Loop correlations in random wire models (Q2304953)

In statistical mechanics, loop soup models describe one-dimensional loops in higher spatial dimensions and provide alternative representations of various particle and spin models. In three and higher dimensions, these models have been recently conjectured [\textit{C. Goldschmidt} et al., Contemp. Math. 552, 177--224 (2011; Zbl 1244.82010)] to exhibit phases with macroscopically long loops and the joint distribution of macroscopic loops is of Poisson-Dirichlet type. This paper introduces a random wire model, which is a special class of loop soup models. This model is defined on the hypercubic lattice and has a random number of links connecting neighboring sites, with a constraint that each site has an even number of links. Such links thus form loop configurations, and the statistical weight defining the partition function is associated with different loop configurations. By using the connection to the classical XY model, this paper proves that for properly chosen parameters, even loop correlations belong to the Poisson-Dirichlet type, thus providing a concrete example for which the above-mentioned conjecture is rigorously established.