Iterating evolutes and involutes (Q2408188)

This article's topic is iterated evolute and involute transforms of planar curves in Euclidean plane. Mostly concerned with the discrete case, it takes into account several possible discretizations, develops their transformation theory and derives some very pleasing results. The article starts off with some considerations on the smooth case where hypocycloids play a special role. They are the only admissible curves that are similar to their evolute and constitute the only class of ``nice'' limit shapes of the iterated evolute transform. (Generically, the number of cusps diverges to infinity.) In the discrete setting, the authors distinguish \(\mathcal P\)-evolutes (whose sides are perpendicular bisectors of the original polygon sides) and \(\mathcal A\)-evolutes (whose sides bisect the original polygon's angles). This means that the evolute vertices correspond, respectively, to circumcenters and in-/excenters of triangles formed by consecutive vertices of the original polygon. Since angle bisectors are not unique, different strategies for a consistent selection of bisectors and hence different versions of \(\mathcal A\)-evolutes are considered. The proper tool for studying iterated \(\mathcal P\)-evolutes is the discrete support function. Defined on the polygon sides, it returns a pair consisting of direction angle and signed distance to the origin. The \(\mathcal P\)-evolute transform simply shifts angles by \(\pi/2\) but it acts linearly on the vector of oriented distances. The analysis of the spectral properties of this linear map results in a description of the limiting behaviour of iterated \(\mathcal P\)-evolutes. Discrete hypocycloids (polygons whose sides are tangent to a smooth hypocycloid) are rather particular. As in the smooth case, their \(\mathcal P\)-evolutes are scaled copies of itself. In the smooth case, curve and evolute share their Steiner point whose role in the discrete setting is partly taken over by the newly introduced ``pseudo Steiner point''. The analogy is, however, not perfect. Investigation for one particular kind of \(\mathcal A\)-evolutes yields comparable results. Throughout the text, pentagons play a special role. We find, for example, a proof of Grünbaum's conjecture that the first and the third \(\mathcal P\)-evolute of a pentagon are homothetic. The authors also observe that one particular kind of the \(\mathcal A\)-evolute transform produces, with positive probability, a four-periodic sequence of limit shapes. The (rather strange) limiting pentagons seem to be independent of the starting value. Invertability of the \(\mathcal P\)-evolute transform and a certain \(\mathcal A\)-evolute transform depends very much on properties of the composition of the reflections in consecutive polygon sides (and in particular on the parity of their number). Generically, the \(\mathcal P\)-involute exists for an even and the \(\mathcal A\)-involute exists for an odd number of sides. Otherwise, the vanishing of the polygon's ``quasiperimeter'' ensures existence of an infinity of involutes. Among them, there is generically one with zero quasiperimeter whence the construction allows iteration. Some results on the limiting behaviour of certain polygons are derived.