The mass, and where it sits, which is what decides the race

Roll a hoop and a solid ball down the same slope from the same line at the same moment. The ball arrives first. Make the hoop lighter and it still loses. Make it smaller and it still loses. Make it out of lead and the ball out of wood, and the ball still wins, by exactly the same margin as before.

The previous rung established that a force’s turning effect depends on where its line of action passes. That is only half of a mechanics of rotation. The other half is the question this rung answers: given a torque, how fast does the thing actually turn? For straight-line motion the answer is Newton’s second law, and the property of the object that appears in it is its mass. For rotation there is an analogous law with an analogous property, and the property is not the mass.

The quantity that replaces mass #

The rotational equation of motion is

$$\tau = I\alpha,$$

with $\alpha$ the angular acceleration and $I$ the moment of inertia. Its definition is a sum over the body:

$$I = \sum_i m_i r_i^2,$$

each piece of mass weighted by the square of its distance from the axis.

That squared distance is the whole content. Mass counts, but distance counts twice over, so a kilogram at the rim of a wheel resists turning four times as hard as a kilogram halfway out and infinitely harder than a kilogram on the axis. Which is why the answer to “how hard is this to spin” is not a property of the object alone: it is a property of the object and a chosen axis, exactly as a torque is a property of a force and a chosen point.

For simple shapes about their symmetry axes the sum can be done once and quoted. It is conventional, and much more informative, to write the result as

$$I = kMR^2,$$

which pulls out the mass and the size and leaves $k$, a pure number that says nothing except how the mass is distributed.

A hoop’s $k$ is exactly one because every part of it is at the full radius, which makes the top row a definition rather than a calculation. Everything below it is that number reduced by the mass that sits closer in.

Why mass and radius cancel #

The race is decided by $k$ alone, and the way $M$ and $R$ disappear is worth doing slowly, because it is the reason the result feels wrong before it is derived.

A body released from a height $h$ arrives at the bottom having converted potential energy $Mgh$ into motion. If it rolls without slipping, that motion is in two places at once: the whole body moves at speed $v$, and it also spins at rate $\omega = v/R$, because rolling means the contact point does not slide.

$$Mgh = \tfrac12 Mv^2 + \tfrac12 I\omega^2 = \tfrac12 Mv^2 + \tfrac12 (kMR^2)\left(\frac{v}{R}\right)^2 = \tfrac12 Mv^2(1+k).$$

The $R^2$ from the moment of inertia is cancelled by the $1/R^2$ from the rolling condition, and the $M$ appears on both sides. What survives is

$$v = \sqrt{\frac{2gh}{1+k}},$$

in which the only property of the body is $k$.

Read that way, the result stops being surprising. Every body has the same energy per kilogram at the bottom. The ones that arrive slowest are the ones that spent the most of it turning, and the ones that spent the most on turning are the ones with their mass furthest out. The race is a competition to avoid rotating, and the shape decides how much of the prize each body has to give up.

The acceleration, and the slope that does not matter #

The same cancellation runs through the dynamics. A body rolling down a slope of angle $\theta$ has

$$a = \frac{g\sin\theta}{1+k},$$

which is the sliding answer $g\sin\theta$ from the inclined plane divided by the same factor.

The weight component along the surface is the same for a rolling body as for a sliding one, so the free-body diagram is unchanged. What differs is where the energy goes: a sliding body puts all of it into translation and a rolling one splits it, and the split is fixed by how far the mass sits from the axis. The forces are identical and the accelerations are not.

Two consequences follow, and both are testable on a kitchen table.

The slope changes every arrival time and no finishing order. The factor $1/(1+k)$ multiplies the whole acceleration, so making the slope gentler slows every body in the same proportion. The ratio of the sphere’s acceleration to the hoop’s is $(1+1)/(1+0.4) = 1.43$ on any slope whatever.

