Stochastic representations for the wave equation on graphs and their scaling limits (Q504901)

The author studies a probabilistic model of the wave equation on weighted (and undirected) graphs under some regularity conditions. The Dirichlet initial boundary value problem of the wave equation is defined as a system of a countable number of linear second-order ordinary differential equations. Then, the probabilistic model is constructed as an interacting particle system, whose particles are however located on both the nodes and the edges of the graph. The author derives a Feynman-Kac type formula that connects the wave equation and the expectation of the state of the interacting particle system at a space-time tuple. Furthermore, \(L^2\) estimates are derived for the interacting particle system, which do not depend on the number of the vertices of the graph. Finally, these estimates are used to establish that, in the high-density hydrodynamic limit, the particle system converges in \(L^2\) to the solution of the wave equation in the Euclidean space. Scaling is with respect to both space and the number of particles in the system. The scaling limit result is sharp due to a phase transition phenomenon, that is also established in the article. If the space grows faster than the particle density, then the aforementioned convergence takes place.