The height of increasing trees.pdf

The height of increasing trees N. Broutin L. Devroye E. McLeish M. de la Salle? May 21, 2007 Abstract We extend results about heights of random trees (Devroye, 1986, 1987, 1998b). In this paper, a general split tree model is considered in which the normalized subtree sizes of nodes converge in distribution. The height of these trees is shown to be in probability asymptotic to c logn for some constant c. We apply our results to obtain a law of large numbers for the height of all polynomial varieties of increasing trees (Bergeron et al., 1992). Keywords and phrases: Height, random tree, branching process, probabilistic analysis, increasing tree. 1 Introduction The present paper gives a general result on the heights of random trees. It applies in particular to general d-ary increasing trees (Bergeron, Flajolet, and Salvy, 1992), a random tree model whose height was not known until now, although the variance of the height is known to be bounded (Drmota, 2006). The general approach we adopt is based on branching processes (Athreya and Ney, 1972; Harris, 1963). The hunt for the height of binary search trees has motivated the development of these branching processes techniques. Pittel (1984) was the first to introduce a crucial continuous embedding and to show that the height of a tree of n nodes is asymptotic to c log n for some positive constant c. Using earlier work by Biggins (1976, 1977), Devroye (1986) proved that c ≥ 2 is given by the solution of c log(2e/c) = 1, hence showing that the height of random binary search trees of size n is asymptotic to 4.311 . . . log n in probability. Using branching random walks Biggins and Grey (1997) were able to generalize these theorems and extend the class of random trees that can be handled using this single method. More recently, Broutin and Devroye (2006) gave further results based on 2-dimensional branching processes that seem to encompass many extremal problems related to paths in trees. In our tree model, we distinguish the