Throw a book into the air. Spin it about its shortest axis (like a frisbee) and it spins cleanly. Spin it about its longest axis and it also spins cleanly. But try to spin it about the middle axis — end over end while also rolling — and it refuses: it tumbles, flips, and catches itself, over and over, no matter how carefully you launch it. This is a real, repeatable, slightly eerie fact about rotating bodies, and by the end of this page you will understand exactly why it happens. Getting there means meeting the machinery of rigid rotation: angular velocity, the inertia tensor, its principal axes, and Euler's equations.

Rotation is where the neat one-line formulas of point-mass mechanics grow up. Mass becomes a tensor, angular velocity and angular momentum stop pointing the same way, and the equations of motion pick up strange cross-terms. Every bit of it is needed to explain the tumbling book.

Angular velocity and the inertia tensor

A rigid body's motion splits into the translation of its centre of mass plus a rotation about it. That rotation is captured by a single vector, the angular velocity \boldsymbol\omega: its direction is the instantaneous axis of spin and its length is the spin rate. Every particle of the body at position \mathbf{r} moves with velocity \mathbf{v} = \boldsymbol\omega \times \mathbf{r}.

Add up the angular momentum of all those particles and something surprising happens. In one dimension we would write L = I\omega with a single number I. In three dimensions the relationship between \boldsymbol\omega and the total angular momentum \mathbf{L} is a matrix — the inertia tensor \mathbf{I}:

The diagonal entries I_{xx}, I_{yy}, I_{zz} are the familiar moments of inertia about each axis; the off-diagonal entries (the products of inertia) measure how lopsided the mass distribution is. The rotational kinetic energy is the natural quadratic form,

The startling consequence: because \mathbf{I} is a matrix, \mathbf{L} is not generally parallel to \boldsymbol\omega. The body can be spinning about one axis while its angular momentum points somewhere else entirely.

Picture it: why L tilts away from ω

Take a flat elliptical plate that is "harder to spin" about one axis than the other (I_2 > I_1). Point the spin axis \boldsymbol\omega somewhere in between and watch the angular momentum \mathbf{L} = \mathbf{I}\boldsymbol\omega: the inertia tensor stretches the component along the "heavy" axis more, so L swings toward the axis of greater inertia and no longer lines up with \boldsymbol\omega.

Slide \boldsymbol\omega until it lies exactly along a principal axis and the two arrows snap into alignment. Those special directions where \mathbf{L}\parallel\boldsymbol\omega are the principal axes, and they are nothing more than the eigenvectors of the inertia tensor.

Principal axes: diagonalising the inertia tensor

Because the inertia tensor is real and symmetric, it can always be diagonalised: there exists a set of three mutually perpendicular principal axes fixed in the body such that, in that frame, the matrix is diagonal:

In the principal frame the kinetic energy and angular momentum simplify to T = \tfrac12(I_1\omega_1^2 + I_2\omega_2^2 + I_3\omega_3^2) and \mathbf{L} = (I_1\omega_1, I_2\omega_2, I_3\omega_3). This is exactly the payoff of diagonalisation: choose the body's own natural axes and the tangled tensor becomes three tidy numbers.

Euler's equations and the tumbling book

Newton's law for rotation is \dot{\mathbf{L}} = \boldsymbol\tau (torque equals rate of change of angular momentum). But \mathbf{L} is simplest in the body frame, which is itself rotating, so converting to that frame adds a \boldsymbol\omega\times\mathbf{L} term. In the principal-axis body frame, with no external torque, this gives Euler's equations:

Now the book. Order the moments I_1 < I_2 < I_3. Spin almost purely about axis 1 (smallest) or axis 3 (largest) and a stability analysis of Euler's equations shows tiny disturbances stay tiny — the spin is stable. But spin about axis 2, the intermediate axis, and the same analysis gives an exponentially growing wobble: the body flips. This is the intermediate axis theorem, better known as the tennis-racket theorem (flip a racket to catch it and it also does a half-twist about its handle). It is a purely mathematical instability baked into Euler's equations — the exact analogue of a negative eigenvalue in the small-oscillation problem.