1: \begin{abstract}

2: The broad motivation of this work is a rigorous understanding of

3: reversible, local Markov dynamics of interfaces, and in particular their

4: speed of convergence to equilibrium,  measured via the

5: mixing time $T_{mix}$.  In the $(d+1)$-dimensional setting, $d\ge2$, this is to a large extent

6: mathematically unexplored territory,

7: especially for discrete interfaces. On the other hand, on the basis of a

8: mean-curvature motion heuristics \cite{Spohn,Henley} and simulations (see

9: \cite{Destainville} and the references in \cite{Wilson,Henley}), one expects

10: convergence to equilibrium to occur on time-scales of order

11: $\approx \delta^{-2}$ in any dimension, with $\delta\to0$ the lattice mesh.

12:   

13: We study the single-flip Glauber dynamics for lozenge tilings of

14: a finite domain of the plane, viewed as $(2+1)$-dimensional surfaces.

15: The stationary measure is the uniform measure on admissible

16: tilings. At equilibrium, by the limit shape theorem \cite{CKP}, the

17: height function concentrates as $\delta\to0$ around a deterministic

18: profile $\phi$,  the unique minimizer of a surface tension

19: functional.  Despite some partial mathematical

20: results \cite{Wilson,LT_ptrf,LT_cmp}, the

21: conjecture  $T_{mix}=\delta^{-2+o(1)}$  has been

22: proven, so far, only in the situation where $\phi$ is an affine

23: function \cite{CMT}.  In this work, we prove the conjecture

24:  under the sole assumption that the limit

25: shape $\phi$ contains no frozen regions (facets).

26: \end{abstract}

27: