Lozenge tilings, Glauber dynamics and macroscopic shape (Q2354006)

Random lozenge tilings is a very natural object in mathematical physics, probability, combinatorics, theoretical computer science, etc. Let \(\mathcal{T}_L\) be the triangular lattice of mesh \(1/L\) and call the union of two adjacent triangular faces a lozenge. For any lozenge tiling of some tileable region \(U\) one can associate with this tiling some height function. Then, it is well known that when the mesh tends to zero, the height function of a random tiling sampled from the uniform measure tends in probability to a certain limit shape \(\overline{\phi}\). Moreover, the limit shape \(\overline{\phi}\) is either analytic or contains so called ``frozen regions''. The example of ``frozen regions'' in random tilings is the famous ``arctic circle theorem''. The Glauber dynamics on lozenge tilings is a natural Markov process, whose updates consist of rotating by an angle \(\pi\) three lozenges that share a vertex. It is widely believed that this dynamics converges to some deterministic macroscopic evolution under some suitable time rescaling. Moreover, it is conjectured that the time to reach equilibrium should scale like \(L^{2+o(1)}\). In the reviewed paper, some partial results on this conjecture are proved. Informally, it is proved that if \(\overline{\phi}\) contains no ``frozen region'' then, whatever the initial condition of Glauber dynamics, at time \(L^{2+o(1)}\) the height function is with high probability at distance \(o(1)\) from \(\overline{\phi}\). Also, it is proved that if the macroscopic shape \(\overline{\phi}\) in some domain \(U\) is smooth in the neighborhood of a point \(u\in U\), then the Taylor expansion of \(\overline{\phi}\) around \(u\) coincides up two second order with the Taylor expansion of the macroscopic shape of some hexagon.