Minimal spanning forests (Q858976): Difference between revisions

Minimal spanning forests on infinite graphs are weak limits of minimal spanning trees from finite subgraphs. These limits can be taken with free or wired boundary conditions and are denoted by FMSF (free minimal spanning forest) and WMSF (wired minimal spanning forest). In the paper authors study these forests and compare them with another natural random forests on infinite graphs which are free uniform spanning forest (FSF), constructed by \textit{R. Pemantle} [Ann. Probab. 19, 1559--1574 (1991; Zbl 0758.60010)] and wired uniform spanning forest (WSF). Among others the following properties of WMSF and FMSF are proved. On any Cayley graph where the critical percolation has no infinite clusters, all component trees in the WMSF have one end a.s. This extends the result of \textit{K. S. Alexander} [Ann. Probab. 23, 87--104 (1995; Zbl 0827.60079)] on \(\mathbb Z^d\). On any graph, the percolation threshold of any component of the WMSF, equals one, a.s. This and the previous result can be interpreted as `thinness' of WMSF. On the other hand FMSF is shown to be `thick': on any connected graph, the union of the FMSF and of independent Bernoulli percolation with arbitrarily small parameter is a.s. connected. In conjunction with the result of \textit{D. Gaboriau} [Geom. Funct. Anal. 15, 1004--1051 (2005; Zbl 1099.60070)], this implies that in any Cayley graph, the expected degree of an vertex of FMSF is at least the expected degree of the FSF. Let \(p_u\) be the critical threshold for having a unique infinite cluster in the Bernoulli percolation. On any graph, the number of infinite components of this percolation at \(p_u\) is at most the number of components of the FMSF. On any connected graph, the both measures FMSF and WMSF coincide iff for almost every \(p\) (in the sense of Lebesgue measure) there is a.s. at most one infinite component in the Bernoulli percolation with parameter \(p\). Finally, an example is given to show that the minimal spanning tree measures does not have negative associations.