Metastable states of spin glasses on random thin graphs

TL;DR

This paper analyzes the number of metastable states in spin glasses on random thin graphs, showing exponential growth with system size and how this number varies with connectivity, connecting finite connectivity models to the fully connected limit.

Contribution

It provides the first detailed calculation of metastable states in spin glasses on random thin graphs, revealing how their number depends on connectivity and converges to the SK model in the infinite limit.

Findings

Number of metastable states grows exponentially with system size.

Average metastable states decrease as connectivity c increases.

Finite connectivity corrections increase metastable states compared to mean field models.

Abstract

In this paper we calculate the mean number of metastable states for spin glasses on so called random thin graphs with couplings taken from a symmetric binary distribution $\pm J$. Thin graphs are graphs where the local connectivity of each site is fixed to some value $c$. As in totally connected mean field models we find that the number of metastable states increases exponentially with the system size. Furthermore we find that the average number of metastable states decreases as $c$ in agreement with previous studies showing that finite connectivity corrections of order $1/c$ increase the number of metastable states with respect to the totally connected mean field limit. We also prove that the average number of metastable states in the limit $c\to\infty$ is finite and converges to the average number of metastable states in the Sherrington-Kirkpatrick model. An annealed calculation for the number of metastable states $N_{MS}(E)$ of energy $E$ is also carried out giving a lower bound on the ground state energy of these spin glasses. For small $c$ one may obtain analytic expressions for $<N_{MS}(E)>$.