Lagrangian Formulation of Rigid-Body Motion

The spinning top is the problem where torque bookkeeping goes to die. Newton's route demands the torque about a moving pivot, angular momentum referred to tumbling axes, fictitious forces if you dare to ride the body — a thicket of vector gymnastics for what is, after all, a single rigid object falling over while spinning. Lagrange's Mécanique analytique (1788), a book famously proud of containing not one figure, dissolved the thicket: choose three orientation angles, write the kinetic energy, subtract the potential, and turn the crank. The top that bears his name — the heavy symmetric top — fell out solved, its precession and nutation reduced to a bead sliding in a one-dimensional effective potential. This article builds that machine for a general rigid body. The kinetic energy splits cleanly into center-of-mass and rotational parts; the rotational part, $\tfrac12\sum I_i\omega_i^2$ in principal axes, becomes an explicit function of the Euler angles $(\phi,\theta,\psi)$ and their rates once the kinematic dictionary $\omega_i(\dot\phi,\dot\theta,\dot\psi)$ is inserted. Then the Euler–Lagrange equations do three things at once: they reproduce Euler's equations of motion exactly (we prove the equivalence, term by term, for the spin axis); they expose every conservation law as a cyclic coordinate — no hunting, no guessing; and they reduce the heavy top to one degree of freedom in four lines. The payoff is strategic as much as computational. Torques, constraints, springs, gravity — anything with a potential — enters through one scalar function, and the formalism decides which of the top's three angles matter. Watching the same equations emerge from $\delta\int L\,dt=0$ that Euler extracted from $\dot{\mathbf{L}}=\boldsymbol{\tau}$ is one of mechanics' great consistency checks — and choosing which face to use is a skill this article teaches explicitly.

Key takeaways

A rigid body's Lagrangian splits as $L=\tfrac12M\dot{\mathbf{R}}^2+\tfrac12\sum_iI_i\omega_i^2-U$ : center-of-mass translation plus rotation about it, with the body-frame principal axes making the rotational term diagonal.
Substituting the Euler-angle kinematics $\omega_i(\phi,\theta,\psi,\dot\phi,\dot\theta,\dot\psi)$ yields an explicit three-coordinate Lagrangian; for a symmetric top it collapses to $\tfrac12I_1(\dot\theta^2+\dot\phi^2\sin^2\theta)+\tfrac12I_3(\dot\psi+\dot\phi\cos\theta)^2-U$ .
The Euler–Lagrange equation for $\psi$ reproduces the third Euler equation $I_3\dot\omega_3-(I_1-I_2)\omega_1\omega_2=\tau_3$ exactly — the two formulations are one dynamics in two costumes.
Cyclic coordinates hand over conservation laws mechanically: for the heavy symmetric top, $\phi$ and $\psi$ are cyclic, giving conserved $p_\phi=L_z$ and $p_\psi=L_3$ and reducing the motion to $\tfrac12I_1\dot\theta^2+U_{\text{eff}}(\theta)=E'$ .
Steady precession, nutation bands, and the sleeping-top threshold all drop out of $U_{\text{eff}}$ ; the fast-top limit recovers the gyroscope rate $\dot\phi\approx Mg\ell/(I_3\omega_3)$ .
The Euler-angle momenta are projections on space and node axes — not body-frame angular-momentum components; conflating them is the classic error of the subject.
Key idea: Write $T=\tfrac12\sum_i I_i\omega_i^2$ in principal body axes, express $\omega_i$ through the Euler angles, form $L=T-U$ , and let the Euler–Lagrange machinery generate the rotational equations of motion — recovering Euler's equations and exposing all conservation laws via cyclic coordinates.

Where this lives today

The Euler-angle Lagrangian is the working chassis of multibody dynamics. Robotics toolchains and physics engines assemble exactly the kinetic energy of equation (1) for every link of an arm or limb — König split, principal-axis inertia, generalized coordinates — and crank the Euler–Lagrange machinery numerically; the "equations of motion" a robot controller inverts each millisecond are this article at industrial scale. In spacecraft engineering, the reduction of the heavy top underwrites momentum-bias attitude designs and control-moment-gyro steering laws, while the fast/slow precession branches of the steady-state quadratic reappear as the operating modes of gyrocompasses and inertial platforms. The formalism's deepest modern echo is conceptual: the rigid body is the canonical example where the configuration space is a group (the rotations), and the Lagrangian route on Euler angles sits beside the Euler–Poincaré route on body-frame velocities as the two standard ways to do mechanics on such spaces — the template for everything from molecular rotational spectra (the quantized symmetric top's $p_\phi,p_\psi$ become the quantum numbers $M$ and $K$ ) to the geometric mechanics of swimming robots and falling cats.

