Eigenvalue method to compute the largest relaxation time of disordered systems

TL;DR

This paper introduces an eigenvalue-based method to compute the largest relaxation time in disordered systems, avoiding dynamic simulations, and applies it to models like self-affine potentials and spin glasses, revealing scaling behaviors and distribution tails.

Contribution

It develops a numerical eigenvalue approach to determine relaxation times in disordered systems, providing new insights into their scaling laws and fluctuation properties.

Findings

Activated scaling in self-affine potentials with exponent H

Growth of relaxation time in spin glasses with exponent 1/3

Distribution tail of relaxation times decays as e^{-u^{1.36}}

Abstract

We consider the dynamics of finite-size disordered systems as defined by a master equation satisfying detailed balance. The master equation can be mapped onto a Schr\"odinger equation in configuration space, where the quantum Hamiltonian $H$ has the generic form of an Anderson localization tight-binding model. The largest relaxation time $t_{eq}$ governing the convergence towards Boltzmann equilibrium is determined by the lowest non-vanishing eigenvalue $E_1=1/t_{eq}$ of $H$ (the lowest eigenvalue being $E_0=0$). So the relaxation time $t_{eq}$ can be computed {\it without simulating the dynamics} by any eigenvalue method able to compute the first excited energy $E_1$. Here we use the 'conjugate gradient' method to determine $E_1$ in each disordered sample and present numerical results on the statistics of the relaxation time $t_{eq}$ over the disordered samples of a given size for two models : (i) for the random walk in a self-affine potential of Hurst exponent $H$ on a two-dimensional square of size $L \times L$, we find the activated scaling $\ln t_{eq}(L) \sim L^{\psi}$ with $\psi=H$ as expected; (ii) for the dynamics of the Sherrington-Kirkpatrick spin-glass model of $N$ spins, we find the growth $\ln t_{eq}(N) \sim N^{\psi}$ with $\psi=1/3$ in agreement with most previous Monte-Carlo measures. In addition, we find that the rescaled distribution of $(\ln t_{eq})$ decays as $e^{- u^{\eta}}$ for large $u$ with a tail exponent of order $\eta \simeq 1.36$. We give a rare-event interpretation of this value, that points towards a sample-to-sample fluctuation exponent of order $\psi_{width} \simeq 0.26$ for the barrier.