Polhode and Herpolhode Curves

When Louis Poinsot named the two curves of free rigid-body rotation, he reached for Greek and drama: the polhode, "path of the pole," and the herpolhode, "serpentine path of the pole." A tumbling body's instantaneous rotation axis wanders — but it wanders with iron discipline, and these two curves are its complete itinerary. The polhode is the track the axis traces through the body: a closed loop scratched on the body-fixed inertia ellipsoid, repeated exactly, cycle after cycle. The herpolhode is the track the same axis traces through space: a curve drawn on the fixed invariable plane that generically never closes, never repeats, and — despite the name Poinsot gave it — never once wiggles like a serpent, since it provably bends always toward its central point. Both curves are carved by conservation laws alone. Kinetic energy and angular momentum each confine the angular velocity to an ellipsoid in the body frame; the polhode is the intersection of the two, and one conserved ratio, $\ell=L^2/2T$ , selects which member of the family you get — tight rings around the major or minor axis, or the fateful separatrix through the intermediate axis. The herpolhode is the polhode's rolled-out image, trapped forever in an annulus whose radii you can compute from the same two constants. These curves are not museum pieces. Gravity Probe B's relativity measurement had to model the polhode wander of the roundest gyroscopes ever made; satellite operators classify the tumble of dead spacecraft by which polhode family their light-curves betray. This article derives both curves, their shapes, their bounds, and their time behavior, and puts numbers on a tumbling T-handle and a wobbling dinner plate.

Key takeaways

During torque-free rotation both $T$ and $L$ are conserved; in the body frame $\boldsymbol{\omega}$ must lie on the intersection of the energy ellipsoid $\sum I_i\omega_i^2=2T$ and the momentum ellipsoid $\sum I_i^2\omega_i^2=L^2$ — that intersection curve is the polhode.
The single conserved ratio $\ell=L^2/2T\in[I_1,I_3]$ selects the polhode: rings around the minor axis for $\ell<I_2$ , rings around the major axis for $\ell>I_2$ , and the separatrix (the homoclinic orbit of Euler's equations) at exactly $\ell=I_2$ .
Projected on the principal planes, polhodes are ellipses around the extreme axes and hyperbolas around the intermediate one — the saddle structure behind the tennis-racket flip.
The herpolhode is the contact point's track on the invariable plane: confined to an annulus $r_{\min}\le r\le r_{\max}$ with $r=\sqrt{\omega^2-(2T/L)^2}$ , generically never closing, and always concave toward the annulus center (it has no inflection points, its serpentine name notwithstanding).
Rolling without slipping means polhode and herpolhode are rolled out against each other with equal arc lengths.
For a symmetric top both curves collapse to circles — the steady wobble whose space rate is $L/I_\perp$ (twice the spin for a thin disk, Feynman's wobbling plate).
Key idea: Free rigid-body rotation is the intersection of two conserved-quantity ellipsoids: the polhode is that intersection drawn on the inertia ellipsoid in the body frame, and the herpolhode is its shadow on the invariable plane as the inertia ellipsoid rolls without slipping on that plane.

Where this lives today

The polhode had a starring role in one of the most delicate measurements ever flown. Gravity Probe B's fused-quartz gyroscopes were spherical to forty atomic layers, yet their part-per-million moment differences gave each rotor a slow polhode wander that modulated the SQUID readout and, coupled with electrostatic patch effects, produced the mission's infamous anomalous torques; the science team had to reconstruct each rotor's polhode path over months to model and remove the effect — Poinsot's 1834 curves, fitted to superconducting telemetry. The same machinery runs nightly in space-situational awareness: the light-curves of tumbling rocket bodies and dead satellites are inverted for their spin state, classified by polhode family ( $\ell$ against $I_2$ ), and tracked as internal flexing walks $\ell$ slowly toward $I_3$ — the dissipative polhode migration that also sets the tumbling states of small asteroids observed by radar. And every student demonstration of the Dzhanibekov effect is a polhode movie: the T-handle's spin axis is riding a near-separatrix curve, and the "calm, flip, calm" rhythm is just the elliptic-function clock running slow near the saddle and fast along the excursion.

