Dynamical arrest — when a system stops relaxing toward equilibrium despite having no apparent reason to — typically requires quenched disorder. Glasses freeze because random impurities or structural irregularities create energy barriers between configurations, trapping the system in metastable states. Remove the disorder and the system flows to equilibrium on normal timescales. The connection between randomness and arrest seems fundamental: you need something external to break ergodicity.
Kundu, Seth, Roy, Bhattacharjee, and Moessner (arXiv 2602.23362, February 2026) demonstrate dynamical arrest in a perfectly ordered system: a classical spin-3/2 ice on a pyrochlore lattice with no quenched disorder, no randomness, no impurities.
Spin ice is a frustrated magnet where spins on corner-sharing tetrahedra obey ice rules — two pointing in, two pointing out — creating an extensive ground-state degeneracy. This degeneracy supports a Coulomb phase: a spin liquid whose long-wavelength fluctuations obey the same equations as the electric field in electrostatics, complete with emergent magnetic monopoles as point-like excitations. Standard spin-1/2 ice on a pyrochlore lattice produces one type of Coulomb phase with monopoles as the only mobile excitations.
Moving to spin-3/2 enlarges the on-site Hilbert space. Each site can point in four directions rather than two, which introduces a new class of excitation beyond monopoles: crystal-field excitations that don't carry magnetic charge but do carry energy. These excitations interact with the monopoles. Crucially, some relaxation pathways that would be available in spin-1/2 ice become kinetically constrained because they require creating crystal-field excitations as intermediate steps. The excitations aren't frozen by disorder — they're frozen by the structure of the allowed moves.
After a thermal quench from high temperature into the Coulomb phase, the system relaxes quickly at first as the easy moves are executed. Then it arrests: spin autocorrelations reach a plateau that persists for exponentially long times in the inverse temperature. The plateau is athermal — it does not correspond to a thermal equilibrium at any temperature. The system has fallen into a configuration where further relaxation requires activated processes involving composite excitation structures, and the activation barriers are set by the crystal-field energy scale, not by disorder.
The arrest is a property of the dynamics, not the statics. The equilibrium Coulomb phase is perfectly well-defined — it has extensive entropy, algebraic correlations, deconfined monopoles. The system simply cannot reach it from the quenched initial condition in any reasonable time, because the pathway requires coordinated multi-spin moves that are individually unfavorable. The equilibrium exists. The kinetics cannot find it.
This is arrest without the usual excuses. No randomness. No long-range interactions to create frustration beyond nearest neighbors. No external pinning. The kinetic constraints are self-generated by the enlarged Hilbert space of the higher-spin system. The same lattice with smaller spins would relax normally. The richer local structure creates its own barriers.