The physical picture

A rigid body's configuration is six numbers: where its center of mass sits, and how it is turned. The translation sector is ordinary particle mechanics; everything distinctive lives in the orientation sector, coordinatized by the Euler angles (constructed in Euler Angles): $\phi$ swings the body about the space-fixed vertical, $\theta$ tips it along the line of nodes, $\psi$ spins it about its own third axis. Each angle's rate contributes a piece of angular velocity about its own instantaneous axis, and the total $\boldsymbol{\omega}$ is their vector sum — the kinematic dictionary assembled in Angular Velocity in Terms of Euler Angles.

Figure

The three Euler-angle rates and the axes they turn about: the raw material of the rigid-body Lagrangian

Why funnel a rotation problem through generalized coordinates at all? Because the Lagrangian formalism trades vector bookkeeping for scalar bookkeeping. Gravity on a tilted top is not three torque components about a moving pivot; it is the single term $Mg\ell\cos\theta$ in $U$ . And the formalism reads structure off that scalar instantly: whatever angles $U$ does not contain are cyclic, and each cyclic angle donates a conserved momentum (the general mechanism of Cyclic Coordinates and Conserved Momenta). For the heavy symmetric top, two of three angles are cyclic — the problem is two-thirds solved before any equation of motion is written. That is the strategic content of this article: Euler's equations (see Euler's Equations of Motion) and the Lagrangian route are the same dynamics, but they surface different structure. Torque-free tumbling questions want Euler; anything with potentials, constraints, or symmetry wants Lagrange.

Building the Lagrangian

Two structural facts make the rigid-body Lagrangian tractable. First, König's decomposition: for any rigid body the kinetic energy splits exactly into translation of the center of mass plus rotation about it,

L=\tfrac12\,M\dot{\mathbf{R}}^2+\tfrac12\,\boldsymbol{\omega}^{\mathsf T}\mathsf{I}\,\boldsymbol{\omega}-U \qquad (1)

where: $M,\dot{\mathbf{R}}$ — total mass and CM velocity; $\mathsf{I}$ — the inertia tensor about the CM (or about a fixed pivot, in which case the translational term is dropped and $\mathsf{I}$ is taken about the pivot). Second, choosing principal axes diagonalizes the rotational term into $\tfrac12(I_1\omega_1^2+I_2\omega_2^2+I_3\omega_3^2)$ — three squares, no products of inertia (the diagonalization story of Principal Axes and Principal Moments of Inertia). The rotational term is not yet a Lagrangian: the $\omega_i$ are not time derivatives of any coordinates. The Euler-angle dictionary supplies the missing change of variables. Decomposing $\boldsymbol{\omega}=\dot\phi\,\hat{\mathbf{z}}+\dot\theta\,\hat{\mathbf{n}}+\dot\psi\,\hat{\mathbf{e}}_3$ and projecting the space vertical $\hat{\mathbf{z}}$ and the node direction $\hat{\mathbf{n}}$ onto body axes gives

\begin{aligned}\omega_1&=\dot\phi\sin\theta\sin\psi+\dot\theta\cos\psi\\ \omega_2&=\dot\phi\sin\theta\cos\psi-\dot\theta\sin\psi\\ \omega_3&=\dot\phi\cos\theta+\dot\psi\end{aligned} \qquad (2)

where: the derivation of these projections is the entire subject of Angular Velocity in Terms of Euler Angles — here they are imported as the dictionary. Substituted into (1) they produce an explicit, if bulky, $L(\phi,\theta,\psi,\dot\phi,\dot\theta,\dot\psi)$ . The bulk melts under symmetry: for $I_1=I_2$ the $\psi$ -dependence cancels via $\sin^2\psi+\cos^2\psi=1$ (the cross terms between $\dot\phi\sin\theta$ and $\dot\theta$ cancel identically), leaving the symmetric-top workhorse

T_{\text{rot}}=\tfrac12\,I_1\big(\dot\theta^2+\dot\phi^2\sin^2\theta\big)+\tfrac12\,I_3\big(\dot\psi+\dot\phi\cos\theta\big)^2 \qquad (3)

where: $I_1=I_2$ — the transverse moments, $I_3$ — the axial moment, all about the pivot for a pivoted top. Equation (3) is worth reading before using: the first bracket is the energy of the axis's own motion on the sphere of directions (tilting at $\dot\theta$ , swinging at $\dot\phi\sin\theta$ ), the second is the spin about the axis — and the only place $\psi$ or $\phi$ appear undifferentiated is $\cos\theta$ . Two coordinates have gone cyclic before any dynamics is invoked. From here the equations of motion are the standard crank $\frac{\mathrm{d}}{\mathrm{d}t}\frac{\partial L}{\partial\dot q_k}=\frac{\partial L}{\partial q_k}$ for $q_k\in\{\phi,\theta,\psi\}$ . Before turning it, settle the equivalence question.

