rho}=0$ ), not a constraint that some force enforces. Nothing can ever slip, at any spin rate, because the statement is an identity. Two further frame traps: the polhode is drawn on the body-fixed ellipsoid (it tumbles with the body — you could paint it on), while the herpolhode is drawn on the space-fixed plane; and the rolling is slippless but not rateless — the contact speeds up and slows down along the polhode, moving fastest along the separatrix portions, which is why the herpolhode, unlike the polhode, generally never closes.

Worked examples

1 · Reading a satellite's tumble off the construction (confidence builder)

A small satellite has principal moments $I_1=20$ , $I_2=30$ , $I_3=40\ \mathrm{kg\,m^2}$ and is released with $\boldsymbol{\omega}=(0.20,\,0.10,\,1.00)\ \mathrm{rad/s}$ in the principal frame. Locate its motion in the Poinsot picture: energy, momentum, the dial $\ell$ , the height $d$ , and which axis its polhode encircles.

Step 1. Kinetic energy from equation (2): $2T=I_1\omega_1^2+I_2\omega_2^2+I_3\omega_3^2=20(0.04)+30(0.01)+40(1.00)=0.80+0.30+40.0=41.1$ , so $T=20.6\ \mathrm{J}$ .

Step 2. Angular momentum squared: $L^2=I_1^2\omega_1^2+I_2^2\omega_2^2+I_3^2\omega_3^2=400(0.04)+900(0.01)+1600(1.00)=16+9+1600=1625$ , so $L=40.3\ \mathrm{kg\,m^2/s}$ .

Step 3. The dial: $\ell=L^2/2T=1625/41.1=39.5\ \mathrm{kg\,m^2}$ . Check the allowed range (4): $20\le39.5\le40$ ✓. Since $\ell>I_2=30$ — indeed $\ell$ is within $0.5\ \mathrm{kg\,m^2}$ of $I_3$ — the polhode is a small ring around the $\hat{\mathbf{e}}_3$ vertex: this is a gentle wobble about the major axis, nowhere near the separatrix. The satellite is safe.

Step 4. The invariable-plane height: $d=\sqrt{2T}/L=\sqrt{41.1}/40.3=6.41/40.3=0.159\ (\mathrm{kg\,m^2})^{-1/2}$ , to be compared with the semi-axes $1/\sqrt{I_3}=0.158$ and $1/\sqrt{I_1}=0.224$ : the plane grazes the ellipsoid just above its shortest vertex, exactly as a near-major-axis spin should. The tumble is a tight polhode ring of the rolling ellipsoid, pressed nearly vertex-down on its invariable plane.

2 · The Earth as a rolling ellipsoid: the Euler wobble (exam level)

Treat the Earth as a free symmetric top with dynamical flattening $(I_3-I_\perp)/I_\perp=1/305$ and spin $\omega_3=7.292\times10^{-5}\ \mathrm{rad/s}$ . The spin axis is observed to sit about $0.2''$ (arcseconds) from the figure axis. (a) Find the period with which the contact point circles the polhode (the body-frame wobble period). (b) Find the radius of the polhode circle drawn on the Earth's surface at the pole. (c) The observed period is about 433 days — what does the discrepancy teach?

Step 1. (a) For the symmetric top the contact circles the polhode at the body-frame rate $\Omega_b=\dfrac{I_3-I_\perp}{I_\perp}\,\omega_3=\dfrac{7.292\times10^{-5}}{305}=2.39\times10^{-7}\ \mathrm{rad/s}$ .

Step 2. Period $T_b=2\pi/\Omega_b=2\pi/2.39\times10^{-7}=2.63\times10^{7}\ \mathrm{s}=\boxed{304\ \text{days}}$ — Euler's celebrated ten-month free-wobble prediction, straight from the rolling-cone rate.

Step 3. (b) The polhode circle's angular radius is the spin-figure axis offset $\theta=0.2''=0.2\times4.848\times10^{-6}=9.70\times10^{-7}\ \mathrm{rad}$ . Drawn on the surface at the pole ( $R=6.37\times10^{6}\ \mathrm{m}$ ), its radius is $r=R\,\theta=6.37\times10^{6}\times9.70\times10^{-7}=\boxed{6.2\ \mathrm{m}}$ . The grandest Poinsot construction available is a circle the size of a living room, traced by the pole among the buildings of an Antarctic station.

