The Gamma-disordered Aztec diamond
Maurice Duits, Roger Van Peski
TL;DR
This work introduces a Gamma-disordered dimer model on the Aztec diamond and proves there is no phase transition in the free energy, while the annealed-quenched energy gap remains positive with a universal $n^2$-scaling. It builds exact distributional bridges between random dimer configurations and integrable directed polymers (log-Gamma, Beta, and strict-weak), enabling transfer of KPZ-class fluctuation results to turning points on the Aztec diamond. The authors develop a unique shuffle-invariant full-plane Gamma family via Lukacs' theorem and extend the analysis to hybrid polymers along vertical and horizontal slices, yielding a unified framework that connects dimer models with multiple polymer universality classes. The results provide precise turning-point statistics, dynamical behavior under the shuffling algorithm, and a robust methodology to port exact polymer results to discrete dimer systems, advancing our understanding of glassy phases and integrable structures in random tilings.
Abstract
We introduce a multi-parameter family of random edge weights on the Aztec diamond graph, given by certain Gamma variables, and prove several results about the corresponding random dimer measures. Firstly, we show there is no phase transition at the level of the free energy. This provides rigorous backing for the physics predictions of Zeng-Leath-Hwa and later works that dimer models with random weights are in the glassy `super-rough' phase at all temperatures with no phase transition. Secondly, we show that the random dimer covers themselves enjoy exact distributional equalities of certain marginals with path locations in new `hybrid' integrable polymers. These reduce to the stationary log-Gamma, strict-weak, and Beta polymer in random environment in certain cases, allowing transfer of known results from integrable polymers to dimers with random weights. As an example application, we prove that the turning points at the boundaries of the Aztec diamond exhibit fluctuations of order $n^{2/3}$, in contrast to the $n^{1/2}$ fluctuations for deterministic weights. Underlying all these is a key integrability property of the weights: they are the unique family for which independence is preserved under the shuffling algorithm.