Derivation AThe Euler-Lagrange equation for the spin angle is the third Euler equation

We are proving that the Lagrangian route and the Newton–Euler route are the same dynamics, by showing that the Euler–Lagrange equation for $\psi$ , applied to the general asymmetric Lagrangian $L=\tfrac12\sum_iI_i\omega_i^2-U(\phi,\theta,\psi)$ , is term-for-term the third Euler equation $I_3\dot\omega_3-(I_1-I_2)\omega_1\omega_2=\tau_3$ . The strategy is chain-rule bookkeeping through the dictionary (2): compute $\partial L/\partial\dot\psi$ and $\partial L/\partial\psi$ by differentiating the $\omega_i$ , and watch two remarkable identities appear — the $\psi$ -derivatives of the transverse spin components rotate into each other. That rotation identity is the point to savor: it is how the nonlinear gyroscopic coupling of Euler's equations hides inside a plain partial derivative.

Step 1. The conjugate momentum. From (2), only $\omega_3$ contains $\dot\psi$ , with coefficient one:

p_\psi=\frac{\partial L}{\partial\dot\psi}=\sum_i I_i\omega_i\,\frac{\partial\omega_i}{\partial\dot\psi}=I_3\,\omega_3

Why this is the body-axis angular momentum: $\hat{\mathbf{e}}_3$ is the axis the coordinate $\psi$ genuinely rotates about, so its conjugate momentum is the angular-momentum projection on that axis — the one Euler-angle momentum with a clean body-frame meaning.

Step 2. The coordinate derivative. Differentiate the dictionary (2) with respect to $\psi$ at fixed rates:

\frac{\partial\omega_1}{\partial\psi}=\dot\phi\sin\theta\cos\psi-\dot\theta\sin\psi=\omega_2,\qquad \frac{\partial\omega_2}{\partial\psi}=-\dot\phi\sin\theta\sin\psi-\dot\theta\cos\psi=-\omega_1

and $\partial\omega_3/\partial\psi=0$ . So the kinetic part of $\partial L/\partial\psi$ is

\sum_i I_i\omega_i\frac{\partial\omega_i}{\partial\psi}=I_1\omega_1\omega_2-I_2\omega_2\omega_1=(I_1-I_2)\,\omega_1\omega_2

Why the derivatives rotate into each other: increasing $\psi$ turns the body axes 1 and 2 about $\hat{\mathbf{e}}_3$ underneath a fixed $\boldsymbol{\omega}$ , so the components transform as a rotation — an infinitesimal rotation of a vector's components is exactly the swap $(\omega_1,\omega_2)\to(\omega_2,-\omega_1)$ .

Step 3. The generalized force. The potential contributes $-\partial U/\partial\psi$ , which is precisely the torque component about $\hat{\mathbf{e}}_3$ : rotating the body by $\delta\psi$ about $\hat{\mathbf{e}}_3$ changes the potential by $-\tau_3\,\delta\psi$ by the definition of torque as the generator of rotations.

Step 4. Assemble the Euler–Lagrange equation $\frac{\mathrm{d}}{\mathrm{d}t}p_\psi-\partial L/\partial\psi=0$ :

I_3\,\dot\omega_3-(I_1-I_2)\,\omega_1\omega_2=\tau_3

— the third Euler equation, exactly, with no symmetry assumption. The other two Euler equations follow from the $\phi$ and $\theta$ equations combined with the kinematic relations, or — slicker — by relabeling which body axis plays the role of "third" in the Euler-angle construction, since the choice was arbitrary. Verified: matches Goldstein §5.7 and Landau–Lifshitz §35 (where this is the standard route to Euler's equations); the rotation identities in Step 2 can be confirmed by direct differentiation of (2); and dimensional check: every term is a torque.