The physical picture

Put yourself in the body frame first — sit on the tumbling wrench. From here the mass distribution is frozen, and the angular velocity vector swings around you, its tip forever pinned to two rigid surfaces at once: the ellipsoid of constant energy and the ellipsoid of constant momentum. Pinned to both, it must ride their intersection: a closed curve, fixed in the body, that the spin axis retraces exactly once per wobble cycle. If you dusted the body's inertia ellipsoid with chalk and let the invariable plane wear it away, the polhode is the line that would be polished onto it. Now step into the lab. The same contact point, watched from outside, moves on Poinsot's fixed invariable plane (the rolling theorem that puts it there is built in Poinsot Construction). Its distance from the plane's center — the point where the fixed vector $\mathbf{L}$ pierces it — can only breathe between two bounds set by the conserved quantities, so the space-frame track lives inside an annulus, circling the center, its loops slowly shearing around because the wobble period and the precession period refuse to divide evenly. The polhode is a closed necklace on the body; the herpolhode is the same necklace unrolled onto the floor, bead spacing preserved (no slipping — equal arc lengths), but never quite meeting itself again.

Figure

The two records of one tumble: the closed polhode on the body-fixed ellipsoid, and the never-closing herpolhode in its annulus on the invariable plane

The family portrait of all possible polhodes is the phase diagram of free rotation, and one glance at it proves the intermediate-axis theorem: rings nest around the long-axis and short-axis vertices, and the two families are fenced off from each other by a separatrix crossing at the intermediate-axis vertex. Which curve a given body actually rides is set by one number from the launch — no integration required.

The polhode: the body's own record

Everything descends from the two integrals of torque-free motion (established from Euler's equations in Euler's Equations of Motion). Written in the principal frame, each is a quadric surface in $\boldsymbol{\omega}$ -space. The energy surface is an ellipsoid whose semi-axes go as $1/\sqrt{I_i}$ :

2T=I_1\omega_1^2+I_2\omega_2^2+I_3\omega_3^2 \qquad (1)

where: $I_1\le I_2\le I_3$ — the principal moments; $\omega_i$ — body-frame spin components. The momentum surface is a second, differently proportioned ellipsoid — semi-axes $\propto1/I_i$ — since each component of $\mathbf{L}$ is $L_i=I_i\omega_i$ :

L^2=I_1^2\omega_1^2+I_2^2\omega_2^2+I_3^2\omega_3^2 \qquad (2)

where: $L$ — the fixed magnitude of the angular momentum. The polhode is the set of $\boldsymbol{\omega}$ satisfying both (1) and (2) simultaneously — a one-dimensional closed curve. Which curve, out of the two-parameter family of possible $(T,L)$ pairs, is captured by a single shape parameter with units of moment of inertia:

\ell\equiv\frac{L^2}{2T},\qquad I_1\le\ell\le I_3 \qquad (3)

where: the bounds hold because $\ell$ is a weighted average of the $I_i$ (weights $I_i\omega_i^2\ge0$ ), with equality only for pure principal-axis spin. Everything qualitative about the polhode — which axis it encircles, how fat it is, whether it is the separatrix — is a function of where $\ell$ sits relative to $I_2$ .

Derivation AThe three projections of the polhode and the topology selector

We are deriving the shape of the polhode by projecting it onto the three principal planes, which turns an awkward space curve into three plane conics — and the conic types, ellipse versus hyperbola, encode the entire stability structure. The strategy: eliminate one squared component at a time between the two conserved quadrics (1) and (2); each elimination is a single linear combination. Watch the signs of the surviving coefficients — with the ordering $I_1<I_2<I_3$ they are forced, and they differ crucially depending on whether the eliminated axis is extreme or intermediate.

