First Order Geometry of Equilateral Small Polygons
José Ayala Hoffmann, Tomás Ayala Coppola
math.MG
Nov 17, 2026 · v1
cs.CG
TL;DR
A Lean 4 formalisation in the ESP1 package verifies the strict local maximality criterion and the maximality results for A16, A32, A64, Z32 and Z64.
Abstract
We develop a first order calculus based on the geometry of realised diameter graphs for perimeter maximisation among convex equilateral small polygons. Polygon side equalities and active diameter constraints define a linearised feasible cone and an associated equality space. We prove that an equilateral small polygon is a strict local maximiser of perimeter modulo Euclidean motions if a multiplier balance holds with strictly positive active diameter multipliers and the active equality space consists only of infinitesimal Euclidean motions. The proof uses a local Euclidean slice and a geometric limiting argument and requires neither independence of the active constraint gradients nor second order information. As an application, we combine our criterion with validated interval arithmetic to establish strict local maximality for Bingane and Audet's polygons $A_{16}$, $A_{32}$, $A_{64}$ and for reconstructions of their polygons $Z_{32}$ and $Z_{64}$. For each $n\in\{32,64\}$, the perimeter of $Z_n$ is strictly larger than that of $A_n$. Thus, strict local maximisers with different perimeters coexist in the space of convex equilateral small polygons with the same number of sides.
Problem
The maximum perimeter of convex equilateral small polygons (diameter at most one) is unknown for n = 2^k with k ≥ 4. Sufficient conditions are needed to certify that a given polygon is a strict local maximiser.
Approach
A first order criterion is derived from realised diameter graphs. Strict local maximality modulo Euclidean motions follows from a multiplier balance with strictly positive diameter multipliers together with an active equality space that contains only infinitesimal motions. The proof uses a local Euclidean slice and a limiting argument. The hypotheses are certified with exact algebra and 256-bit interval arithmetic in Julia, and a Lean 4 formalisation independently checks the criterion and the polygon results.
Results
Strict local maximality is established for Bingane–Audet polygons A16, A32, A64 and for reconstructions Z32 and Z64. Z_n has strictly larger perimeter than A_n for n = 32 and 64, so strict local maximisers with different perimeters coexist.
| n | z_n | nℓ |
|---|
| 16 | 0.1962703721 | 3.1352878881 |
| 32 | 0.0981721090 | 3.1402460942 |
| 64 | 0.0490872986 | 3.1412717079 |
Constructor root and perimeter of A_n