Which formulation when

the problem	reach for	because
torque-free tumbling, stability, polhodes	Euler's equations	body-frame $\boldsymbol{\omega}$ is the natural variable; no potential to exploit
top or gyroscope in gravity	Euler-angle Lagrangian	$U=Mg\ell\cos\theta$ makes two angles cyclic — eq. (4) below
rotor on gimbals, springs, constraints	Lagrangian	constraints eliminate coordinates or enter via multipliers
attitude simulation	Euler eqs + quaternion kinematics	avoids the $\theta=0,\pi$ coordinate singularity of Euler angles
conservation-law hunting	Lagrangian	cyclic coordinates and Noether do the hunting for you

The showcase: reducing the heavy symmetric top

The heavy symmetric top — pivoted at a point on its axis a distance $\ell$ below the CM, gravity pulling — is the formalism's demonstration piece. Its Lagrangian is (3) minus $U=Mg\ell\cos\theta$ , and neither $\phi$ nor $\psi$ appears undifferentiated: both momenta

p_\psi=I_3(\dot\psi+\dot\phi\cos\theta)=I_3\omega_3,\qquad p_\phi=\big(I_1\sin^2\theta+I_3\cos^2\theta\big)\dot\phi+I_3\dot\psi\cos\theta \qquad (4)

where: $p_\psi$ — the spin angular momentum about the body axis; $p_\phi$ — the angular momentum about the space vertical ( $L_z$ ): gravity's torque is horizontal, so the vertical component is conserved, and the formalism knows it without being told. These two constants strip the problem to one coordinate.

Go deeper · Derivation B · 1 minTwo cyclic angles collapse the top to one dimension

Derivation BTwo cyclic angles collapse the top to one dimension

We are reducing the heavy symmetric top from three degrees of freedom to a single equation for the tilt $\theta(t)$ , using the two cyclic momenta plus energy conservation. The strategy is elimination: solve the two momentum integrals (4) for $\dot\phi$ and $\dot\psi$ as functions of $\theta$ , then substitute into the conserved energy, which becomes kinetic-in- $\dot\theta$ plus an effective potential. This is the same maneuver that turns the Kepler problem into a radial problem — angular integrals first, then a one-dimensional landscape. Watch what happens to the $\dot\phi^2$ energy: it converts into a $\theta$ -dependent barrier, and every qualitative feature of top motion is a reading of that barrier.

Step 1. Solve the momentum pair for the rates. Subtract $\cos\theta$ times the first of (4) from the second:

\dot\phi=\frac{p_\phi-p_\psi\cos\theta}{I_1\sin^2\theta},\qquad \dot\psi=\frac{p_\psi}{I_3}-\dot\phi\cos\theta

Why the combination $p_\phi-p_\psi\cos\theta$ rules everything: it is the part of the vertical angular momentum not supplied by the spin itself — the amount the precession must furnish, at a $1/\sin^2\theta$ price that blows up at the poles.

Step 2. Write the conserved energy (the Lagrangian has no explicit time dependence) and substitute:

E=\tfrac12 I_1\dot\theta^2+\tfrac12 I_1\dot\phi^2\sin^2\theta+\frac{p_\psi^2}{2I_3}+Mg\ell\cos\theta

The spin term $p_\psi^2/2I_3$ is a constant; absorb it into $E'\equiv E-p_\psi^2/2I_3$ and insert $\dot\phi(\theta)$ :

E'=\tfrac12 I_1\dot\theta^2+\underbrace{\frac{(p_\phi-p_\psi\cos\theta)^2}{2I_1\sin^2\theta}+Mg\ell\cos\theta}_{U_{\text{eff}}(\theta)}

Why this is a complete solution in principle: one degree of freedom with a conserved energy integrates by quadrature — $\theta(t)$ from inverting an integral, then $\phi$ and $\psi$ by integrating Step 1's expressions along it.

Step 3. Read the landscape. $U_{\text{eff}}$ diverges at $\theta\to0,\pi$ (unless $p_\phi=\pm p_\psi$ ) and has a single minimum $\theta_0$ between them: the tilt oscillates in the well between turning points $\theta_1\le\theta\le\theta_2$ — nutation — while $\dot\phi(\theta)$ carries the axis around the vertical — precession, possibly reversing sign within the band if $p_\phi-p_\psi\cos\theta$ changes sign. A top parked exactly at the minimum precesses steadily at constant tilt. Verified: matches Goldstein §5.7 and Landau–Lifshitz §35, problem 1; the Kepler-style reduction structure (two cyclic + energy → 1D effective problem) is the standard integrable-top mechanism; limiting check: for $g\to0$ the well flattens into the free symmetric top's steady precession.

