Penny graphs in the hyperbolic plane
Ádám Sagmeister, Konrad J. Swanepoel
TL;DR
This work analyzes penny graphs arising from packings of congruent circles in the hyperbolic plane, focusing on the maximum number $e_d(n)$ of touching pairs and its asymptotic density $c(d)=\lim_{n\to\infty} e_d(n)/n$. Using a combination of spiral constructions tied to hyperbolic tilings, Euler–Gauss–Bonnet type area bounds, and isoperimetric arguments, the authors derive a suite of upper bounds $\gamma_i(d)$ that sharpen the trivial $2\le c(d)\le 3$ and prove tight results for distances $d(k)$ corresponding to order-$k$ tilings; they establish a self-contained Bowen-type upper bound and extend the analysis to subgraphs of $\{p,q\}$-tilings. They also provide nontrivial lower bounds via constructive spiral packings, showing $c(d)\ge \gamma_6(d)$ with $\gamma_6(d)=2+\frac{1}{4q-14}=2+\frac{\alpha}{8\pi}+O(\alpha^2)$ as $d\to\infty$, and demonstrate that $\lim_{d\to\infty} c(d)=2$. The results imply that, except for a small set of distances, the order-$7$ tiling spiral is extremal among penny graphs in the hyperbolic plane, and they discuss the horocycle case as a complementary extremal scenario.
Abstract
We consider the problem of finding the maximum number $e_d(n)$ of pairs of touching circles in a packing of $n$ congruent circles of diameter $d$ in the hyperbolic plane of curvature $-1$. In the Euclidean plane, the maximum comes from a spiral construction of the tiling of the plane with equilateral triangles (Harborth 1974), with a similar result in the hyperbolic plane for the values of $d$ corresponding to the order-$k$ triangular tilings (Bowen 2000). We present various upper and lower bounds for $e_d(n)$ for all values of $d > 0$. In particular, we prove that if $d > 0.66114\dots$ except for $d=0.76217\dots$, then the number of touching pairs is less than the one coming from a spiral construction in the order-$7$ triangular tiling, which we conjecture to be extremal. We also give a lower bound $e_d(n) > (2+\varepsilon_d)n$ where $\varepsilon_d > 1$ for all $d > 0$.
