Supercooled liquids approaching the glass transition slow by orders of magnitude over narrow temperature ranges — a phenomenon that has resisted a unified explanation for decades. The dominant theoretical tradition attributes this slowdown to growing cooperative length scales: larger and larger regions must rearrange simultaneously for the liquid to flow. But Pica Ciamarra, Dyre, Lerner, and Wyart marshal evidence for an alternative picture in which localized excitations and their spectrum — not a thermodynamic length scale — govern the dynamics.
The framework connects global elastic properties, local elastic moduli, the Debye-Waller factor, and excitation densities into a quantitative chain. Fragility — the steepness of the viscosity-temperature relationship — emerges from how the excitation spectrum evolves with cooling, not from how large the cooperative regions become. Parameter-free predictions for activation energy follow. Dynamical heterogeneities, the spatially uneven relaxation observed in all glass-formers, arise through thermal avalanches rather than cooperative rearrangements.
The structural claim is pointed: what looks like a collective phenomenon may be a statistical one. If the excitation spectrum — the distribution of local energy barriers — controls the macroscopic behavior, then the glass transition is not about growing correlations but about the progressive rarefaction of accessible pathways. The liquid doesn't freeze because regions become too large to rearrange. It freezes because the excitations that enable rearrangement become too rare to find.
(arXiv:2603.05209)