A sliding block beats everything that rolls. With $k = 0$ the divisor is one and the acceleration is the full $g\sin\theta$. A frictionless block on the same slope outruns every rolling body, because it has nothing to spin up. That is the useful edge case: rolling is not a way of going faster, it is a way of going without slipping, and it is paid for out of the same energy budget.

Where the torque comes in #

The energy argument gives the answer without mentioning torque at all, which is efficient and slightly unsatisfying, because it leaves out the mechanism. The force that actually spins the body up is friction at the contact point, and it is the only force with a moment arm about the centre of mass.

Friction acts at the contact point, a distance $R$ from the centre and perpendicular to the line joining them, so it exerts a torque $fR$ — and that torque is what spins the body up. Its size is not $\mu N$: it takes whatever value keeps the contact point from slipping, so it is an unknown to be solved for alongside the acceleration rather than a number to be looked up.

Gravity acts at the centre of mass, so its line passes through the axis and its torque about that axis is zero. The normal force points at the centre too. Only friction is offset, and $\tau = fR = I\alpha$ is what converts it into spin.

Two things worth extracting from that. The first is that a frictionless slope produces no rotation whatever — a ball placed on ice slides down without turning, which is easy to say and hard to picture. The second is that this friction does no work: the contact point is instantaneously at rest, so the force acts through no displacement. Static friction is what makes rolling possible and is not what makes it slow, and the two claims are often confused.

What the figure cannot show #

The picture draws a spoke on each body, turned by the distance rolled divided by the radius, and that is exactly what rolling without slipping means. What the picture cannot show is the assumption underneath.

Rolling without slipping is a constraint, not a force law. It says the contact point has zero velocity, and it holds only while static friction is able to supply whatever force the constraint demands. That demand rises with the slope: the required friction is $f = kMg\sin\theta/(1+k)$, while the maximum available is $\mu_s Mg\cos\theta$. Set them equal and the slope at which rolling fails is

$$\tan\theta_{\max} = \frac{\mu_s(1+k)}{k}.$$

Above it the body rolls and slides, the two speeds decouple, and every equation on this page stops applying. For a solid sphere on a surface with $\mu_s = 0.5$, that limit is about $74°$, which is why the effect is rarely met with a ball and readily met with a hoop on a polished floor, where $k = 1$ gives a much lower ceiling.

The other unstated assumption is that the bodies are rigid and the contact is a point. A real tyre flattens where it meets the road, the contact is an area, and the deformation is not perfectly recovered as the wheel turns — the energy loss is rolling resistance, it is not friction in the sliding sense, and it is what actually stops a wheel that is left to run.

The same number, doing a different job #

The moment of inertia was introduced as the thing that resists a torque, which makes it sound like a nuisance. It becomes an asset the moment the object is meant to keep turning rather than to start.

A flywheel is a moment of inertia bought deliberately: mass placed as far from the axis as the material will survive, so that the stored energy $\tfrac12 I\omega^2$ is as large as possible. A potter’s wheel, an engine’s crankshaft and a grid-scale energy store are the same device at three sizes, and all three put the mass at the rim for the reason the hoop loses the race.

The same quantity also decides an oscillation. A pendulum that is not a point mass on a string — a swinging rod, a clock’s balance wheel, a ship rolling in a swell — has a period set by the ratio of its moment of inertia to its restoring torque, and the small-angle result becomes

$$T = 2\pi\sqrt{\frac{I}{mgd}},$$

with $d$ the distance from the pivot to the centre of mass. The simple pendulum is the special case $I = md^2$, which recovers $T = 2\pi\sqrt{d/g}$ and explains why the amplitude-independence argument survives the generalisation untouched: nothing in it depended on the mass being at a point.

For a body with extent rather than a point mass, the kinetic half of an energy exchange is split between moving and spinning — in a ratio fixed by the same number. That is where the moment of inertia turns up doing a different job: it decides how a rolling body descends a slope, and it decides the period of anything that swings, and both are the same statement about how much of the motion is rotation.