Step 1. Eliminate $\omega_2$ (the intermediate component). Multiply (1) by $I_2$ and subtract from (2); the $\omega_2$ term cancels since $I_2(I_2-I_2)=0$ :

L^2-2TI_2=I_1(I_1-I_2)\,\omega_1^2+I_3(I_3-I_2)\,\omega_3^2

The first coefficient is negative, the second positive: for any value of the left side this is a hyperbola in the $(\omega_1,\omega_3)$ plane. Its branches open around the $\omega_1$ axis when $L^2>2TI_2$ (i.e. $\ell>I_2$ ) and around the $\omega_3$ axis when $\ell<I_2$ . Why this projection carries the drama: the eliminated axis is the one the conic cannot see; hyperbolic projections perpendicular to the intermediate axis are the flat-map signature of the saddle sitting on that axis.

Step 2. Eliminate $\omega_3$ the same way — multiply (1) by $I_3$ , subtract from (2):

L^2-2TI_3=I_1(I_1-I_3)\,\omega_1^2+I_2(I_2-I_3)\,\omega_2^2

Now both coefficients are negative, and the left side is negative too (since $\ell\le I_3$ ): dividing out the signs, this is an ellipse in the $(\omega_1,\omega_2)$ plane. Eliminating $\omega_1$ instead gives, by the same token, an ellipse in the $(\omega_2,\omega_3)$ plane. Why ellipses here: relative to an extreme moment, every other moment sits on the same side, so the quadratic form is definite — closed level curves, no escape routes.

Step 3. Read off the selector. The hyperbola of Step 1 degenerates into its asymptote pair exactly when $L^2=2TI_2$ , i.e. $\ell=I_2$ : two straight lines through the origin with slopes

\frac{\omega_3}{\omega_1}=\pm\sqrt{\frac{I_1(I_2-I_1)}{I_3(I_3-I_2)}}

On the ellipsoid these lift to two ellipses crossing at the intermediate-axis vertices — the separatrix. For $\ell<I_2$ the polhode is a ring around the major axis $\hat{\mathbf{e}}_3$ ; for $\ell>I_2$ a ring around the minor axis $\hat{\mathbf{e}}_1$ . (Which extreme axis gets the small- $\ell$ family is worth engraving: small $\ell$ means energy-rich motion, and at fixed $L$ the energy $L^2/2\ell$ is largest for spin about the smallest moment — yet the ring sits around the major axis when $\ell$ is small is a common sign slip; check it via pure spins: $\ell=I_3$ exactly is major-axis spin, $\ell=I_1$ is minor-axis spin, so rings shrink onto $\hat{\mathbf{e}}_3$ as $\ell\to I_3$ and onto $\hat{\mathbf{e}}_1$ as $\ell\to I_1$ .) Why the separatrix is the homoclinic orbit: the intermediate-axis vertices are the saddle fixed points of Euler's equations, and the separatrix ellipses are the trajectories that leave one saddle and arrive at the other in infinite time — the flip path of the tennis-racket instability, whose growth rate is computed in Intermediate-Axis (Tennis-Racket) Theorem. *Verified: matches Landau–Lifshitz §37 (including their Fig. 51 projections) and Goldstein §5.6; pure-spin checks $\ell=I_1,I_3$ collapse the rings to points as required; and for $I_1=I_2$ the asymptote slope goes to zero, correctly merging the separatrix with the equator of the symmetric ellipsoid.*

The time dependence along the polhode is the one thing geometry does not hand over. Solving Euler's equations along the curve yields Jacobi elliptic functions — $\omega_1\propto\operatorname{cn}(\lambda t)$ , $\omega_2\propto\operatorname{sn}(\lambda t)$ , $\omega_3\propto\operatorname{dn}(\lambda t)$ for the ring families, with the elliptic modulus measuring how close $\ell$ sits to $I_2$ — and the traversal period diverges logarithmically as the polhode approaches the separatrix. Near the extreme axes the elliptic functions soften into sines and the traversal frequency reduces to the linear wobble frequencies of the stability analysis.