Figure

The effective potential of the heavy symmetric top: nutation is a bead oscillating in this well while the precession rate rides along

The full phenomenology of that well — the shapes of nutation loops, the sleeping top's wake-up, the cusp motion of a released gyroscope — is developed in Heavy Symmetric Top and the fast-spin story in Gyroscopic Precession. What belongs to this article is the piece of it that the Lagrangian delivers in closed form: the steady state.

Go deeper · Derivation C · 1 minSteady precession: the quadratic that governs every gyroscope

Derivation CSteady precession: the quadratic that governs every gyroscope

We are finding the condition for the top to precess at constant tilt $\theta_0$ — no nutation — and extracting the two possible precession rates. The strategy: demand $\ddot\theta=0$ in the $\theta$ Euler–Lagrange equation (equivalently, sit at the minimum of $U_{\text{eff}}$ ), which yields a quadratic in the precession rate $\dot\phi$ . The quadratic structure is the physics: gravity, the spin coupling, and the centrifugal-like $\dot\phi^2$ term must balance, and two-term balances are possible in two different ways.

Step 1. The $\theta$ equation of motion from the symmetric-top Lagrangian, written with $p_\psi=I_3\omega_3$ :

I_1\ddot\theta=\big(I_1\dot\phi\cos\theta-p_\psi\big)\,\dot\phi\sin\theta+Mg\ell\sin\theta

Set $\ddot\theta=0$ at $\theta=\theta_0$ and divide by $\sin\theta_0\neq0$ :

I_1\dot\phi^2\cos\theta_0-p_\psi\dot\phi+Mg\ell=0

Why a quadratic and not a linear balance: the precession itself carries angular momentum whose vertical tilt feeds back a $\dot\phi^2$ gyroscopic term — the top leans on its own precession.

Step 2. Solve:

\dot\phi_{\pm}=\frac{p_\psi\pm\sqrt{p_\psi^2-4I_1Mg\ell\cos\theta_0}}{2I_1\cos\theta_0}

Real solutions require $p_\psi^2\ge4I_1Mg\ell\cos\theta_0$ — a minimum spin for steady precession at a given tilt. Above it, two steady rates coexist: slow and fast precession. Why two rates: the balance can be gravity-versus-spin-coupling (slow branch) or spin-coupling-versus-centrifugal (fast branch, essentially free precession that barely notices gravity). Real released gyroscopes select the slow branch because typical initial conditions sit near it energetically.

Step 3. Expand the slow root for a fast top ( $p_\psi^2\gg4I_1Mg\ell\cos\theta_0$ ):

\dot\phi_{\text{slow}}\approx\frac{Mg\ell}{p_\psi}=\frac{Mg\ell}{I_3\,\omega_3}

— the textbook gyroscope rate: tilt-independent, inversely proportional to spin. The fast root approaches $p_\psi/(I_1\cos\theta_0)$ , the free-top precession in disguise. Verified: matches Taylor Ch. 10 and Morin Ch. 9; the discriminant condition reproduces the sleeping-top threshold at $\theta_0\to0$ ; and Worked Example 1 confirms the exact and approximate slow roots agree to the expected $O(4I_1Mg\ell\cos\theta_0/p_\psi^2)$ .

Limiting cases

Case	Result	Meaning
spherical top $I_1=I_2=I_3$	$T=\tfrac12I(\dot\phi^2+\dot\theta^2+\dot\psi^2+2\dot\phi\dot\psi\cos\theta)$	maximal symmetry; free motion is uniform rotation about a fixed axis
symmetric top, $U=U(\theta)$	$\phi,\psi$ cyclic → eq. (4), 1D reduction	the integrable heavy top: precession + nutation from $U_{\text{eff}}$
asymmetric top, torque-free	no cyclic angle survives in general	Lagrangian exists but Euler's equations are the practical tool
fast spin $p_\psi^2\gg4I_1Mg\ell$	$\dot\phi\to Mg\ell/I_3\omega_3$	the gyroscope regime; nutation shrinks to a fast small ripple
$g\to0$	$U_{\text{eff}}$ barrier only	free symmetric top: steady precession at constant tilt recovered
$\theta\to0$ with $p_\phi=p_\psi$	barrier term finite: sleeping top	stable upright spin iff $p_\psi^2>4I_1Mg\ell$

Common trapThe Euler-angle momenta are the body-frame angular momentum components — three momenta, only one of them is

