Random walks and induced Dirichlet forms on self-similar sets (Q2411352): Difference between revisions

The authors study random walks on the augmented trees \((\Sigma^*,\mathfrak E)\) and the Dirichlet forms induced by them on a self-similar set \(K\) (generated by an iterated function system IFS). \(\Sigma^*\) is the symbolic space (coding space) which gives a representation of any member of \(K\). It is assumed that the IFS, resp., \(K\) satisfies the open set condition which entails that the representation of generic points of \(K\) is unique. The augmented tree refers to adding certain horizontal edges on \(\Sigma^*\). In the present paper a class of reversible random walks is investigated such that \(K\) can be identified with the Martin boundary of the walk and the the hitting distribution becomes normalized Hausdorff measure on \(K\) with a Naim kernel of order \(|x-y|^{-\alpha-\beta}\); here \(\alpha\) is the Hausdorff dimension of \(K\) and \(\beta\) is a number depending on the return ratio of the walk. For suitable \(\beta\), there is a regular non-local Dirichlet form on \(K\) which is studied in the last part of the paper.