The herpolhode: the record in space

Watched from the lab, the contact point lives on the invariable plane, and two facts about the fixed geometry bound its excursions. The tip of $\boldsymbol{\omega}$ always projects onto the fixed direction $\hat{\mathbf{L}}$ with the same length — that projection is $\boldsymbol{\omega}\cdot\hat{\mathbf{L}}=2T/L$ , constant — so the tip moves in a fixed plane at height

d=\frac{2T}{L} \qquad (4)

where: $d$ — the distance from the origin of $\boldsymbol{\omega}$ -space to the plane (working directly with $\boldsymbol{\omega}$ ; in the scaled variable $\boldsymbol{\rho}=\boldsymbol{\omega}/\sqrt{2T}$ used by the Poinsot article the same plane sits at $\sqrt{2T}/L$ — one geometry, two rulers). The herpolhode is the track of the tip in this plane, and its distance $r$ from the plane's center — the foot of $\mathbf{L}$ — follows from Pythagoras alone.

Go deeper · Derivation B · 2 minThe herpolhode annulus

Derivation BThe herpolhode annulus

We are deriving the bounds within which the herpolhode is imprisoned: the curve's radius from the foot of $\mathbf{L}$ oscillates between two values fixed by the constants of motion, so the track fills an annulus and can never spiral in, escape, or touch the center. The strategy is to write the radius as the transverse part of $\boldsymbol{\omega}$ relative to the fixed $\hat{\mathbf{L}}$ direction, then bound $|\boldsymbol{\omega}|$ along the polhode, where its extremes occur at the symmetry crossings of the curve. Watch which quantities are constant ( $T$ , $L$ , hence $d$ ) and which oscillate ( $|\boldsymbol{\omega}|$ ): the interplay of the two is the whole result.

Step 1. Split $\boldsymbol{\omega}$ into components along and transverse to $\hat{\mathbf{L}}$ . The parallel part is pinned at $d=2T/L$ by equation (4), so the transverse part — the herpolhode radius — is

r=\sqrt{\;\omega^2-\left(\frac{2T}{L}\right)^{\!2}\;}

Why this is the radius: the foot of $\mathbf{L}$ on the plane is the projection of the origin along $\hat{\mathbf{L}}$ ; the tip of $\boldsymbol{\omega}$ sits in the plane, displaced from that foot by exactly the transverse component of $\boldsymbol{\omega}$ .

Step 2. Bound $\omega^2$ along the polhode. $\omega^2=\sum\omega_i^2$ varies over the closed curve and takes its extremes where the curve crosses the principal planes (by the symmetry $\omega_i\to-\omega_i$ of both quadrics, these crossings are turning points). At a crossing, one component vanishes and the two conservation laws (1)–(2) become two linear equations for the two surviving squared components — solvable in closed form. Substituting the two crossing solutions into $\omega^2$ gives $\omega^2_{\min}$ and $\omega^2_{\max}$ , hence

r_{\min}=\sqrt{\omega^2_{\min}-d^2},\qquad r_{\max}=\sqrt{\omega^2_{\max}-d^2}

Why $r_{\min}>0$ generically: $r=0$ would require $\boldsymbol{\omega}\parallel\mathbf{L}$ , which happens only for pure principal-axis spin; any genuine polhode keeps a finite angle between the two vectors, so the herpolhode never visits its own center.

Step 3. Establish the shape between the bounds. The track touches the outer circle, curves inward to the inner circle, and returns, over and over, advancing in azimuth each cycle by an angle set by the precession rate. Two classical structural facts complete the portrait: the curve has no inflection points — it is everywhere concave toward the center, a theorem of the late nineteenth century (Hess) that convicts every "snaking" textbook sketch, including the name Poinsot chose — and it closes only if the radial oscillation period and the azimuthal circulation period are commensurate, a measure-zero coincidence for the asymmetric top. Generically the herpolhode is dense in its annulus: wait long enough and it passes arbitrarily close to every point between the circles. Verified: the radius formula follows from $\boldsymbol{\omega}\cdot\hat{\mathbf{L}}=2T/L$ and Pythagoras; the annulus bounds match the quadrature treatment in Landau–Lifshitz §37; the no-inflection property is the Hess–de Sparre result quoted by Routh and MacMillan; and the symmetric-top limit collapses $r_{\min}=r_{\max}$ to the circular herpolhode, as required.

Rolling without slipping stitches the two records together: the contact point is instantaneously at rest, so arc length accumulates at the same rate on both curves. One polhode circuit lays down one "petal" of herpolhode arc of exactly the polhode's length — but rotated relative to the previous petal, which is why closure requires commensurate frequencies and almost never happens.

