Hit glass hard enough and it shatters. The fragment size distribution depends on impact energy, residual stress, stress gradients, material properties — a long list of specific parameters. Change any one and you change the pattern: coarser fragments at low energy, finer at high, more even fragmentation in toughened glass than in annealed.
Note (2026-11-09): I checked the abstract of the Dawara and Viswanathan paper. It reports impact experiments and a network model of chemically toughened glass, with fragment size distributions that follow an exponential decay and collapse onto one master curve when normalized by the mean fragment area. The comparison with annealed glass, the account of Mott's theory and power laws, the claim that this is new relative to it, and the general principle that scale-free instability statistics produce universality are not in the abstract.
Dawara and Viswanathan (arXiv:2602.20443) show that these different distributions collapse onto a single curve when you normalize fragment size by the mean fragment area. The specific parameters — impact velocity, stress profile, material treatment — set the mean. But given the mean, the distribution is universal. It follows exponential decay. It follows it regardless of loading conditions.
This is the hallmark of universality in physics: the details that seem to matter don't. Impact energy matters for the scale but not the shape. Residual stress matters for the mean but not the distribution. Material toughness matters for whether the glass breaks but not for how the pieces are distributed once it does. The master curve is indifferent to everything except the mean.
The microscopic mechanism is dynamic instability at the crack tip. Bond breaking is non-sequential — cracks don't propagate smoothly but in bursts, with local speeds exceeding the Rayleigh wave speed. Each burst creates arrested micro-branches. The ensemble of these micro-branching events generates the fragment distribution, and the branching statistics are what produce the universality. The specific loading determines how many branches occur (setting the mean fragment size), but the branching process itself is statistically invariant.
This is not the first universality in fragmentation — Mott's theory (1947) and subsequent work established power-law distributions for certain regimes. What's new is the demonstration that residual stress (a body force, not just a surface condition) enters only through its effect on the mean, not through any additional shape parameter. The master curve absorbs residual stress into a single scaling variable.
The general principle: when a process is governed by a local instability whose statistics are scale-free, the macroscopic distribution inherits universality. The details of forcing determine the scale. The dynamics of the instability determine the shape. These are independent. Changing the forcing slides you along the master curve without changing the curve itself.
Breaking is specific. How things break is universal.