It is fatally tempting to read $(p_\phi,p_\theta,p_\psi)$ as $(L_1,L_2,L_3)$ dressed in new letters. They are not: each Euler-angle momentum is the angular-momentum projection on that coordinate's own rotation axis, and the three axes ( $\hat{\mathbf{z}}$ space vertical, $\hat{\mathbf{n}}$ line of nodes, $\hat{\mathbf{e}}_3$ body axis) are neither mutually orthogonal nor body-fixed as a set. Only $p_\psi=I_3\omega_3=L_3$ has a clean body-frame identity; $p_\phi$ is the space-vertical component $L_z$ ; $p_\theta=I_1\dot\theta$ (symmetric case) is the node-axis projection. Corollaries that bite: you cannot get $L^2$ by summing squares of the three conjugate momenta; the kinetic energy is not $\sum p_k^2/2I_k$ ; and conservation of $p_\phi$ is a statement about the lab vertical, not about any body axis. A second, separate trap: the dictionary (2) degenerates at $\theta=0,\pi$ (gimbal lock — $\phi$ and $\psi$ rotate about the same axis), so numerical work near the poles switches to quaternions even though the physics is perfectly regular there.

Worked examples

1 · A toy gyroscope, exactly and approximately (confidence builder)

A toy gyroscope has a disk rotor of mass $M=0.15\ \mathrm{kg}$ with axial moment $I_3=3.0\times10^{-5}\ \mathrm{kg\,m^2}$ ; about the pivot the transverse moment is $I_1=2.5\times10^{-4}\ \mathrm{kg\,m^2}$ , and the CM sits $\ell=4.0\ \mathrm{cm}$ from the pivot. It spins at $\omega_3=400\ \mathrm{rad/s}$ and precesses steadily at tilt $\theta_0=60°$ . Find the slow precession rate, exactly and in the fast-top approximation.

Step 1. The constants: $p_\psi=I_3\omega_3=3.0\times10^{-5}\times400=1.2\times10^{-2}\ \mathrm{kg\,m^2/s}$ and $Mg\ell=0.15\times9.8\times0.040=5.88\times10^{-2}\ \mathrm{N\,m}$ .

Step 2. Check the steady-precession discriminant at $\cos60°=0.5$ : $p_\psi^2=1.44\times10^{-4}$ versus $4I_1Mg\ell\cos\theta_0=4\times2.5\times10^{-4}\times5.88\times10^{-2}\times0.5=2.94\times10^{-5}$ . Comfortably real — and the ratio $\approx4.9$ says "fast, but not absurdly fast": expect visible corrections to the naive formula.

Step 3. Exact slow root: $\dot\phi_-=\dfrac{p_\psi-\sqrt{p_\psi^2-4I_1Mg\ell\cos\theta_0}}{2I_1\cos\theta_0}=\dfrac{0.0120-\sqrt{1.44\times10^{-4}-2.94\times10^{-5}}}{2\times2.5\times10^{-4}\times0.5}=\dfrac{0.0120-0.01071}{2.5\times10^{-4}}=\boxed{5.2\ \mathrm{rad/s}}$ — a stately 0.82 revolutions per second of precession.

Step 4. Fast-top approximation: $\dot\phi_{\text{slow}}\approx Mg\ell/p_\psi=5.88\times10^{-2}/1.2\times10^{-2}=4.9\ \mathrm{rad/s}$ — 6% below the exact value, exactly the size the discriminant ratio predicted. (The coexisting fast branch: $\dot\phi_+=(0.0120+0.01071)/2.5\times10^{-4}\approx91\ \mathrm{rad/s}$ — the near-free precession no released toy ever selects.)

2 · When does the top sleep? (exam level)

The same gyroscope is stood vertically ( $\theta=0$ ) and spun up. (a) Find the minimum spin rate for the upright position to be stable ("sleeping"). (b) Friction bleeds the spin at 2% per second; if it starts at $400\ \mathrm{rad/s}$ , how long until it wakes? (c) What happens then?

Step 1. (a) At $\theta=0$ the vertical is an equilibrium; expanding $U_{\text{eff}}$ (with $p_\phi=p_\psi$ , the sleeping condition) to second order in $\theta$ gives stability iff $p_\psi^2>4I_1Mg\ell$ . The threshold spin: $\omega_{3,\min}=\dfrac{\sqrt{4I_1Mg\ell}}{I_3}=\dfrac{\sqrt{4\times2.5\times10^{-4}\times5.88\times10^{-2}}}{3.0\times10^{-5}}=\dfrac{7.67\times10^{-3}}{3.0\times10^{-5}}=\boxed{256\ \mathrm{rad/s}}$ — about 2440 rpm.

Step 2. (b) Exponential decay $\omega_3(t)=400\,e^{-0.02t}$ reaches 256 rad/s when $t=\ln(400/256)/0.02=\ln(1.564)/0.02=0.447/0.02=\boxed{22\ \mathrm{s}}$ .

