Prof. Dr. Benedikt Stufler

Simulations: Encoding processes of simply generated trees


In a rooted plane (ordered) tree, list the vertices v1, …, vn in depth-first order. The Łukasiewicz path is the walk (Wj)0 ≤ j ≤ n with W0=0 and increments Wj−Wj−1 = (number of children of vj) − 1; it stays nonnegative up to time n−1 and ends at −1, and it encodes the whole tree. The height process is (Hj)1 ≤ j ≤ n where Hj is the depth (graph distance from the root) of vj. The (discrete) looptree of a plane tree is the planar graph on the same vertex set obtained by, for each vertex with k children, connecting consecutive children by edges and also connecting the parent to its first and last child (and discarding the original tree edges).

The following illustrations display a randomly generated tree in the upper left with the associated looptree in the upper right corner, the encoding Łukasiewicz path in the lower left corner, and the height process in the lower right corner.

These pictures were published in the following paper:
B. Stufler, On the maximal offspring in a subcritical branching process, Electronic Journal of Probability (2020), Vol. 25, paper no. 104, 1–62.

A Bienaymé—Galton—Watson tree conditioned on having 500k vertices, with offspring distribution having average value 1 and a power-law tail with exponent -1.5.


A Bienaymé—Galton—Watson tree conditioned on having 500k vertices, with offspring distribution having average value 1 and a power-law tail with exponent -3.


A Bienaymé—Galton—Watson tree conditioned on having 100k vertices, with offspring distribution having average value 1 and a regularly varying tail with exponent -1.


A Bienaymé—Galton—Watson tree conditioned on having 100k vertices, with offspring distribution having average value 0.75 and a regularly varying tail with exponent -1.5.