Reading the arrangement off the arrival time #

Because $k$ is the only property of the body that survives, the race runs backwards as a measurement. Time a body over a measured distance on a measured slope, extract its acceleration from $s = \tfrac12 at^2$, and

$$k = \frac{g\sin\theta}{a} - 1$$

reports how its mass is distributed — without opening it, weighing it, or knowing its radius.

That is a genuinely useful instrument, and it is the standard way the distribution of mass inside an inaccessible object is established. A sealed can of soup and a sealed can of set custard have the same mass and the same size and lose the race by different margins, because the liquid one does not fully rotate with its container: the contents slip, the effective $k$ falls, and the can arrives faster than its shape predicts. Rolling a can down a plank is a test of whether it has set.

The same inference at a much larger scale is how the interior of a planet is known. A planet’s moment of inertia factor is measured from the way its spin axis precesses and from the way it responds to being pulled on, and the number that comes back is compared with the $0.4$ a uniform sphere would give. The Earth’s is $0.3307$, well below uniform, and the deficit is the statement that the mass is concentrated toward the middle — which is the primary evidence for an iron core, arrived at without any seismology. The Moon’s is $0.3929$, barely below uniform, which says its core is small. Both numbers are the $k$ of this page, measured on an object nobody can put on a slope.

The history, and the length of a pendulum that is not a point #

The moment of inertia entered physics through a problem that looks unrelated: what length of simple pendulum keeps time with a swinging body of some awkward shape?

Huygens asked it in the 1670s because he was building clocks, and the pendulum in a clock is a rod with a bob, not an idealised mass on a weightless string. His answer defined the centre of oscillation — the point at which all the mass could be concentrated without changing the period — and locating it is exactly the calculation that later became $I = \sum m_i r_i^2$. The quantity was discovered as a correction to a clock and only afterwards recognised as the rotational counterpart of mass.

His result carries a symmetry that Kater turned into the most precise measurement of gravity available for a century: the centre of oscillation and the pivot are interchangeable. Hang the same body from its centre of oscillation and it swings with the period it had before. A pendulum with two adjustable knife edges, tuned until the periods about both agree, therefore has a known effective length — the distance between the edges — without anyone needing to know its mass, its shape or where its centre of mass is. Every awkward quantity cancels, in the same way and for the same reason that $M$ and $R$ cancel in the race.

The flywheel whose size does not matter either #

The flywheel was named above as a moment of inertia bought on purpose, which invites the conclusion that a bigger and heavier one stores more energy per kilogram. It does not, and the reason is a cancellation of the same kind as the race’s.

Store energy in a spinning rim and the energy per unit mass is $\tfrac12 v^2$, with $v$ the rim’s speed. Nothing about the radius appears — a large slow wheel and a small fast one with the same rim speed store the same energy per kilogram.

What limits the rim speed is not the bearings or the motor. It is the material. A rotating hoop has to supply its own centripetal force out of its own tension, and working that balance through gives a hoop stress of $\rho v^2$: the stress in the rim depends on the density and the speed, and again not on the radius. Setting that equal to the material’s strength gives a maximum rim speed, and substituting back gives a maximum energy per unit mass of

$$\frac{E}{m} = \frac{\sigma}{2\rho}.$$

A flywheel’s energy density is its material’s specific strength and nothing else. Not its size, not its mass, not its moment of inertia — the very quantity the device is built out of has cancelled from the answer, exactly as the mass and radius cancel from the rolling race, and for the same reason: a term proportional to $R^2$ meeting a constraint proportional to $1/R^2$.

The numbers make the design decision for anybody who has them. High-strength steel at a thousand megapascals and 7,800 kilograms per cubic metre gives about 18 watt-hours per kilogram. Carbon fibre at twice the strength and a fifth of the density gives nearly ten times as much, and composite rotors reach a hundred watt-hours a kilogram or more — comparable with a lithium battery, with a power density orders of magnitude higher and no limit on the number of cycles.