Step 3. (c) At the threshold the upright equilibrium turns from a minimum of $U_{\text{eff}}$ into a maximum: the top wakes — the axis spirals out to a growing nutation-plus-precession wobble, settling toward the finite-tilt minimum of the tilted-well problem of Example 1, at an ever-slower spin. The familiar biography of every playground top — long sleep, sudden wobble, staggering collapse — is the effective potential deforming in real time as $p_\psi$ decays.

Origins · 1 minJoseph-Louis Lagrange and how this was discovered

Joseph-Louis Lagrange

1736–1813 · Turin, Berlin, Paris

Lagrange built the machine this article runs. Self-taught in Turin, he was doing capital work on the calculus of variations at nineteen — the letter that made Euler delay publishing his own version — and by the Mécanique analytique of 1788 he had recast the whole of mechanics as consequences of a single variational structure: generalized coordinates, the function now bearing his name, and equations that generate themselves from it. He boasted the book contained no figures ("on ne trouvera point de Figures dans cet Ouvrage"), mechanics reduced to analysis — a striking philosophy to apply to the most visual object in the science, the spinning top. Yet apply it he did: the pivoted symmetric top in gravity, solved there by exactly the cyclic-coordinate reduction of this article, is known simply as the Lagrange top, the second of the three integrable heavy tops (Euler's free body before, Kovalevskaya's a century after). The same book delivered the multiplier method for constraints and the planetary perturbation theory that filled the Paris academy's prize lists. Napoleon made him a count; he is buried in the Panthéon; and every time a physicist writes $L=T-U$ and lets symmetry harvest the conservation laws, the figure-free book is working precisely as designed.

History

1765

— Euler's rigid-body treatise supplies the equations of motion and principal axes the Lagrangian route will later reproduce

1788

— Lagrange's Mécanique analytique solves the heavy symmetric top by generalized coordinates and cyclic momenta: the Lagrange top

1852

— Foucault names the gyroscope and demonstrates Earth's rotation with it, making the top's precession an instrument

1888

— Kovalevskaya finds the third and last integrable heavy top, winning the Prix Bordin with a case no symmetry foretells

1897

— Klein and Sommerfeld's Kreisel treatise systematizes top theory, marrying the analytic and geometric traditions

Going further

References. Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §4.9–§5.7 · Landau & Lifshitz, Mechanics, 3rd ed., §32–§35 · Morin, Introduction to Classical Mechanics, Ch. 9 · Taylor, Classical Mechanics, Ch. 10

Provenance. Equations and derivations cross-checked against Goldstein, Poole & Safko 3rd ed. §5.5–§5.7 (Euler-angle Lagrangian, heavy top, steady precession), Landau & Lifshitz Mechanics §35 (Euler equations from the Lagrangian, asymmetric top), Morin Ch. 9 and Taylor Ch. 10 (gyroscope, precession rate $Mg\ell/I_3\omega_3$ ). A verifier should recompute the angular-velocity components from the Euler-angle composition, confirm the symmetric-top kinetic-energy reduction via $\sin^2\psi+\cos^2\psi=1$ , verify $\partial\omega_1/\partial\psi=\omega_2$ and $\partial\omega_2/\partial\psi=-\omega_1$ to recover the third Euler equation, and check the sleeping-top threshold $p_\psi^2>4I_1Mg\ell$ by expanding $U_{\mathrm{eff}}$ to second order in $\theta$ . Worked-example numbers on this page (toy gyroscope $p_\psi=1.2\times10^{-2}\ \mathrm{kg\,m^2/s}$ , exact slow precession $5.2$ rad/s vs approximate $4.9$ rad/s, fast branch $\approx91$ rad/s; sleeping threshold $256$ rad/s, wake time $22$ s) are illustrative arithmetic recomputed from the boxed formulas.

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

If the Lagrangian route reproduces Euler's equations anyway, what did I actually gain?

Structure, not new physics. The Newton–Euler route hands you three coupled ODEs and leaves conservation laws to your ingenuity; the Lagrangian route makes them typographically visible — a coordinate absent from $L$ is a conserved momentum, before any equation is solved. For the heavy top that converts a three-degree-of-freedom problem into a one-dimensional effective potential in four lines, which is why every textbook solves the top this way and none solves it by torque bookkeeping. The gain compounds when constraints appear (gimbals, contacts, linkages): constraints eliminate generalized coordinates cleanly, while in the vector formulation they spawn unknown constraint forces you must carry and then cancel.