Step 4. (c) The observed Chandler wobble period, $\approx433$ days, is 40% longer than the rigid-body 304 days. The lengthening is the fingerprint of non-rigidity: the elastic yielding of the mantle and the response of the oceans effectively reduce $(I_3-I_\perp)/I_\perp$ for the wobble mode. The Poinsot construction supplies the rigid baseline; the deviation from it is data about the Earth's interior — an early and still-running example of using rigid-body theory as a null hypothesis.

Origins · 1 minFelix Klein and how this was discovered

Felix Klein

1849–1925 · Göttingen

Klein is remembered for the Erlangen program — the 1872 manifesto that reorganized all of geometry around groups of transformations — but he also co-wrote the longest love letter geometry ever sent to a spinning body. With Arnold Sommerfeld he produced Über die Theorie des Kreisels (1897–1910), four volumes and nearly a thousand pages on the top, in which Poinsot's rolling ellipsoid is not a pedagogical aside but a load-bearing method: Klein insisted that the "kinematical content" of a mechanical law should be visible before it is computed. His own contribution ran deeper than exposition — the Cayley–Klein parameters he championed represent rotations by pairs of complex numbers, anticipating by decades the quaternion and spinor machinery of modern attitude control and quantum mechanics. Klein's larger campaign, at Göttingen and through his reform of mathematics teaching, was that intuition and rigor are allies, not rivals; the Kreisel volumes are that philosophy applied to the same problem Poinsot had rescued from pure analysis sixty years earlier, and the modern habit of drawing phase portraits before integrating anything descends from exactly this tradition.

History

1758

— Euler derives the body-frame equations of free rotation, solving the motion analytically but opaquely

1834

— Poinsot's Théorie nouvelle de la rotation des corps recasts free rotation as an ellipsoid rolling on a fixed plane, naming the polhode and herpolhode

1849

— Jacobi expresses the asymmetric top's motion in elliptic functions, completing the analytic solution Poinsot had complained about

1897

— Klein and Sommerfeld's Über die Theorie des Kreisels makes the geometric method the backbone of the definitive top treatise

1958

— Explorer 1's tumble teaches spacecraft engineers to read Poinsot's picture with energy dissipation added: the polhode spirals to the major axis

Going further

References. Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §5.6 · Landau & Lifshitz, Mechanics, 3rd ed., §37 · Morin, Introduction to Classical Mechanics, Ch. 9 · Taylor, Classical Mechanics, Ch. 10 · Klein & Sommerfeld, Über die Theorie des Kreisels, Vol. I The key relations are: (i) $\boldsymbol\rho^{\mathsf T}I\boldsymbol\rho=1$ from $\boldsymbol\rho=\boldsymbol\omega/\sqrt{2T}$ ; (ii) the ellipsoid normal $\propto I\boldsymbol\rho\parallel\mathbf L$ ; (iii) $d=\sqrt{2T}/|\mathbf L|$ ; (iv) zero contact velocity $\boldsymbol\omega\times\boldsymbol\rho=0$ ; (v) the polhode as intersection of $\sum I_i\omega_i^2=2T$ and $\sum I_i^2\omega_i^2=|\mathbf L|^2$ , with sign analysis giving stability of extreme axes; (vi) symmetric-top rates $\Omega_b=(I_3-I_\perp)\omega_3/I_\perp$ and $\Omega_{\text{prec}}=|\mathbf L|/I_\perp$ . Worked-example numbers on this page (satellite $2T=41.1\ \mathrm{J}$ , $L=40.3\ \mathrm{kg\,m^2/s}$ , $\ell=39.5\ \mathrm{kg\,m^2}$ , $d=0.159$ ; Earth wobble $\Omega_b=2.39\times10^{-7}\ \mathrm{rad/s}$ , 304 days, polhode radius 6.2 m) are illustrative arithmetic computed from the boxed formulas; the Chandler-wobble 433-day figure and the flattening $1/305$ are standard geophysical values.

Builds on: Inertia Ellipsoid · Angular Momentum of a Rigid Body · Rotational Kinetic Energy · Principal Axes and Principal Moments of Inertia · Euler's Equations of Motion Leads to: Classification of Rigid Bodies: Spherical, Symmetric, and Asymmetric Tops · Euler Angles · Angular Velocity in Terms of Euler Angles

Is the inertia ellipsoid the same thing as the energy ellipsoid I meet in the ω-space analysis?

