Trie Partitioning Process: Limiting Distributions

This paper is devoted to the well-known trie structure. We consider two basic parameters: depth of the leaves and height when the trie is formed with n items. We prove the convergence of their distributions and of their moments of any order when n → ∞ to a limit distribution. We exhibit the limits : a periodic distribution or a normal distribution. The results are given for uniform or biased data distributions for Bernoulli and Poisson models. Our reasoning is based on generating and characteristic functions. We make an extensive use of analytic functions and asymptotic methods.