On the limit of random hives with GUE boundary conditions

TL;DR

This paper proves that random hives with GUE boundary conditions converge in probability to a continuum hive, extending previous results by establishing full convergence without subsequences and linking the limit to a supremum of a functional on lozenge tilings.

Contribution

It demonstrates full convergence of random hives with GUE boundary conditions to a continuum limit and characterizes this limit via a variational principle.

Findings

Convergence in probability to a continuum hive as size increases.

The continuum hive value equals a supremum over certain functional on lozenge tilings.

Extension of previous variance results to full convergence without subsequences.

Abstract

We show that hives chosen at random with independent GUE boundary conditions on two sides, weighted by a Vandermonde factor depending on the third side (which is necessary in the context of the randomized Horn problem), when normalized so that the eigenvalues at the edge are asymptotically constant, converge in probability to a continuum hive as $n \rightarrow \infty.$ It had previously been shown in joint work with Sheffield and Tao \cite{NST} that the variance of these scaled random hives tends to $0$ and consequently, from compactness, that they converge in probability subsequentially. In the present paper, building on \cite{NST}, we prove convergence in probability to a single continuum hive, without having to pass to a subsequence. We moreover show that the value at a given point $v$ of this continuum hive equals the supremum of a certain functional acting on asymptotic height functions of lozenge tilings.