Classical Mechanics » Rigid Body Dynamics

The inertia tensor and principal axes, Euler’s equations and Euler angles, the symmetric and asymmetric top, gyroscopic precession and nutation, rolling constraints, and the tennis-racket theorem.

From Point Particles to Extended Bodies

A point particle has three numbers — its position — and three momenta. A rigid body is a collection of particles whose mutual distances are frozen, so the whole object is pinned down by just six numbers: three to locate a reference point (the center of mass) and three to specify the body’s orientation. The first three obey the familiar center-of-mass theorem, $\vec{F}{\text{ext}} = M\ddot{\vec{R}}{\text{cm}}$, and reduce to the projectile motion of the Newtonian chapter. The interesting physics — and most of the surprises — lives in the rotational three.

The central message of this chapter is a single subtle fact: for a spinning body the angular velocity $\vec{\omega}$ and the angular momentum $\vec{L}$ generally point in different directions. The object that converts one into the other is a matrix, the moment-of-inertia tensor. Once you accept that $\vec{L}$ and $\vec{\omega}$ are not parallel, every rigid-body phenomenon — the wobble of a thrown book, the steady drift of a gyroscope, the tumbling of a satellite — follows with mathematical inevitability.

Article: Rigid Body Dynamics - Wikipedia

Video: The Bizarre Behavior of Rotating Bodies (Dzhanibekov Effect)

Two Frames: Space and Body

Every rigid-body problem juggles two reference frames sharing the same origin (taken at the center of mass or a fixed pivot):

• The space frame (inertial, lab fixed): where Newton’s laws hold and where $d\vec{L}/dt = \vec{\tau}$ is true.
• The body frame (rotating, glued to the object): where the mass distribution — and therefore the inertia tensor — is constant in time.

A vector $\vec{A}$ has time derivatives in the two frames related by the kinematic transport equation, the foundation for everything that follows:

The strategy throughout is: write the physics (torque = rate of change of $\vec{L}$) in the space frame, but evaluate the inertia in the body frame where it is constant, and bridge the two with this equation.

The Moment-of-Inertia Tensor

Why a Single Number Is Not Enough

For rotation about a fixed axis, the elementary $L = I\omega$ of the Newtonian chapter is all you need: $I = \sum_i m_i r_{\perp i}^2$ is a single scalar. But a freely tumbling body has no fixed axis. The angular momentum of a collection of particles is

Expanding the double cross product with $\vec{a}\times(\vec{b}\times\vec{c}) = \vec{b}(\vec{a}\cdot\vec{c}) - \vec{c}(\vec{a}\cdot\vec{b})$ gives $\vec{L} = \sum_i m_i\left[ r_i^2\,\vec{\omega} - \vec{r}_i(\vec{r}_i\cdot\vec{\omega})\right]$. This is linear in $\vec{\omega}$, but the second term mixes the components: $L_x$ depends not only on $\omega_x$ but on $\omega_y$ and $\omega_z$ too. The proportionality “constant” is therefore a $3\times 3$ matrix, the inertia tensor $\mathbf{I}$:

Components of the Tensor

For a continuous body of density $\rho(\vec{r})$, the diagonal moments of inertia and off-diagonal products of inertia are

with the other components following by cyclic permutation. Compactly, with $\delta_{jk}$ the Kronecker delta and indices running over $x, y, z$,

The tensor is real and symmetric ($I_{jk} = I_{kj}$), a fact with enormous consequences: any real symmetric matrix can be diagonalized by an orthogonal transformation, and its eigenvalues are real.

Rotational Kinetic Energy

The same tensor governs energy. Summing $\tfrac{1}{2}m_i v_i^2$ with $\vec{v}_i = \vec{\omega}\times\vec{r}_i$ yields a clean quadratic form:

Note that $T_{\text{rot}} = \tfrac{1}{2}\vec{\omega}\cdot\vec{L}$ is not $\tfrac{1}{2}L^2/I$ in general, precisely because $\vec{L}$ and $\vec{\omega}$ are not parallel.

Two Indispensable Theorems

Parallel-axis (Steiner) theorem. The inertia tensor about any point $P$ displaced by $\vec{d}$ from the center of mass is

This lets you tabulate inertia about the center of mass once, then shift to any pivot.

Perpendicular-axis theorem (for flat laminae in the $xy$-plane): $I_{zz} = I_{xx} + I_{yy}$, because every mass element has $z = 0$.

