Graph Laplace and Markov operators on a measure space (Q2010219): Difference between revisions

\((G, c)\) is called a weighted network if \(G = (V,E)\) is a countable connected locally finite graph with no loops, and \(c = c_{xy}\) is a symmetric function defined on pairs of connected vertices. In general, the theory of weighted networks is built around two important operators acting on the space of functions \(f : V \to {\mathbb R}\). They are the Laplace operator \(\Delta=\sum\limits_{x\sim y} c_{xy}(f(x)-f(y))\) and the Markov operator \(P=\sum\limits_{x\sim y}p(x,y)f(y)\), \(p(x,y)=\frac{c_{xy}}{c(x)}\), \(c(x)=\sum\limits_{x\sim y}c_{xy}\). The Laplacian generates the operator in the Hilbert spaces \(\ell^2(V)\), \(\ell^2(V,c)\), and the \textit{finite energy space} \(\mathcal{H}\) which is formed by functions \(f : V \to {\mathbb R}\) such that \(\|f\|_{\mathcal{H}}^2=\frac{1}{2}\sum\limits_{x\sim y} c_{xy}(f(x)-f(y))^2 < \infty\). One can think of a countable network as a discrete measure space \((V,m)\) with the counting measure \(m\). There are many applications where the framework of countable discrete sets of vertices \(V\) and edges \(E\) is much too restrictive. This circumstance motivated the authors to build a measurable analogue to the theory of weighted networks on infinite graphs. In the paper under review, the basic setting is an infinite \(\sigma\)-finite measure space \((V, \mathcal{B}, \mu)\) and a symmetric measure \(\rho\) on \((V \times V,\mathcal{B} \times \mathcal{B})\) supported by a measurable symmetric subset \(E \subset V \times V\). As in the theory of weighted networks, the authors consider the Hilbert spaces \(L^2(\mu), L^2(c\mu)\) and define two other Hilbert spaces, the dissipation space \(\mathrm{Diss}\) and the finite energy space \(\mathcal H_E\). The graph Laplacian \(\Delta\) and Markov operators \(P\) are defined in this setting. The following topics are studied: spectral properties of the operator \(P\), harmonic functions for \(P\), the Markov process generated by \(P\), the path spaces \(\Omega\) and \(\Omega_x\) and the corresponding path measures, reversibility of the Markov process. The final part of the paper is devoted to the study of spectral properties of the Laplacians \(\Delta_2\) and \(\Delta_{\mathcal{H}}\), the Laplace operators acting in \(L^2(\mu)\) and \(\mathcal{H}_E\), respectively. It turns out that \(\Delta_2\) is a positive definite and self-adjoint unbounded operator in \(L^2(\mu)\). On the other hand, \(\Delta_{\mathcal{H}}\) is a symmetric operator that admits many selfadjoint extensions. For the entire collection see [Zbl 1423.46002].