Euler's Disk

The whirr that won't quit

Spin a coin on a table and listen. For a second it spins quietly on its edge, then it tips over a little and starts to roll on its rim. As it settles, the rolling sound rises in pitch — faster, faster, into a frantic buzz — and then, abruptly, it slaps flat and goes silent. Euler's disk is that same motion, engineered to last: a thick polished steel disk on a mirror that drags the accelerating finale out to a full minute or more.

The headline puzzle is that the disk seems to speed up as it dies. That feels like a free-energy violation. It isn't. The sound you track is the precession (the rate at which the point of contact races around the rim), not the disk's own spin. Total energy is steadily falling the whole time. But as the disk flattens, the geometry funnels whatever motion is left into an ever-faster wobble — so the audible pitch climbs even as the energy drains away.

It is named for Leonhard Euler, who first wrote the rigid-body equations for a rolling disk in the 18th century. The toy version was popularized in the 1980s, and the physics of the dramatic finale was sharpened by Keith Moffatt in a 2000 Nature paper that pinned the disappearing energy on a thin film of squeezed air.

How the motion works

A disk on a surface has three motions layered together:

• Spin — the disk rotating about its own symmetry axis.
• Roll — the rim rolling along the table without slipping, like a tilted coin.
• Precession — the contact point traveling around the table, which makes the disk's axis sweep out a cone.

Define the tilt angle α as the angle between the disk's face and the table. When α is large (≈ 90°), the disk is nearly upright and spinning. As α shrinks toward 0, the disk lies almost flat and the dominant motion is rolling/precession. The remarkable feature is that the precession rate Ω grows as the disk flattens — it scales as Ω ∝ α^−1/2. So the smaller the tilt, the faster the contact point laps the rim, and the higher the whirr's pitch.

Crucially, the disk's center of mass barely moves and barely descends until the very end. The motion is almost pure rolling, so sliding friction does very little work. That is why the disk can keep going for a minute: there is almost nothing to dissipate energy quickly — until air viscosity finishes the job.

The governing physics

Take a uniform disk of radius a, mass m, tilted at angle α to the horizontal, rolling without slipping. Moffatt's energy bookkeeping gives the total mechanical energy as a function of α, and it goes to zero as α → 0. The key relations:

The center of mass sits a height h above the table:

h = a · sin(α)

For steady rolling without slipping, the precession (rolling) angular rate Ω relates to gravity and tilt. In the small-α limit Moffatt obtained:

Ω² ≈ (4 g) / (a · α)     ⇒     Ω ∝ α^(−1/2)

So as α → 0, Ω → ∞. The total energy E (potential + kinetic) of the steadily rolling disk scales linearly with the tilt:

E(α) ≈ (3/2) · m · g · a · α        (small α)

Now add dissipation. If a thin film of air of viscosity μ is squeezed in the gap between rim and table, the rate of viscous energy loss scales steeply with the precession rate. Equating the energy-loss rate to dE/dt gives a closed equation for α(t). Integrating it yields a finite settling time t* at which α reaches zero:

α(t) ∝ (t* − t)^(1/3)        ⇒      Ω(t) ∝ (t* − t)^(−1/6)

That is the finite-time singularity: at t = t* the tilt is exactly zero and the precession rate has formally diverged. In Moffatt's worked example, the rolling rate reaches the order of 100 Hz only in the final hundredth of a second before t*. (The exact exponents depend on the dissipation model; air-film viscosity gives the (t*−t)^1/3 collapse above, and the abrupt stop is what you hear and see.)

Spinning vs. rolling — two regimes

The disk's life splits into two phases, and confusing them is the single most common error in understanding the toy.

Quantity	Early phase (large α, upright)	Late phase (small α, near-flat)
Dominant motion	Spin about own axis	Rolling / precession of contact point
Tilt angle α	≈ 60°–90°	≈ 10° → 0°
What you hear	Quiet, steady spin	Rising whirr → buzz
Audible frequency	The spin rate (low)	The precession rate Ω (climbing as 1/√α)
Center-of-mass height	h = a·sin α (high)	Collapsing toward 0
Energy loss rate	Slow (little slip, little air shear)	Runs away near t* (air film + thin gap)

The audible pitch is the contact point's lap rate, not the spin. A common misconception is that the disk "spins faster at the end." Its spin actually does not run away; the precession does.

