Iteration of quadrilateral foldings (Q5917557)

Let \(q_{0}\) be a quadrilateral of \(\mathbb{R}^{2}\) such that its sidelengths are \(a_{1},a_{2},a_{3},a_{4}\). We assume that \((a_{1},a_{2})\) is different from \((a_{4},a_{3})\) and \((a_{1},a_{4})\) is different from \((a_{2},a_{3})\) which implies that the quadrilaterals \(q_{n}=\phi_{4}\circ \phi_{3}\circ \phi_{2}\circ \phi_{1}(q_{n-1})\) obtained by iteration of foldings (where the folding \(\phi_{j}\) replaces the vertex number \(j\) by its symmetric with respect to the opposite diagonal) is well defined. By means of numerical experimentations, Charter and Rogers described some phenomena about the dynamical behavior of the sequence \((q_{n})\) when \(a_{1}=a_{3}\), i.e., for isosceles quadrilaterals and they conjectured that the sequence \((q_{n})\) is bounded whenever \(a_{2}+a_{4}\leq p\) where \(p\) denotes the semiperimeter of the quadrilateral (for details see \textit{K. Charter} and \textit{T. Rogers} [Exp. Math. 2, 209--222 (1993; Zbl 0802.58035)]). The goal of the paper under review is to show the validity of these phenomena and to give an answer to Charter-Rogers' conjecture. As an example of the results stated in the paper we have: Theorem 1. The drift \(v(q_{0})=\lim_{n\rightarrow \infty}(1/n)q_{n}\) exists. Theorem 2. If \(q_{0}\) is of nonperiodic type, then \(v(q_{0})=0\) if and only if \(\max_{1\leq j\leq 4}\{a_{j}\}+\min_{1\leq j\leq 4}\{a_{j}\} \leq (1/2)\sum_{1\leq j\leq 4}a_{j}\). Theorem 3. If \(q_{0}\) is of nonperiodic type, then the sequence \((q_{n})\) converges to infinite if and only if \(v(q_{0})\) is different from zero. In relation to Charter-Rogers' conjecture it is shown that it is true in the case that the vertices of \(q_{0}\) are algebraic, and that it is also true for Lebesgue almost all \(q_{0}\) but that is false for Baire generic \(q_{0}\).