Same surface, two costumes. The energy ellipsoid in $\boldsymbol{\omega}$ -space is $\sum I_i\omega_i^2=2T$ , whose size depends on the motion through $T$ ; the inertia ellipsoid is its normalized copy $\sum I_i\rho_i^2=1$ in the scaled variable $\boldsymbol{\rho}=\boldsymbol{\omega}/\sqrt{2T}$ , a purely geometric body-fixed object independent of how the body happens to be spinning. Poinsot works with the normalized version precisely so that one rigid surface serves every possible motion, with the launch conditions entering only through the plane's distance $d=\sqrt{2T}/L$ and the track dial $\ell=L^2/2T$ .

What exactly is invariable about the invariable plane?

Both of the things a plane can do — turn and translate along its normal — are frozen. Its orientation is slaved to $\hat{\mathbf{L}}$ , which zero torque nails to the inertial frame; its distance from the body's center is $\sqrt{2T}/L$ , built entirely from constants of the motion. Note the reference point: the plane is at fixed distance from the center of mass, so in the CM frame it is literally motionless. If the body's CM also translates uniformly, the tidy statement lives in the comoving frame — the rolling picture describes the rotational sector only.

The rolling is without slipping — does the ellipsoid roll at a constant rate along its track?

No, and this is a favorite subtlety. Slippless means the contact point has zero velocity, not that it advances along the polhode uniformly. The contact speed along the track is set by the curvature of the trajectory in $\boldsymbol{\omega}$ -space — Euler's equations underneath — and it varies strongly: crawling where the polhode passes near a stable vertex, sprinting along near-separatrix stretches. That non-uniformity is exactly why the herpolhode generally fails to close: each polhode circuit deposits an arc of different angular extent on the plane, and the arcs only line up for rational frequency ratios.

Does the herpolhode really never close? The polhode always does?

The polhode is always a closed curve: it is the intersection of two body-fixed quadrics, and the body-frame motion is exactly periodic along it. The herpolhode closes only if the polhode period and the precession period about $\mathbf{L}$ are commensurate — a measure-zero coincidence for the asymmetric top, so generically it wanders forever inside an annulus between two circles centered on the $\mathbf{L}$ -piercing point (and, a classical curiosity, it is always concave toward that center — it has no inflection points). The detailed anatomy of both curves, including the annulus radii and the symmetric-top degeneration to circles, is Polhode and Herpolhode Curves.

How does energy dissipation show up in the Poinsot picture?

As a slow leak between tracks. Internal flexing or fuel slosh drains $T$ while leaving the space-frame $\mathbf{L}$ untouched, so the dial $\ell=L^2/2T$ rises through the family of polhodes: the contact spirals from ring to ring, crosses the separatrix if it started on the minor-axis side, and asymptotes to the ring family around the major axis, ending in pure spin about $I_3$ (the minimum-energy state at fixed $L$ , $d\to1/\sqrt{I_3}$ ). One picture thus contains the entire "major-axis rule" that spacecraft engineers learned from Explorer 1 — the rigid construction gives the instantaneous motion, and dissipation just walks the state slowly across the map.

Poinsot gives me the geometry — where did the time dependence go?

Deliberately out of frame. The construction tells you the complete shape of the motion — which curves, which stability, which frames — but the clock along the polhode still comes from integrating Euler's equations, which for the asymmetric top yields Jacobi elliptic functions. This division of labor is the construction's philosophy: qualitative questions (is this spin safe? which axis does the wobble encircle? what does the motion look like from space?) get answered without a single elliptic integral, and the analytic machinery is reserved for the one question geometry cannot answer, when the body reaches a given point of its track.

Is there anything special about using the inertia ellipsoid rather than some other quadric?

The inertia ellipsoid is the unique quadric that makes both conservation laws geometric at once: its radius encodes $T$ (putting $\boldsymbol{\rho}$ on the surface) and its normal encodes $\mathbf{L}$ (nailing the tangent plane). A closely related alternative does exist — the Binet (angular-momentum) ellipsoid $\sum L_i^2/I_i=2T$ in $\mathbf{L}$ -space, on which the body-frame $\mathbf{L}$ traces the polhode's twin on the momentum sphere; that version is often handier for stability proofs, and it is the one used in the sphere-picture of the intermediate-axis analysis. The rolling-on-a-plane theater, though, belongs to the inertia ellipsoid alone.