Numbers: the commercial disk vs. a coin

The branded Euler's disk is engineered to maximize spin-down time. Compare it with a one-euro coin spun on a tabletop:

Property	Euler's disk (toy)	€1 coin	Why it matters
Diameter	≈ 7.5 cm	≈ 2.3 cm	Larger a → larger angular momentum, longer roll
Thickness	≈ 1.3 cm	≈ 0.2 cm	Thick rim keeps mass and stability
Mass	≈ 440 g	≈ 7.5 g	More inertia = slower drain of energy
Surface	Hard concave mirror	Whatever table you have	Hard + smooth = minimal slip & deformation
Spin-down time	≈ 60–120 s	≈ 1–3 s	The toy stretches the same physics ~50×
Peak audible whirr	passes ~hundreds of Hz	a quick rising buzz	Ω ∝ 1/√α climbs as the disk flattens

The mirror's slight concavity is not decoration: it gently recenters the disk so the contact point doesn't wander off the edge, letting all that angular momentum bleed off slowly instead of being lost to a wobble that walks the disk away.

Where the energy actually goes

For decades the obvious culprits were rolling friction and micro-slip at the contact. Moffatt's 2000 analysis showed those are too weak to match the observed, abrupt finale on a polished surface — the disk's motion is too close to pure rolling. Instead he proposed the dominant brake is viscous dissipation in the thin layer of air being repeatedly squeezed and pumped in the narrowing gap between the rim and the table as the disk precesses.

Two pieces of evidence support this:

• Vacuum test. Run the disk in a partial vacuum and the spin-down time lengthens markedly — remove the air and you remove the main energy sink.
• The collapse exponent. The air-film model predicts α(t) ∝ (t*−t)^1/3, a steep final collapse, matching the suddenness of the real stop better than a friction-only model, which predicts a gentler decay.

Later work (e.g. van den Engh et al., and follow-ups to Moffatt) argued that rolling friction and adhesion also contribute and can dominate on rougher surfaces, so the precise mix depends on the materials. The qualitative picture — a near-pure-rolling disk drained slowly by air, ending in a fast collapse — is robust.

Where this shows up

• Coins, washers, and lids. Every spun coin, dropped bottle cap, or rattling pot lid runs through the same spin-then-rolling-precession sequence — Euler's disk is just the cleanest, longest-lived version.
• Rolling-without-slipping dynamics. The disk is a textbook nonholonomic system: the no-slip constraint can't be integrated into a position constraint, which is exactly what makes the math rich and the motion counter-intuitive.
• Finite-time singularities. The Ω → ∞ behavior is a clean, tabletop example of a real physical quantity formally diverging in finite time — the same mathematical flavor that appears in fluid-jet pinch-off, bubble collapse, and certain astrophysical models.
• Granular and bearing physics. Squeezed-film air damping between a moving surface and a substrate is a real engineering effect in hard-disk drive heads, MEMS resonators, and precision bearings; the disk is a vivid demonstration of it.
• Teaching gyroscopic motion. Because spin, roll, and precession are simultaneously visible, the disk is a favorite for introducing precession and angular-momentum bookkeeping without abstract diagrams.

Common misconceptions and edge cases

• "It's gaining energy at the end." No. Total energy falls monotonically. Only the precession rate rises, because the geometry concentrates the dwindling energy into a faster wobble (Ω ∝ 1/√α).
• "The disk spins faster as it dies." The thing accelerating is the rolling/precession of the contact point, not the disk's own spin. The whirr you hear is the contact point lapping the rim.
• "Friction stops it." On a polished surface, sliding friction does almost no work because the disk nearly rolls without slipping. Air-film viscosity (and on rough surfaces, rolling friction/adhesion) is what removes the energy.
• "It's a true singularity." The idealized model predicts Ω → ∞ in finite time, but physics intervenes: the air-film theory breaks down, the disk can momentarily bounce, and surface roughness sets a floor. The stop is extremely fast, not literally infinite.
• "Any flat disk works as well." Mass, hardness, smoothness, and a slightly concave base matter enormously. A light, thin, rough disk on a soft surface stops in a second; the engineered combination is what produces the minute-long, dramatic finale.
• "The mirror is just for looks." Its concavity recenters the disk so it doesn't walk off the surface, and its hardness minimizes surface deformation — both essential to the long spin-down.