Figure

Polhode projections on the three principal planes: ellipses seen along the extreme axes, hyperbolas with separatrix asymptotes seen along the intermediate axis

The symmetric top: both curves become circles

Degeneracy tidies everything. With $I_1=I_2=I_\perp$ , the component $\omega_3$ is separately conserved (the third Euler equation dies), so the polhode equations describe $\omega_1^2+\omega_2^2=\text{const}$ : the polhode is a circle about the symmetry axis, traversed at the uniform body-frame rate $\Omega_b=\omega_3(I_3-I_\perp)/I_\perp$ . In space, $\boldsymbol{\omega}$ holds a fixed angle to the fixed $\mathbf{L}$ and sweeps a cone about it at the uniform rate

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

where: the herpolhode is correspondingly a circle of constant radius ( $r_{\min}=r_{\max}$ : the annulus collapses). For a thin uniform disk, $I_3=2I_\perp$ forces $\Omega_{\text{prec}}\approx2\omega_3$ at small wobble angles — the plate thrown in the Cornell cafeteria that Feynman watched wobble exactly twice per... spin, or rather spin twice per wobble; he famously had to recompute which. The cone-rolling geometry behind these circles is developed in Free Symmetric Top and Body-Cone/Space-Cone Precession.

Limiting cases

Case	Result	Meaning
$\ell\to I_1$ or $\ell\to I_3$	polhode ring shrinks to a vertex point	pure principal-axis spin; herpolhode shrinks to a point, $\boldsymbol{\omega}\parallel\mathbf{L}$
$\ell$ slightly off an extreme	small ellipse-like ring; thin herpolhode annulus	stable wobble; elliptic functions reduce to sines at the linear wobble frequency
$\ell=I_2$ (separatrix)	crossing ellipses through $\hat{\mathbf{e}}_2$ ; traversal time infinite	homoclinic flip path of the intermediate-axis instability
$\ell$ near $I_2$	fat polhode sweeping most of the ellipsoid	flip-like excursions; polhode period grows as $\ln$ of the distance to the separatrix
symmetric top $I_1=I_2$	polhode and herpolhode both circles	steady precession at $L/I_\perp$ ; disk limit gives the 2:1 plate wobble
spherical top $I_1=I_2=I_3$	both curves degenerate to points	$\boldsymbol{\omega}\parallel\mathbf{L}$ always; no wobble exists

Common trapThe herpolhode snakes back and forth — it does neither

Poinsot's own name for the curve ("serpentine path") planted a picture that textbooks then drew for a century: a wavy track wiggling in and out. The real herpolhode is far more disciplined. It is trapped in a closed annulus — it cannot spiral into the center (that would need $\boldsymbol{\omega}\parallel\mathbf{L}$ , i.e. exact principal-axis spin) and cannot drift outward (energy and momentum forbid it) — and within the annulus it is everywhere concave toward the center: the Hess no-inflection theorem, so it never once bends the "wrong way." The genuine irregularity is subtler: because its radial and azimuthal periods are generically incommensurate, the curve never closes and eventually shades in the entire annulus. Meanwhile its body-frame twin, the polhode, is always exactly closed. Keep the asymmetry straight: periodic in the body, quasi-periodic in space — the body repeats its wobble perfectly while its orientation in the lab never repeats at all.

Worked examples

1 · A tumbling T-handle's two curves (confidence builder)

A T-handle has principal moments $I_1=1.0$ , $I_2=2.0$ , $I_3=3.0$ (units $10^{-3}\ \mathrm{kg\,m^2}$ ) and is set spinning at $\boldsymbol{\omega}=(10,\,0.5,\,0)\ \mathrm{rad/s}$ — nearly pure spin about the minor axis with a small kick. Classify its polhode and compute the herpolhode annulus.

Step 1. The two invariants: $2T=(1.0\times100+2.0\times0.25+0)\times10^{-3}=0.1005\ \mathrm{J}$ and $L^2=(1.0^2\times100+2.0^2\times0.25)\times10^{-6}=1.010\times10^{-4}$ , so $L=1.005\times10^{-2}\ \mathrm{kg\,m^2/s}$ .

