The scaling limit geometry of near-critical 2D percolation (Q867710)

The main purpose is to propose and then to analyze a geometric framework for scaling limits of near-critical models \(2D\) percolation, where \(p=p_c+\lambda\delta^{\theta}\) as \(\delta\to 0\) with \(\lambda\in (-\infty,\infty)\) and \(\theta\) chosen so that macroscopic connectivity functions in the scaling limit have a nontrivial dependence on \(\lambda\). The scaling theory and the other authors results indicate the correct choice \(\theta=\frac{1}{V}=\frac{V_3}{4}\), where \(V\) is the correlation length exponent. The main attention is paid on site percolation on the triangular lattice, or equivalently, random colorings of the hexagonal lattice. The proposed framework provides a random marking of countable many double points each of these labeled by a number in \((-\infty, \infty)\) representing the value of \(\lambda\) at which that double point changes its state and hence correspondingly changes macroscopic connections, loops. This yields a realization on a single probability space of all the scaling limits as \(\lambda\) varies in \((-\infty, \infty)\). The most double points are not marked since they do not change their state for a finite value of \(\lambda\) in the limit \(\delta\to 0\), it is only the marked ones that change. It is explained how the original full scaling limit at the critical point and the marking of double points are together sufficient to yield the scaling limit, simultaneously for all \(\lambda\), of connectivity probabilities and cluster boundary loops in the lattice model with \(p=\frac{1}{2}+\lambda\delta^{\theta}\). Further the percolation transition is analyzed for the continuum model as \(\lambda\) varies and the description of the situation is given for \(\lambda\neq 0\) on a spatial scale where the system continues to look critical. Another natural scaling limit is discussed that should be constructible from the \(O\)-loop process combined with marked double points, namely the continuum minimal spanning tree.