Discrete diffusion-type equation on regular graphs and its applications

TL;DR

This paper derives explicit formulas for the fundamental solution of the discrete diffusion equation on regular graphs, and applies these results to spectral analysis, return time distributions, and counting specific walks.

Contribution

It provides explicit formulas for the diffusion equation on regular graphs and develops applications in spectral theory, return time analysis, and walk enumeration.

Findings

Explicit formula for the fundamental solution on regular trees.

Closed-form expression for return time distribution on regular graphs.

Asymptotic convergence of return time distributions to the tree case.

Abstract

We derive an explicit formula for the fundamental solution $K_{T_{q+1}}(x,x_{0};t)$ to the discrete-time diffusion equation on the $(q+1)$-regular tree $T_{q+1}$ in terms of the discrete $I$-Bessel function. We then use the formula to derive an explicit expression for the fundamental solution $K_{X}(x,x_{0};t)$ to the discrete-time diffusion equation on any $(q+1)$-regular graph $X$. Going further, we develop three applications. The first one is to derive a general trace formula that relates the spectral data on $X$ to its topological data. Though we emphasize the results in the case when $X$ is finite, our method also applies when $X$ has a countably infinite number of vertices. As a second application, we obtain a closed-form expression for the return time probability distribution of the uniform random walk on any $(q+1)$-regular graph. The expression is obtained by relating $K_{X}(x,x_{0};t)$ to the uniform random walk on a $(q+1)$-regular graph. We then show that if $\{X_{h}\}$ is a sequence of $(q+1)$-regular graphs whose number of vertices goes to infinity and which satisfies a certain natural geometric condition, then the limit of the return time probability distributions from $\{X_{h}\}$ is equal to the return time probability distribution on the tree $T_{q+1}$. As a third application, we derive formulas which express the number of distinct closed irreducible walks without tails on a finite graph $X$ in terms of moments of the spectrum of its adjacency matrix.