Frozen percolation in two dimensions (Q892165)

The paper is devoted to the so called frozen percolation process on a triangular lattice. This process can be described as follows. Let \(G\) be a graph in the triangular lattice and \(n\in\mathbb{N}\). For every vertex \(v\) independently from each others we assign a random value \(\tau_v\) uniformly distributed on \([0,1]\). At time \(t=0\) all of the vertices of \(G\) are closed. The vertex \(v\) becomes open at time \(t=\tau_v\) if all its neighbours' open clusters at time \(\tau_v\) have size less than \(N\) in the \(L^{\infty}\)-metric. The frozen percolation process was introduced by \textit{D. J. Aldous} [Math. Proc. Camb. Philos. Soc. 128, No.~3, 465--477 (2000; Zbl 0961.60096)] for an infinite binary tree. From classical percolation theory, it follows that at time \(t<1/2\) for all \(K>0\) the probability that there is an open cluster with diameter at least \(N\) in a box with side length \(KN\) goes to \(0\) as \(N\to\infty\). The main result of the paper states that this effect does not take place at times \(t>1/2\). More precisely, it is proved that, for all \(K>0\) and \(t>1/2\), the probability that after time \(t\) a frozen cluster forms which intersects a given box with side length \(KN\) goes to \(0\) as \(N\to\infty\). Also, some quantitative estimates for this probability are established. As a corollary, it is proved that for any \(\varepsilon>0\) there exist \(a, b\in (0,1)\), \(N_0\) depending on \(\varepsilon\), such that, for any \(N>N_0\), the probability that the diameter of an open cluster at time \(1\) lies in the interval \((aN,bN)\) is greater than \(1-\varepsilon\). Some generalizations to frozen percolation on square lattices are discussed.