Principal Axes

Diagonalizing the Tensor

Because $\mathbf{I}$ is real and symmetric, there exists an orthonormal set of body-fixed axes — the principal axes $\hat{e}_1, \hat{e}_2, \hat{e}_3$ — in which the tensor is diagonal:

The eigenvalues $I_1, I_2, I_3$ are the principal moments of inertia, and the eigenvectors are the principal axes. In this frame the messy vector relation collapses to three decoupled scalar equations:

The geometric meaning is sharp: $\vec{L}$ is parallel to $\vec{\omega}$ if and only if the body spins about a principal axis. Spin about any other direction and the two vectors splay apart, which is the seed of every wobble and precession.

Symmetry Finds the Axes for You

You rarely diagonalize a matrix by hand. Any axis of geometric symmetry is automatically a principal axis, and any plane of symmetry contains two of them. A cube about its center has $I_1 = I_2 = I_3$ (a “spherical top”); a cylinder or any body of revolution has $I_1 = I_2 \neq I_3$ (a “symmetric top”); a generic brick has all three distinct (an “asymmetric top”). This classification drives the rest of the chapter.

The inertia ellipsoid. Plotting $\vec{\omega}^{\mathsf{T}}\mathbf{I}\,\vec{\omega} = 1$ traces an ellipsoid whose semi-axes lie along the principal axes with lengths $1/\sqrt{I_k}$. Poinsot showed that torque-free motion is exactly this ellipsoid rolling without slipping on a fixed plane (the “invariable plane” perpendicular to the conserved $\vec{L}$) — a purely geometric picture of free rotation.

Euler’s Equations of Motion

Torque in the Rotating Frame

The physical law is $\vec{\tau} = (d\vec{L}/dt)_{\text{space}}$. But $\mathbf{I}$ is only constant in the body frame, so we apply the transport equation to $\vec{L}$:

Resolving along the principal axes (where $L_k = I_k\omega_k$ and the $I_k$ are constants) gives Euler’s equations, the rotational analogue of $\vec{F}=m\vec{a}$ for a rigid body:

The nonlinear cross terms — present even with zero torque — are the entire source of the rich, sometimes counterintuitive, dynamics. Notice they vanish when two of the moments are equal or when motion is purely about one principal axis.

Article: Euler's Equations (Rigid Body Dynamics) - Wikipedia

The Two Conserved Quantities of Free Motion

For torque-free motion ($\vec{\tau}=0$) two scalars are conserved and tightly constrain the dynamics:

In angular-momentum space these are two surfaces: a sphere of radius $L$ and an ellipsoid of fixed energy. The motion of $\vec{L}$ (as seen in the body) is confined to their intersection curves — the polhodes. This geometric fact alone predicts which spins are stable, as we will see in the tennis-racket theorem.

Euler Angles: Parametrizing Orientation

Three Angles, Three Rotations

To connect the body frame to the space frame we need three independent numbers for orientation. The standard choice is the Euler angles $(\phi, \theta, \psi)$, built from three successive rotations:

1. Rotate by $\phi$ about the space $z$-axis (precession).
2. Rotate by $\theta$ about the new $x$-axis, the line of nodes (nutation).
3. Rotate by $\psi$ about the new $z$-axis, the body symmetry axis (spin).

Article: Euler Angles - Wikipedia

Angular Velocity in Euler Angles

The three angular rates $\dot\phi, \dot\theta, \dot\psi$ are not orthogonal, so resolving $\vec{\omega}$ onto the body’s principal axes for a symmetric top ($I_1=I_2$) takes some care. The result is

These plug into the Lagrangian $L = T_{\text{rot}} - V$ and make heavy-top problems tractable in the Lagrangian formulation, where the cyclic coordinates $\phi$ and $\psi$ immediately hand you two conserved momenta.

Gimbal lock. When $\theta = 0$, the precession and spin axes coincide and $\phi$ and $\psi$ become indistinguishable — the parametrization is singular. This gimbal lock is why aerospace and graphics code prefers quaternions or rotation matrices, which have no such coordinate singularity. Euler angles remain unbeatable for analytic insight, though.

The Symmetric Top

Free Symmetric Top: Steady Precession

Take a torque-free symmetric body, $I_1 = I_2 \equiv I_\perp$ and $I_3$ the symmetry axis. The third Euler equation gives $I_3\dot\omega_3 = 0$, so the spin component $\omega_3$ along the symmetry axis is constant. The remaining two equations become a simple linear oscillator:

So $(\omega_1, \omega_2)$ rotates in a circle at angular rate $\Omega$: in the body frame, $\vec\omega$ traces a cone about the symmetry axis (the body cone). Meanwhile, viewed from space, the symmetry axis and $\vec\omega$ both precess around the fixed $\vec{L}$ (the space cone). This free precession is exactly what makes a wobbling thrown frisbee or a misshot American football “spiral wobble.” For Earth, the analogous effect is the Chandler wobble, a ~433-day free precession of the rotation axis.

Heavy Symmetric Top: Precession and Nutation

Now stand the top on a fixed pivot under gravity — the classic spinning top or gyroscope. Gravity exerts torque $\vec\tau = \vec{r}{\text{cm}}\times M\vec{g}$, perpendicular to both the symmetry axis and the vertical. Using the Lagrangian in Euler angles, $\phi$ and $\psi$ are cyclic, yielding two conserved momenta $p\phi$ and $p_\psi$ plus energy. The tilt angle $\theta$ then obeys a one-dimensional effective-potential problem identical in spirit to the orbital problem of the Newtonian chapter:

The motion splits into two pieces:

• Precession — the slow circling of the symmetry axis about the vertical at rate $\dot\phi$. For a fast top, the steady precession rate is approximately $\dot\phi \approx Mg\ell/(I_3\omega_3)$: spin faster and the top precesses slower.
• Nutation — a superimposed nodding of $\theta$ between two turning points where $\dot\theta = 0$, as the axis bobs up and down while precessing.

Whether you see a smooth precession, a cusped trochoidal path, or looping nutation depends entirely on the initial release conditions, all encoded in $V_{\text{eff}}(\theta)$.

Article: Precession - Wikipedia

Video: Gyroscopic Precession Explained

The Intuition Behind Gyroscopic Precession

Why does a gyroscope defy gravity and circle instead of toppling? Because $\vec\tau = d\vec{L}/dt$ is a vector equation. Gravity’s torque is horizontal and perpendicular to $\vec{L}$, so it does not change $\vec L$’s magnitude — it only swings its direction sideways. The tip of $\vec L$ chases the torque around a circle. The faster the spin (larger $\vec L$), the smaller the fractional change $d\vec L/L$ a given torque produces, hence the slower precession. This is the same physics that keeps a bicycle upright and makes the equinoxes precess over 26,000 years as the Sun and Moon torque Earth’s equatorial bulge.

The Asymmetric Top and the Tennis-Racket Theorem

Stability of Free Rotation

When all three moments differ ($I_1 < I_2 < I_3$), the cross terms in Euler’s equations never vanish and the dynamics is genuinely nonlinear. The key question: spin the body about a principal axis and give it a tiny nudge — does it keep spinning steadily or tumble? Linearizing Euler’s torque-free equations about steady rotation $\vec\omega = \omega\hat{e}_k$ gives the small-perturbation growth rate. The algebra delivers a clean verdict:

The Tennis-Racket (Intermediate Axis) Theorem

Rotation about the axis of largest or smallest moment of inertia is stable; rotation about the intermediate axis is unstable. Toss a tennis racket, a book, or a phone, trying to spin it about the intermediate axis, and it executes a half-flip — the tennis-racket theorem, also called the Dzhanibekov effect after the cosmonaut who noticed a wing-nut spontaneously flipping in orbit.

The geometric explanation is the polhode picture from earlier: on the intersection of the energy ellipsoid and the angular-momentum sphere, the curves near the intermediate axis are hyperbolic (saddle-like), so the tip of $\vec\omega$ races away along the unstable manifold and back, producing the periodic flips. Near the extreme axes the curves are small closed loops, so the spin merely jitters.

Worked example — stability of the spinning book. Consider a thin rectangular book with principal moments $I_1 < I_2 < I_3$, thrown with zero external torque so Euler’s equations are homogeneous. Spin it primarily about axis 1 with a small wobble: $\omega_1 \approx \omega_0$ (large, nearly constant) and $\omega_2, \omega_3$ tiny. Drop the products of the small quantities and use $\dot\omega_1 \approx 0$:

Differentiate the first and substitute the second to decouple them:

Rolling Constraints

Rolling Without Slipping

Many rigid-body problems — a wheel, a ball on a table, a coin spinning to rest — add a constraint linking translation and rotation at the contact point. Rolling without slipping demands that the contact point have zero instantaneous velocity:

For a wheel of radius $R$ this reduces to the familiar $v_{\text{cm}} = R\omega$, with acceleration $a_{\text{cm}} = R\alpha$. The constraint is what lets a ball convert translational into rotational kinetic energy as it rolls down an incline.

