Ultrametric and tree potential (Q1028627)

An ultrametric matrix \(U = (U_ {i j} : i, j \in I )\) is a symmetric nonnegative matrix verifying the ultrametric inequality \(U_ {i j} \geq \min \{ U_ {i k}, U_ {k j}\}\) for all \(i, j, k \in I\). Tree matrices are associated with a (locally finite) rooted tree \((I, {\mathcal T})\), with root \(r\) and a strictly increasing function \(w:\{ |k| : k \in I \} \rightarrow \mathbb {R}_ {+}\) by \(U_ {i j} = w_ {|i\wedge j|}\) with \(i\wedge j\) being the farthest vertex from the root that is common to the geodesic from \(i\) and \(j\) to \(r\). Tree matrices are special case of ultrametric matrices. Conversely, each ultrametric matrix posseses a minimal tree matrix extension. One of the purposes of this paper is to extend to countably infinite ultrametric and tree matrices the results obtained by \textit{C. Dellacherie}, \textit{S. Martinez} and \textit{J. San Martin} [Adv. Appl. Math. 17, No. 2, 169--183 (1996; Zbl 0861.60077)] for the finite case. The existence of an associated symmetric random walk is proved for each tree matrix. Its Green potential is studied and a representation formula for (increasing) harmonic function is given. For ultrametric matrices, probabilistic conditions are supplied in order to study its potential properties when immersed in its minimal tree matrix extension.