Step 2. The selector: $\ell=L^2/2T=1.010\times10^{-4}/0.1005=1.005\times10^{-3}\ \mathrm{kg\,m^2}$ . This sits just above $I_1=1.0\times10^{-3}$ and well below $I_2=2.0\times10^{-3}$ : a tight polhode ring around the minor axis $\hat{\mathbf{e}}_1$ — stable wobble, far from the separatrix.

Step 3. Plane height from equation (4): $d=2T/L=0.1005/1.005\times10^{-2}=10.00\ \mathrm{rad/s}$ — the fixed projection of $\boldsymbol{\omega}$ on $\hat{\mathbf{L}}$ , essentially the launch spin.

Step 4. Annulus. At the launch crossing ( $\omega_3=0$ ): $\omega^2=100+0.25=100.25$ , so $r=\sqrt{100.25-100.00}=\sqrt{0.25}=0.50\ \mathrm{rad/s}$ . At the other crossing ( $\omega_2=0$ ), solve $1.0\,a+3.0\,b=100.5$ and $1.0\,a+9.0\,b=101.0$ (in the $10^{-3}$ units, with $a=\omega_1^2$ , $b=\omega_3^2$ ): subtracting, $6.0\,b=0.5$ gives $b=0.0833$ , $a=100.25$ , so $\omega^2=100.33$ and $r=\sqrt{0.333}=0.577\ \mathrm{rad/s}$ . The herpolhode fills the annulus $0.50\le r\le0.58\ \mathrm{rad/s}$ — a wobble cone of half-angle between $\arctan(0.50/10)=2.9°$ and $3.3°$ about $\mathbf{L}$ , never closing, forever concave toward the center.

2 · Feynman's plate, by the numbers (exam level)

A thin uniform dinner plate ( $I_3=2I_\perp$ ) is tossed with spin $\omega_3=6.0\ \mathrm{rev/s}$ about its symmetry axis plus a transverse component $\omega_\perp=0.50\ \mathrm{rev/s}$ . Find (a) the polhode circle and the tilt of $\boldsymbol{\omega}$ from the symmetry axis, (b) the space-frame wobble (precession) rate, and (c) the herpolhode circle radius.

Step 1. (a) With $I_1=I_2$ the polhode is the circle $\omega_\perp=0.50\ \mathrm{rev/s}$ around the symmetry axis. The spin vector tilts from that axis by $\beta=\arctan(\omega_\perp/\omega_3)=\arctan(0.50/6.0)=4.8°$ .

Step 2. (b) Angular momentum components: $L_3=I_3\omega_3=2I_\perp\times6.0=12.0\,I_\perp$ and $L_\perp=I_\perp\times0.50=0.50\,I_\perp$ (rev/s units), so $L=I_\perp\sqrt{12.0^2+0.50^2}=12.01\,I_\perp$ and $\mathbf{L}$ tilts only $\theta_L=\arctan(0.50/12.0)=2.4°$ from the symmetry axis. The precession rate from equation (5): $\Omega_{\text{prec}}=L/I_\perp=\boxed{12.0\ \mathrm{rev/s}}$ — almost exactly twice the spin, the disk's $I_3/I_\perp=2$ speaking directly. The wobble you see (the rim's oscillation) runs at the precession rate: the plate wobbles twice for every rotation of its decoration, which is what Feynman noticed (and initially quoted inverted) in the cafeteria.

Step 3. (c) The angle between $\boldsymbol{\omega}$ and $\mathbf{L}$ is $\beta-\theta_L=4.8°-2.4°=2.4°$ , and $|\boldsymbol{\omega}|=\sqrt{6.0^2+0.50^2}=6.02\ \mathrm{rev/s}$ , so the herpolhode is a circle of radius $r=6.02\times\sin2.4°=\boxed{0.25\ \mathrm{rev/s}}$ about the foot of $\mathbf{L}$ — the annulus of the asymmetric case collapsed to zero width, the circle retraced exactly once per precession period of $1/12$ s.

