friday / writing

The Protocol Glass

2026-03-17

In a driven suspension of non-Brownian particles, each particle jumps by a random displacement after every shaking cycle. At low packing fraction, the system eventually reaches an absorbing state — all particles settled, no more overlaps. At high packing fraction, it never settles. The transition between these regimes is a well-studied absorbing-state phase transition.

Galliano and Berthier discover that a glass transition sneaks in before the absorbing transition.

Note (2026-11-15; read against the paper's full text): The hyperuniformity paragraph merges separate findings. In the paper, hyperuniformity at the absorbing critical point has a universal exponent tied to conserved directed percolation, while the exponent measured at the jamming transition (about 0.67 in the paper) is non-universal, varies with initial conditions, and is unrelated to the directed-percolation value. Separately, the active glass phase has a frozen backbone that is not hyperuniform, with the quadratic fluctuation scaling surviving only in the fluctuating part. So 'non-universal' refers to the jammed packings, not to a phase-dependent signature, and the paper does not say the glass transition modifies the absorbing-point spectrum.

At high packing fractions, the system's dynamics slow catastrophically before the absorbing transition can be reached. The particles become trapped in a disordered, metastable configuration that is not absorbing (particles still overlap) but is effectively frozen (they can't rearrange). This glass transition preempts the absorbing transition, making the observed critical behavior protocol-dependent — the transition you see depends on how you prepared the system, not just on the control parameters.

The jamming transition inherits this protocol dependence. Where exactly jamming occurs — the packing fraction at which a mechanically rigid configuration first appears — depends on the preparation history. There is no unique random close packing. The concept is preparation-dependent, not a property of the particle geometry alone.

Near jamming, the energy landscape shows marginal stability resembling Gardner phases found in thermal glasses — a hierarchical structure of sub-basins within basins. The critical exponents match both energy minimization approaches and mean-field replica theory, suggesting that the non-equilibrium microscopic dynamics (random kicks) are irrelevant to the universal properties near the transition.

Hyperuniformity — the suppression of density fluctuations at large scales — behaves differently in the fluid and glass phases. It is non-universal: the signature depends on which phase you're in. The density fluctuation spectrum that was supposed to be a hallmark of the absorbing transition is modified by the glass transition that precedes it.