That is why every serious flywheel built since the 1980s is a filament-wound composite rim rather than a lump of steel, and why the useful figure of merit for the material is a ratio the mechanical engineer already had: specific strength.

There is one further consequence of the same expression, and it decides how such a device fails. Since the stress does not depend on the radius, a large composite rotor and a small one are equally close to their limit at the same rim speed — and both are storing a great deal of energy in a body that is trying to tear itself apart. A steel wheel that fails throws fragments; a wound composite one delaminates into fibres and resin, which is a great deal easier to contain, and the choice of material is made partly for that reason.

Weighing the arrangement #

The rolling race was offered as a way of measuring how a body’s mass is arranged without opening it. There are better instruments, and they are used routinely on objects whose inertia has to be known to a few per cent before they are allowed to fly.

The standard laboratory method is a torsional oscillation. Hang the object from a wire, twist it, and let it oscillate; the period is $2\pi\sqrt{I/\kappa}$ with $\kappa$ the wire’s torsional stiffness. That still requires knowing $\kappa$ — so in practice the measurement is made twice, once with a calibration body of computable inertia and once with the unknown, and the stiffness cancels out of the ratio of the squared periods.

For anything larger than a laboratory object the arrangement is a trifilar pendulum: a platform hung on three equal vertical wires, which twists about the vertical with a period set by the total inertia of platform plus load. Weigh the empty platform’s period, add the object, weigh it again, subtract. It handles objects weighing tonnes and needs no wire of known stiffness at all — the restoring torque comes from gravity and the geometry.

A third method uses the essay’s own pendulum result backwards. Swing the object about a knife edge, measure the period, and combine it with the mass and the measured distance to the centre of mass; the compound pendulum expression then gives the inertia about the pivot, and the parallel-axis relation transfers it to the centre.

The reason for the trouble is control. An aircraft’s or a spacecraft’s attitude-control system is designed around its inertia tensor, and a control loop tuned for the wrong number can be unstable rather than merely sluggish. So the tensor is measured — on a purpose-built swing rig, about several axes, with the off-diagonal terms extracted from the differences — before anything flies. The number that comes out is the same $k$ this page is about, for an object nobody can roll down a slope, obtained because the arrangement of the mass decides the response and nothing else does.

Where it stops being a number #

Everything above treats $I$ as a single number, which works only because every body drawn here is spun about an axis of symmetry.

In general the moment of inertia is a tensor — a three-by-three array — and the reason is visible in an object as simple as a book. Spun about its longest axis or its shortest, a thrown book turns stably; spun about the intermediate axis it tumbles chaotically, and no scalar can express that. The rotational equation of motion in three dimensions is correspondingly less friendly than $\tau = I\alpha$, and the surprising behaviour of gyroscopes and tumbling spacecraft lives there.

There is also the matter of the axis. The values of $k$ tabulated above are for rotation about the centre of mass, and the moment of inertia about any parallel axis a distance $d$ away is $I_{\text{cm}} + Md^2$ — the parallel-axis theorem, which is the reason the centre of mass is the natural place to quote the number from, in the same way and for the same reason that it is the natural frame for a collision.

Where the ladder goes next #

The immediate next rung is angular momentum, $L = I\omega$: what is conserved when no external torque acts, why a skater speeds up on pulling in, and why that is not a violation of energy conservation but a purchase paid for by the work of pulling. After it: the parallel-axis theorem in earnest; the gyroscope, where a torque produces motion at right angles to itself and every intuition built on this page fails; rolling as a constraint, and the wheel as the machine that exploits it; the tennis-racket theorem and unstable axes; and the rigid body in three dimensions, where the tensor becomes unavoidable.

The claim to carry forward is the one the race makes. A body’s resistance to being spun is not a property of how much of it there is, but of how far out it is — and because that dependence is quadratic while the rolling constraint is linear, the size and the mass cancel completely, leaving a competition that a marble wins against a hoop of any weight, on any slope, every time.