Holonomic vs Nonholonomic Constraints

The distinction matters for which machinery from the Lagrangian chapter applies:

• A wheel rolling in a straight line has a holonomic constraint — $x = R\phi$ integrates to a relation among coordinates alone, so you can eliminate a coordinate outright.
• A ball or a coin free to steer has a nonholonomic constraint: the rolling condition links velocities in a way that cannot be integrated to a relation among coordinates. The accessible configuration space is larger than the instantaneous freedom — which is exactly why you can parallel-park a ball into any position and orientation by rolling, and why a coin traces beautiful curves as it spins down. Nonholonomic systems require Lagrange multipliers or the Euler-Lagrange equations with constraint forces, not naive coordinate elimination.

Worked example — race down the incline. A solid sphere ($I = \tfrac{2}{5}MR^2$) and a hoop ($I = MR^2$) roll without slipping from rest down a ramp of height $h$. Which wins? Energy conservation splits the released potential energy between translation and rotation, and the rolling constraint $v = R\omega$ ties them together:

Solving for the bottom speed, $v = \sqrt{\dfrac{2gh}{1 + I/MR^2}}$. The factor $I/MR^2$ is $\tfrac{2}{5}$ for the sphere and $1$ for the hoop, so the sphere arrives at $v = \sqrt{10gh/7}$ and the hoop at the slower $v = \sqrt{gh}$. The sphere wins, and the answer is independent of mass and radius — only the shape (how mass is distributed relative to the axis) decides the race. The hoop loses because it must invest a larger fraction of its energy in spinning its rim.

Putting It Together: A Computational View

Euler’s equations are a clean nonlinear ODE system, ideal for numerical integration with the methods of the Chaos and Computation chapter. The free asymmetric top is a good test bed for watching the tennis-racket flips emerge from the equations themselves:

import numpy as np
from scipy.integrate import solve_ivp

# Principal moments: I1 < I2 < I3  (axis 2 is the unstable intermediate axis)
I1, I2, I3 = 1.0, 2.0, 3.0

def euler_eqs(t, w):
    """Torque-free Euler equations in the principal-axis (body) frame."""
    w1, w2, w3 = w
    dw1 = (I2 - I3) * w2 * w3 / I1
    dw2 = (I3 - I1) * w3 * w1 / I2
    dw3 = (I1 - I2) * w1 * w2 / I3
    return [dw1, dw2, dw3]

# Spin almost purely about the INTERMEDIATE axis, with a tiny perturbation
w0 = [0.01, 1.0, 0.01]
sol = solve_ivp(euler_eqs, [0, 60], w0, rtol=1e-9, atol=1e-12, dense_output=True)

# w2 periodically collapses and reverses sign: the Dzhanibekov flips.
# Conserved checks (should stay constant to integrator tolerance):
w = sol.y
L2 = (I1*w[0])**2 + (I2*w[1])**2 + (I3*w[2])**2   # |L|^2
E  = I1*w[0]**2 + I2*w[1]**2 + I3*w[2]**2          # 2T
print("spread in |L|^2:", L2.max() - L2.min())
print("spread in 2T:   ", E.max() - E.min())

Key Takeaways

• Inertia is a tensor. A single number $I$ suffices only for fixed-axis rotation. In general $\vec{L} = \mathbf{I}\vec\omega$, and $\vec L$ need not be parallel to $\vec\omega$.
• Principal axes simplify everything. The symmetric tensor diagonalizes along principal axes, where $L_k = I_k\omega_k$ and $\vec L \parallel \vec\omega$ only when spinning about one of them.
• Euler’s equations govern the spin. The nonlinear cross terms $(I_j - I_k)\omega_j\omega_k$, present even at zero torque, drive precession, nutation, and tumbling.
• Gyroscopes precess, not topple. Because torque changes the direction of $\vec L$, a fast spin precesses slowly — the principle behind gyrocompasses and the precession of the equinoxes.
• The intermediate axis is unstable. Free rotation is stable about the largest and smallest moments but unstable about the middle one — the tennis-racket / Dzhanibekov theorem.
• Rolling can be nonholonomic. A steering ball or coin obeys a velocity constraint that does not integrate to a coordinate relation, expanding its reachable configurations.

See Also

• Newtonian Mechanics & Conservation Laws — the force-based foundation, fixed-axis $L = I\omega$, torque, and the conservation of angular momentum this chapter generalizes.
• Lagrangian & Hamiltonian Mechanics — the energy method that tames the heavy top via Euler angles and cyclic coordinates, and handles rolling constraints with Lagrange multipliers.
• Chaos & Nonlinear Dynamics — numerical integration of Euler’s equations and the nonlinear dynamics of tumbling bodies.
• Classical Mechanics Hub — browse all classical mechanics topics.