Hydrodynamical limits and geometric measure theory: mean curvature limits from a threshold voter model (Q1969450)

The author considers a generalization of the model studied by Spohn and develops necessary tools within the framework of the geometric measure theory to prove the existence of hydrodynamical limits in the system of interacting particles. To describe a particle system for each configuration \(\sigma\) on the configuration space \(\mathcal X=\{0,1\}^{Z^2}\) the author constructs a process \[ n_\sigma(x)=\sum_{y\in Z^2, \|x-y\|_1=1}\sigma^x(y)-\sigma(x),\;\;x\in {\mathcal X}, \quad \text{and}\quad \sigma^x(y)=\begin{cases} 1-\sigma(x)&\text{ if } y=x,\\ \sigma(y)&\text{ else},\end{cases} \] and considers an operator \[ Lf(\sigma)=\sum_{x\in Z^2}\chi_{\{n_\sigma(x)\geq 2\}}[f(\sigma^x)-f(\sigma)] \] acting on bounded measurable functions defined on \(\mathcal X\). An evolution of an island \({\mathcal S}_\sigma \) of 1's \[ {\mathcal S}_\sigma =\bigcup _{x\in Z^2,\sigma(x)=1}\{y\in R^2:\|y-x\|_\infty \leq 1/2\} \] is described next in terms of currents \[ {\mathcal F}_F(\varphi)= \int_{x\in {\mathcal S}_{\sigma/n}}\langle\varphi(x),e_1^*\wedge e_2^*\rangle dx\quad \text{and}\quad (\partial{\mathcal F})_F(\varphi)={\mathcal F}_F(d\varphi) \] defined for a 2-form \(\varphi\) and a basis \(\{e_1,e_2\}\) in \(R^2\) by putting \(T_\sigma^n=\partial {\mathcal F}_{{\mathcal S}_\sigma/n}\). Specifying initial data \(\pi_0^n, n\geq 1\) (a set of probability measures on \({\mathcal X}\)), and fixing the evolution the author considers a Markov process on the probability space \(D_{\mathcal X}[0,\infty)\) with \(\sigma_t(\omega)=\omega(t)\), and defines for each \(n\) \( P_n\in {\mathcal P}(D_{\mathcal X}[0,\infty))\) as a solution to a martingale problem with the generator \(n^2L\) and initial data \(\pi_0^n\). The main goal of the paper is to define a macroscopic description of \(P_n\) law of the process \(\{T^n_{\sigma_t}\), \(t\geq 0\}\). The author proves that under some density and local indecomposability assumptions \(T^n_\sigma\) converges as \(n\to\infty\) to a unique solution of an integral equation which is equivalent to the mean curvature flow in a Hausdorff measure, i.e. the gradient flow of the length function.