Macroscopic loops in the random loop model on sparse random graphs

Abstract

We study the random loop model with crosses and bars on sparse random graphs. Our main objective is to prove the existence of macroscopic loops, in the sense that a loop visits a positive proportion of the vertices. We develop a deterministic drift method on arbitrary finite graphs based on three ingredients: a local split--merge--rewire analysis for the loop number, an exact differential identity for the partition function, and a slice estimate reducing the relevant same-loop insertion volume to induced edge counts of small vertex sets. This yields a general criterion in terms of a small-set sparsity condition on the underlying graph. We then verify this condition for random regular graphs, sparse Erdős--Rényi graphs, and simple bounded-degree configuration models, obtaining averaged lower bounds on the probability of a macroscopic loop whenever the edge density exceeds an explicit threshold depending on the loop weight \(θ\) and the cross parameter \(u\). For integer values of \(θ\), a trace representation of the partition function implies log-convexity, which upgrades the averaged bounds to pointwise-in-time results away from the threshold time.