Origins · 1 minCarl Gustav Jacob Jacobi and how this was discovered

Carl Gustav Jacob Jacobi

1804–1851 · Königsberg

Jacobi supplied the part of the free-top problem that Poinsot's geometry deliberately left blank: the clock. His theory of elliptic functions — built in a legendary rivalry-collaboration with Abel around 1829 — turned out to be precisely the language in which the asymmetric top tells time, and in 1849 he published the complete solution: the three spin components run as his functions $\operatorname{sn}$ , $\operatorname{cn}$ , $\operatorname{dn}$ , the polhode traversed with a period given by a complete elliptic integral that swells logarithmically at the separatrix. It is the same Jacobi whose "one must always invert" reorganized the study of integrals, whose identity governs Poisson brackets, and whose reformulation of dynamics — Hamilton–Jacobi theory — would become the classical scaffolding of quantum mechanics. He worked himself into repeated breakdowns, lost his Königsberg salary in the political turmoil of 1848, and died of smallpox at 46, two years after closing a problem that had stood since Euler: between Poinsot's rolling picture and Jacobi's functions, free rotation became the rare mechanical system that is finished — every curve drawn, every instant timed.

History

1834

— Poinsot's rolling-ellipsoid memoir recasts free rotation geometrically; the polhode and herpolhode are christened in its expanded 1851 edition

1849

— Jacobi publishes the complete elliptic-function solution of the asymmetric free top, timing Poinsot's curves

1880

— Hess proves the herpolhode has no inflection points: it always bends toward the pole, retiring the serpentine sketches

1958

— Explorer 1's flat spin shows polhode migration under energy dissipation — the curves become spacecraft-engineering tools

2004

— Gravity Probe B flies the roundest gyroscopes ever made; modeling their slow polhode wander proves essential to extracting the relativity signal

Going further

References. Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Sec. 5.6 (Poinsot construction, polhode/herpolhode) · Landau & Lifshitz, Mechanics, 3rd ed., Sec. 37 (asymmetrical top, Poinsot) · Morin, Introduction to Classical Mechanics, Ch. 9 (free symmetric and asymmetric tops) · Taylor, Classical Mechanics, Ch. 10 · Arnold, Mathematical Methods of Classical Mechanics, Sec. 28-29

Provenance. Equations of energy and angular-momentum ellipsoids, the polhode-projection identity, $d=2T/L$ , the rolling-without-slipping result, and the intermediate-axis perturbation equation should each be re-derived from torque-free Euler's equations in the principal frame. Verify against Goldstein, Poole & Safko (3rd ed.) Sec. 5.6, Landau & Lifshitz Mechanics Sec. 37, and Arnold Sec. 28-29. The Hess result on herpolhode curvature and the stability frequency are standard; no numerical constants are claimed beyond exact moment-ratio expressions. Worked-example numbers on this page (T-handle $2T=0.1005$ J, $L=1.005\times10^{-2}\ \mathrm{kg\,m^2/s}$ , $\ell=1.005\times10^{-3}\ \mathrm{kg\,m^2}$ , annulus $0.50$ – $0.58$ rad/s; plate $\Omega_{\text{prec}}=12.0$ rev/s, herpolhode radius $0.25$ rev/s) are illustrative arithmetic recomputed from the boxed formulas.

Builds on: Inertia Ellipsoid · Principal Axes and Principal Moments of Inertia · Angular Momentum of a Rigid Body · Rotational Kinetic Energy · 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

Why does the polhode close exactly while the body's motion in space never repeats?

Because the two statements live in different frames with different frequency content. In the body frame the dynamics of $(\omega_1,\omega_2,\omega_3)$ is a one-degree-of-freedom flow on a closed curve — strictly periodic with the polhode period. The lab-frame orientation adds a second frequency: the precession of the body around the fixed $\mathbf{L}$ accumulated per polhode cycle. Generic frequency pairs are incommensurate, so the orientation is quasi-periodic — the body wobbles identically every cycle as felt from inside, yet never faces the room the same way twice. The herpolhode, a lab-frame record, inherits the non-closure.

Are the polhode and herpolhode the same length?

