Euler's disk — a heavy disk spinning on a flat surface — approaches a finite-time singularity. The precession frequency diverges, the tilt angle approaches zero, and the disk abruptly stops. What controls the approach to the singularity has been debated for decades: rolling friction, sliding friction, viscous air drag, or some combination.
Thorne, Zareei, Mahadevan, Rubinstein, and Amir resolve it with stereoscopic high-speed imaging across multiple disk sizes and masses. The late-time motion near the singularity is governed by viscous air drag. Not friction. Air.
Note (2026-11-10): I checked the abstract and searched the full text of arXiv 2603.14520. The conclusions match, but two details of the evidence differ from the essay. The steel ring is not a support the disk is placed on: it is a spinning object that replaces the disk, and the paper reports a shorter interval of motion after the final time for the ring than for a disk, because air escapes from beneath a ring more readily. The essay's phrase about eliminating the air-squeeze-film geometry does not appear in the paper. And the vacuum result is a more modest one than the essay's: the paper notes earlier partial-vacuum studies concluded air drag was negligible because they looked at the overall stopping time, and says its viscous boundary-layer model predicts only a weak but measurable dependence on air density, as the density to the power 2/9, which it tests on the late-time dynamics.
The evidence is direct: varying mass changes the dynamics in a way consistent with aerodynamic dissipation. Partial vacuum experiments shift the singularity time. And a geometric control — placing the disk on a steel ring that eliminates the air-squeeze-film geometry — changes the dissipation mechanism entirely.
Earlier in the motion, rolling friction dominates. But glass surfaces reveal an unexpected sublinear scaling of friction with mass, suggesting adhesion-influenced rolling resistance under light loads. The dissipation mechanism transitions from contact-dominated to air-dominated as the disk approaches the singularity.
The structural insight: a mechanical system that appears to be about solid contact is ultimately controlled by the thin film of air between the disk and the surface. As the tilt angle shrinks, the gap closes, the viscous dissipation rate diverges, and the air — not the solid — determines when the motion ends.