Equilibrium avalanches in spin glasses

TL;DR

This paper analyzes equilibrium avalanches in spin glasses, deriving their distribution and exponents for the SK model and finite-range models, revealing criticality and suggesting experimental and numerical investigations.

Contribution

It introduces a multi-component Parisi-Duplantier equation to compute avalanche distributions and derives a universal formula for the avalanche exponent in finite-range models.

Findings

The avalanche size distribution follows a power law with exponent tau=1 in the SK model.

The density of overlap q in avalanches scales as 1/(1-q).

Finite-range models predict tau=(d_f + theta)/d_m, linking geometry and critical exponents.

Abstract

We study the distribution of equilibrium avalanches (shocks) in Ising spin glasses which occur at zero temperature upon small changes in the magnetic field. For the infinite-range Sherrington-Kirkpatrick model we present a detailed derivation of the density rho(Delta M) of the magnetization jumps Delta M. It is obtained by introducing a multi-component generalization of the Parisi-Duplantier equation, which allows us to compute all cumulants of the magnetization. We find that rho(Delta M) ~ (Delta M)^(-tau) with an avalanche exponent tau=1 for the SK model, originating from the marginal stability (criticality) of the model. It holds for jumps of size 1 << Delta M < N^(1/2) being provoked by changes of the external field by delta H = O(N^[-1/2]) where N is the total number of spins. Our general formula also suggests that the density of overlap q between initial and final state in an avalanche is rho(q) ~ 1/(1-q). These results show interesting similarities with numerical simulations for the out-of-equilibrium dynamics of the SK model. For finite-range models, using droplet arguments, we obtain the prediction tau= (d_f + theta)/d_m, where d_f,d_m and theta are the fractal dimension, magnetization exponent and energy exponent of a droplet, respectively. This formula is expected to apply to other glassy disordered systems, such as the random-field model and pinned interfaces. We make suggestions for further numerical investigations, as well as experimental studies of the Barkhausen noise in spin glasses.