Intermediate axis instability
A Socratic walk-through of intermediate axis instability — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does a book tossed spinning about its middle axis keep flipping over when the other two spins stay steady?
Take a book, tape it shut, and toss it spinning. About its long axis, like a rolling pin, it spins cleanly. About the axis through the covers, like a frisbee, it also spins cleanly. About the remaining axis it refuses: within a turn or two it flips end over end, then flips back, and keeps doing so all the way to your hand.
Nothing touched it. No torque acted; gravity pulls on the whole body and produces none about its own centre. Three axes, the same free body, the same physics — and one of them cannot be spun about. What distinguishes the middle one?
Reasoning it through
REASONING #The distinguishing feature has to be inertial, since that is all a free rigid body has. Every body has three principal moments of inertia; call them I1 < I2 < I3, one for each of three mutually perpendicular axes. The middle axis is the one whose moment is neither greatest nor least. That is the only thing "middle" can mean here, so the answer must come out of the ordering.
For a body with no torque on it, Euler's equations govern how the three spin components trade with each other:
I1 w1' = (I2 - I3) w2 w3 I2 w2' = (I3 - I1) w3 w1 I3 w3' = (I1 - I2) w1 w2
where the primes are rates of change. Notice the structure: each axis's spin is driven by the product of the other two, scaled by a difference of moments. Spin purely about one axis and the other two are zero, so nothing changes — all three are steady states. The question is what happens to a small imperfection.
Take spin mainly about axis 2, the middle one: w2 is a large constant W, and w1 and w3 are small. Drop the product of two small quantities in the second equation, and it says w2 barely changes. The other two become
I1 w1' = (I2 - I3) W w3 and I3 w3' = (I1 - I2) W w1.
Differentiate the first and substitute the second:
w1'' = (I2 - I3)(I1 - I2) W^2 / (I1 I3) x w1.
Now count signs. Since I2 < I3, the factor (I2 - I3) is negative. Since I1 < I2, the factor (I1 - I2) is also negative. Two negatives multiply to a positive, so the whole coefficient is positive: w1'' = +k^2 w1. That equation has an exponentially growing solution. Any wobble, however small, doubles and doubles again.
Repeat for spin about axis 1, the smallest moment. The two factors become (I3 - I1), positive, and (I1 - I2), negative. Their product is negative, so w'' = -k^2 w — simple oscillation. The wobble goes round and comes back. The same happens about axis 3. So two axes give bounded wobble and one gives runaway, and the whole difference is a sign that flips only when the moment sits between the other two. Nothing was assumed about books.
How fast? The growth rate is k = W sqrt((I3 - I2)(I2 - I1) / (I1 I3)). Take a uniform book, 20 by 15 by 3 cm; its principal moments go as the sums of squared dimensions, 0.0234, 0.0409 and 0.0625. Putting those in gives k = 0.51 W, so the wobble grows by a factor e in about two radians of spin — roughly a third of a revolution — and a hundredfold within a turn and a half. That is exactly what you see: it does not creep, it goes almost immediately.
The analogy
THE ANALOGY #A ball placed on a mountain pass. Nudge it along the ridge and it rolls back to the saddle point; nudge it across the pass and it accelerates away. The same point is a minimum in one direction and a maximum in the other, so "stable" is not a property of the point but of the direction you disturb it in.
The ball has a landscape of potential energy to roll down and eventually settles somewhere lower. The tumbling book has no such landscape — it is torque-free, its energy and its angular momentum are both exactly constant, and the flip is not an escape but a redistribution of the same fixed energy among the three axes. That is why it does not run away but returns and flips again, forever.
Clarifying the model
THE MODEL #It is worth separating this sharply from precession, which the collection treats elsewhere. A spinning top precesses because an external torque acts — gravity, applied at a distance from the pivot — and the motion is stable and forced. Here there is no torque at all. The instability is internal, a property of how a free body's own spin components feed each other, and it needs the body to be asymmetric in all three directions. That is the fixed point of difference: precession is a forced response, this is a free instability.
Which gives the sharpest test. Make two moments equal — a cylinder, a disc, anything with an axis of symmetry — and there is no "middle" moment for the mechanism to work with, so such a body should tumble about no axis whatsoever. If a symmetric body flipped, the argument would be dead. A second, quantitative test: the growth rate is linear in W, so flips per revolution is fixed and flips per second should double when you toss it twice as fast. If flip rate were independent of spin rate, the derivation would be wrong.
Two honest caveats. The linearisation above describes only the start of the growth; the flip itself is fully nonlinear, and the exact torque-free motion is periodic, expressible in Jacobi elliptic functions. It is not chaotic, whatever the tumbling looks like. Second, the rigid-body result assumes no dissipation. A real body that can flex, or that carries fluid, sheds kinetic energy while its angular momentum is conserved, and the lowest-energy state at fixed angular momentum is rotation about the greatest-inertia axis. So dissipative bodies eventually settle there regardless — which is what happened to Explorer 1 in 1958, when the satellite was spun about its long axis and its flexible antennas bled off enough energy to leave it tumbling flat.
A picture of it
THE PICTURE #How to readThe three states fanning out from the start dot are the three ways to spin the same body; each is a genuine steady motion until disturbed. Follow the self-loops on the outer two: a nudge oscillates and returns, so the spin looks clean. Follow the middle one and there is no loop back to it — the nudge grows, carries the body through a half turn, and lands it in the flipped state, from which the growth starts over. The absence of a return edge into the middle state is the whole result.
What became clearer
WHAT CLEARED #Stability here is decided by a sign, and the sign is decided by an ordering. In Euler's torque-free equations each spin component is driven by the product of the other two times a difference of moments of inertia. Spin about the largest or smallest moment and those two differences carry the same sign, so a disturbance oscillates. Spin about the middle one and they carry opposite signs, the coefficient turns positive, and the disturbance grows exponentially — at a rate proportional to the spin itself, which is why a hard toss flips just as many times per turn as a gentle one.
Where to go next
ONWARD #- Why energy dissipation drives a real spacecraft toward spinning about its flattest axis, and how spin stabilisation is designed around that.
Key terms
TERMS #| Term | What it means |
|---|---|
| Principal moments of inertia | the three numbers describing how hard a body is to spin about each of three perpendicular axes through its centre of mass. |
| Euler's equations | the equations of motion for a rigid body's spin components in its own rotating frame. |
| Dzhanibekov effect | the name often given to the same flip when it is observed in weightlessness. |
Every term the collection defines is gathered in the glossary.