Anderson transition at two-dimensional growth rate on antitrees and spectral theory for operators with one propagating channel (Q2628869): Difference between revisions

The paper deals with an Anderson model on a class of discrete graphs, so-called antitrees. Antitrees arise as graphs where the vertices are grouped into shells (discrete distance classes) around some vertices called roots, and edges are precisely connecting all vertices of one shell with all vertices in the neighbouring shells. Edge weights are allowed. Such graphs obey a one-dimensional structure due to the radial direction through the shells. The correponding operators are called ``operators with one propagating channel''. The Anderson model in particular describes operators of the form \[ \mathcal{A} + \lambda \mathcal{V}, \] where \(\mathcal{A}\) is the adjacency operator of the graph describing the kinetic part, \(\mathcal{V}\) is a multiplication operator whose values are i.i.d. random variables at the vertices and \(\lambda\) is a coupling constant. The notions of growth and growth rate for such antitrees are introduced, describing the degree of polynomial growth of balls and shells, respectively. The main result in this paper states that there is an Anderson transition when varying the growth rate of the antitree: for uniform growth rate in \((1,2)\) it is shown that there is dense pure-point spectrum, for growth rate equal to two there is dense pure point and dense purely singular continuous spectrum, and for uniform growth rate larger than 2 there is purely absolutely continuous spectrum. All these spectral results are proven in a particular interval depending on the coupling constant, which is non-empty for small values of \(\lambda\). The result is proven by exploiting the one-dimensional structure (the propagating channel) and transferring results on transfer matrices for Jacobi matrices to this case.