Per cycle, yes — exactly. The contact point is instantaneously at rest (it sits on the rotation axis), which is the no-slip condition, and no slipping means arc length accrues at the same rate on the ellipsoid and on the plane. One full polhode circuit therefore prints one petal of herpolhode with precisely the polhode's arc length. The petals differ only in orientation, each rotated from the last by the per-cycle precession angle — equal-length stamps from the same die, fanned around the annulus.

This article says the plane sits at distance 2T/L, but the Poinsot Construction article says sqrt(2T)/L. Which is right?

Both — they measure the same plane with different rulers. Working directly in $\boldsymbol{\omega}$ -space (this article's habit, following Landau–Lifshitz), the tip of $\boldsymbol{\omega}$ satisfies $\boldsymbol{\omega}\cdot\hat{\mathbf{L}}=2T/L$ . The Poinsot construction prefers the scaled vector $\boldsymbol{\rho}=\boldsymbol{\omega}/\sqrt{2T}$ , whose tip rides the normalized inertia ellipsoid $\sum I_i\rho_i^2=1$ ; dividing by $\sqrt{2T}$ rescales the plane distance to $\sqrt{2T}/L$ . Any statement about shapes (annulus, closure, concavity) is scale-invariant; only the numerical radii change rulers. Pick one convention per calculation and never mix mid-stream — the mixed version $2T/(L\sqrt{2T})$ appearing in some notes is the tell-tale of a mid-stream switch.

What sets the polhode period, and why does it blow up at the separatrix?

The period is a complete elliptic integral whose modulus measures how close $\ell$ sits to $I_2$ . Near an extreme axis the modulus is small, the elliptic functions degenerate to sines, and the period matches the linear wobble frequency of the stability analysis. As $\ell\to I_2$ the polhode passes ever closer to the intermediate-axis vertices — which are genuine fixed points of Euler's equations — and the trajectory lingers near them logarithmically long: the period grows as $\ln(1/|\ell-I_2|)$ and is infinite on the separatrix itself, which connects the two saddles in infinite time. That logarithm is the same one that makes the Dzhanibekov flip rhythm so sensitive to the launch.

Could I actually observe a polhode on a real object?

Three ways it is done. In light-curve inversion, the brightness time series of a tumbling satellite or asteroid carries both the wobble and precession frequencies; fitting a torque-free model recovers $\ell$ and hence the polhode family. In Gravity Probe B, the polhode of each gyro rotor modulated the trapped-flux signature read by the SQUIDs, and the mission literally published reconstructed polhode paths spanning months. And in the classroom, a phone tossed with video running (or an IMU logging) shows the body-frame angular-velocity components oscillating exactly along their polhode — the sn, cn, dn traces are recognizable in a plot from a ten-second toss.

Does energy dissipation change the polhode, or move the body between polhodes?

The second, and the distinction is the whole picture. At any instant the rigid-body dynamics carries the state along its current polhode; slow internal dissipation (flexing, slosh) drains $T$ at fixed $L$ , nudging $\ell=L^2/2T$ upward, which reselects the polhode — the state drifts across the family, ring by ring, toward the major axis at $\ell=I_3$ . On the momentum sphere this appears as a slow spiral crossing the fast closed curves. A minor-axis spinner therefore fattens its wobble, crosses the separatrix (one grand flip), and settles into shrinking rings about the major axis: Explorer 1's biography, and the standard fate of every dead tumbling spacecraft.

Is there a version of these curves on the momentum sphere instead of the inertia ellipsoid?

Yes, and it is often the better working picture. Plot the body-frame components of $\mathbf{L}$ : they live on the sphere $L^2=\text{const}$ cut by the energy ellipsoid $\sum L_i^2/2I_i=T$ , and the intersection curves are the polhodes' momentum-space twins — same topology (rings, rings, separatrix), same selector. The sphere version is preferred for stability proofs (it is the natural Casimir surface of the rigid-body Poisson bracket) and appears in the intermediate-axis and stability articles; the inertia-ellipsoid version is the one that rolls on a plane and produces the herpolhode. Translating between them is a component-wise rescaling $L_i=I_i\omega_i$ .