Bi-Laplacians on graphs and networks

TL;DR

This paper investigates the behavior of bi-Laplacian operators on graphs and networks, revealing conditions under which their associated parabolic equations exhibit Markovian properties despite the lack of maximum principles.

Contribution

It provides a novel analysis of bi-Laplacian operators on graphs, showing how certain transmission conditions induce Markovian features in their parabolic equations.

Findings

Parabolic equations driven by bi-Laplacians can display Markovian features after transient time.

Transmission conditions at vertices influence the Markovian nature of solutions.

Complete graphs are characterized by the Markovian property of the semigroup generated by the squared Laplacian.

Abstract

We study the differential operator $A=\frac{d^4}{dx^4}$ acting on a connected network $\mathcal{G}$ along with $\mathcal L^2$, the square of the discrete Laplacian acting on a connected discrete graph $\mathsf{G}$. For both operators we discuss well-posedness of the associated {linear} parabolic problems \[ \frac{\partial u}{\partial t}=-Au,\qquad\frac{df}{dt}=-\mathcal L^2 f, \] on $L^p(\mathcal{G})$ or $\ell^p(\mathsf{V})$, respectively, for $1\leq p\leq\infty$. In view of the well-known lack of parabolic maximum principle for all elliptic differential operators of order $2N$ for $N>1$, our most surprising finding is that, after some transient time, the parabolic equations driven by $-A$ may display Markovian features, depending on the imposed transmission conditions in the vertices. Analogous results seem to be unknown in the case of general domains and even bounded intervals. Our analysis is based on a detailed study of bi-harmonic functions complemented by simple combinatorial arguments. We elaborate on analogous issues for the discrete bi-Laplacian; a characterization of complete graphs in terms of the Markovian property of the semigroup generated by $-\mathcal L^2$ is also presented.