Why can't I just write Euler-Lagrange equations directly in the omega variables and skip Euler angles?

Because the $\omega_i$ are not derivatives of any coordinates — there is no function $\alpha_i(t)$ with $\dot\alpha_i=\omega_i$ (if there were, rotations would commute, and they do not). Variables like these are called quasi-velocities, and the plain Euler–Lagrange operator does not apply to them. The honest options: use genuine coordinates (Euler angles, as here) and accept trigonometric bulk, or use the generalized machinery for quasi-velocities — the Euler–Poincaré / Boltzmann–Hamel equations — which, applied to the rigid body with $T=\tfrac12\sum I_i\omega_i^2$ , outputs Euler's equations directly. That second route is elegant and is the doorway to geometric mechanics, but it is a genuinely different variational calculus, not a shortcut through this one.

What exactly goes wrong at theta = 0, and is it physical?

Purely a coordinate disease. At $\theta=0$ or $\pi$ the precession axis $\hat{\mathbf{z}}$ and the spin axis $\hat{\mathbf{e}}_3$ coincide, so $\phi$ and $\psi$ rotate about the same line: only their sum is geometrically meaningful, the dictionary (2) loses rank, and $\dot\phi$ , $\dot\psi$ can individually blow up while the physical $\boldsymbol{\omega}$ stays finite — gimbal lock. The body itself is perfectly happy spinning upright (the sleeping top does exactly that). Analytics near the pole regroups variables (e.g. $p_\phi=p_\psi$ for the sleeping analysis); numerics switches to quaternions, which cover the rotation group without any singular orientation.

How do the OTHER two Euler equations come out of the Lagrangian, if psi only gives the third?

Two honest answers. Mechanically: the $\theta$ and $\phi$ Euler–Lagrange equations are linear combinations of the first two Euler equations with the kinematic identities (2) — invert the combination and both drop out; the algebra is unilluminating but finite. Structurally: the labeling of body axes was a choice. Rebuild the Euler angles with axis 1 (or 2) playing the "figure axis" role, and the same one-line derivation delivers the first (or second) Euler equation. Since the Lagrangian formalism is covariant under relabeling, proving one cyclic-permutation representative proves the family — the standard trick, and worth internalizing as a habit of symmetry exploitation.

Is p-theta conserved for the free (torque-free) symmetric top, since there is no potential at all?

No — and the reason is instructive. Even with $U=0$ , the kinetic energy depends explicitly on $\theta$ through $\sin^2\theta$ and $\cos\theta$ factors, so $\partial L/\partial\theta\neq0$ and $p_\theta=I_1\dot\theta$ evolves. Cyclicity is a statement about the whole Lagrangian, not the potential alone. The free top instead conserves $p_\phi$ , $p_\psi$ , and $E$ — three integrals for three degrees of freedom, which is why its motion (steady precession of the axis about the fixed $\mathbf{L}$ ) is completely solvable, as the Poinsot geometry shows from the other direction.

The reduction produced an effective potential just like the Kepler problem's. Is that a coincidence?

It is the same theorem wearing two hats. Both systems have a symmetry group big enough to supply one conserved momentum per unwanted angle (rotations about the vertical for the top's $\phi$ and about the figure axis for $\psi$ ; rotations about the force center for Kepler), and eliminating cyclic angles at fixed momentum always deposits their kinetic energy as a repulsive barrier — $(p_\phi-p_\psi\cos\theta)^2/2I_1\sin^2\theta$ here, $L^2/2mr^2$ there. The recipe is general (Routh's procedure for ignorable coordinates): symmetry → momentum integral → one-dimensional landscape. Whenever you meet a $1/\sin^2\theta$ or $1/r^2$ barrier, you are looking at an angle that was integrated out.

Which of the three integrable heavy tops does this article's method actually crack, and what makes the others special?

The Lagrange top — symmetric body, pivot on the symmetry axis — is the one whose integrability is visible in the coordinates: two cyclic angles plus energy. Euler's case (no gravity torque: pivot at the CM or free body) is integrable by the momentum integrals instead, and is best handled by Euler's equations and the Poinsot picture. Kovalevskaya's case ( $I_1=I_2=2I_3$ , CM in the equatorial plane) has no cyclic coordinate to betray it; its fourth integral is a hidden quartic invariant she found in 1888 by demanding the solutions be meromorphic in complex time — a method that founded modern integrability analysis. Every other mass geometry is provably chaotic in general: the top family is the textbook exhibit of how rare and precious integrable structure is.