Bootstrap percolation on homogeneous trees has 2 phase transitions (Q960137)

Bootstrap percolation may serve as a mathematically idealized model of nucleation and growth. It may as well serve as a prototype cellular automaton visualizing the epidemic infection spreading from a randomly set number of initially infected lattice sites/vertices (with a prescribed initial density on a graph or tree) that is induced by a deterministic spreading rule. To follow the present paper, one needs first to consult [\textit{J. Balogh, Y. Peres} and \textit{G. Pete}, Comb. Probab. Comput. 15, No. 5, 715--730 (2006; Zbl 1102.60086)] paper on bootstrap percolation on infinite trees. New results are obtained for the bootstrap percolation process on a homogeneous tree. It is shown that in addition to the known critical point (e.g. critical density) above which the tree becomes fully occupied for almost every initial configuration, there is another critical point above which the occupied sites percolate, i.e. for almost every initial configuration there are infinite clusters of occupied vertices in the final configuration. The probability of percolation is proved to be continuous and analytic in a suitable subinterval of density values which belongs to \([0,1]\).