Moduli spaces of pentagonal subdivision tilings (Q6858030)

The classification of edge-to-edge tilings of the sphere by congruent polygons has been a sustained research programme spanning the last century, beginning with Sommerville's work on triangular tilings in 1924 and completed for triangles by Ueno and Agaoka in 2002. The corresponding classification for pentagons -- which is the most case-rich polygon type, since any polygon appearing in an\Nedge-to-edge spherical tiling by congruent polygons must be a triangle, quadrilateral, or pentagon -- was carried out across a series\Nof papers by (subsets of) Gao, Shi, Akama, Cheung, Luk, Liao, Qian, Shu, Hu, Liang, Wang, Xu, and Yan over the period 2013--2026. In particular, an edge-to-edge tiling of the sphere by congruent pentagons admits one of five possible edge combinations: \(a^2b^2c\), \(a^3bc\), \(a^3b^2\), \(a^4b\), and~\(a^5\). The cases \(a^2b^2c\) and \(a^3bc\) were resolved by \textit{E. Wang} and \textit{M. Yan} [Adv. Math. 394, Article ID 107866, 36 p. (2022; Zbl 1484.52016)], who proved in particular that for the edge combination~\(a^2b^2c\), the only tilings are three two-parameter families of \textit{pentagonal subdivisions} \(PP_4\), \(PP_8\), \(PP_{20}\) of the Platonic solids (tetrahedron, octahedron, icosahedron), producing tilings with \(12\), \(24\), and \(60\)~tiles respectively. The case~\(a^3b^2\) was handled by \textit{E. Wang} and \textit{M. Yan} [Adv. Math. 394, Article ID 107867, 68 p. (2022; Zbl 1484.52017)] and the case~\(a^5\) by \textit{Y. Akama} et al. [Adv. Math. 394, Article ID 107881, 41 p. (2022; Zbl 1484.52014)]. The remaining and most difficult case \(a^4b\) is treated by \textit{Hoi Ping Luk} and \textit{Min Yan} [``Tilings of the sphere by congruent pentagons. IV: Edge combination $a^4b$'', Preprint, \url{arXiv:2307.11453}] and further refined\Nin \textit{Junjie Shu} et al. [``Tilings of the sphere by congruent pentagons. IV: Edge combination $a^4b$ with general angles'', Preprint, \url{arXiv:2412.08492}] and \textit{Jinjin Liang} et al. [``Tilings of the sphere by congruent pentagons. V: Edge combination $a^{4}b$ with rational angles'', Preprint, \url{arXiv:2507.07038}]; these last three are currently preprints, but the upshot is that the full classification of edge-to-edge tilings of the sphere by congruent pentagons is now essentially complete.\N\NAmong all the spherical tilings by congruent polygons that have been classified, exactly five families admit two free parameters: the\Ntetrahedron~\(P_4\), the first quadrilateral earth map tiling \(E_{\square}1\), and the three pentagonal subdivisions \(PP_4\), \(PP_8\), \(PP_{20}\). The moduli space of~\(P_4\) is elementary (all spherical triangles with angle sum~\(2\pi\)), and that of~\(E_{\square}1\) was described in [\textit{Ho Man Cheung} et al., ``Tilings of the sphere by congruent quadrilaterals or triangles'', Preprint, \url{arXiv:2204.02736}, Figure~51] and [\textit{Y. Liao} et al., Chin. Ann. Math., Ser. B 45, No. 5, 733--766 (2024; Zbl 1558.52031), Figure~13]. The important question that remained was to give a precise geometric description of the moduli spaces of the three pentagonal subdivision families. This is the problem addressed by the paper under review.\N\NTo begin with, the authors recall the pentagonal subdivision construction. One fixes an \emph{anchor point} \(V \in S^2\) within\Na~regular triangular face~\(F\) of a Platonic solid, with centre~\(A\), vertex~\(B\), and edge midpoint~\(C\). The pentagon is\Nconstructed by connecting~\(V\) to~\(A\) and to~\(B\) by great arcs of length~\(<\pi\) to produce edges \(a_1\) and~\(b_1\); rotating~\(a_1\) around~\(A\) by~\(-\tfrac{2}{3}\pi\) to produce~\(a_2\); rotating~\(b_1\) around~\(B\) by~\(\tfrac{2}{n}\pi\) to produce~\(b_2\) (where \(n = 3, 4, 5\) for tetrahedron, octahedron, icosahedron); and connecting the free ends of~\(a_2\) and~\(b_2\) to~\(C\) to form a single arc~\(c\) with \(C\)~as its midpoint. The resulting congruent pentagons have edge lengths\N\(a, a, b, b, c\). A tiling of the sphere by congruent polygons requires each tile to be simple -- its boundary must be a simple closed\Ncurve. The moduli space is then defined as the set of all anchor points \(V \in S^2\) for which the resulting pentagon is simple. Since\Nthe tiling is entirely determined by~\(V\), the moduli space is an open subset of~\(S^2\), and each of \(PP_4\), \(PP_8\), \(PP_{20}\) constitutes a two-real-parameter family of tilings.\N\NThe main theorem of the paper gives a complete description of this moduli space. Concretely, the moduli space is an open region of the sphere bounded by exactly five components: two arcs of great circles (the boundary arcs \(\Omega_3 \cap \Omega_9\) and \(\Omega_7 \cap \Omega_9\) in a natural decomposition of the sphere into \(6n\) triangular regions), and three algebraic curves \(\gamma_A\), \(\gamma_B\), \(\gamma_C\). The curve~\(\gamma_A\) lies inside~\(\Omega_5\) and connects the two endpoints of the arc~\(\Omega_3 \cap \Omega_5\); \(\gamma_B\) lies inside~\(\Omega_{13}\) (or \(\Omega_{13} \cup\Omega_{19}\) for the icosahedron) and connects the two endpoints of~\(\Omega_7 \cap \Omega_{13}\); and \(\gamma_C\) passes through \(\Omega_4\) and \(\Omega_8\) (or \(\Omega_8 \cup \Omega_{14}\) for the icosahedron), connecting~\(A\) to~\(C\) to~\(B\). Each boundary curve\Nadmits two equivalent geometric descriptions: it is the intersection of a hyperbolic cylinder with~\(S^2\), and equivalently the intersection of a quadratic cone with~\(S^2\).\N\NThe key observation that drives the boundary analysis is that each curve is characterised by a concyclicity condition: four points\N\((P, w, v, P^*)\), where \(P\) and~\(P^*\) are an antipodal pair, \(v\)~is the anchor point, and \(w\)~is its image under a rotation, must lie on a common great circle. As it turns out, this condition yields a degree-3 algebraic curve in the Cartesian stereographic\Ncoordinates~\((x, y)\), whose general form in an \(X\)-projection is \N\[\N2\lambda(x^2 + y^2) + (x \cos\alpha + y \sin\alpha)(x^2 + y^2 - 1) = 0,\N\]\Nwhere \(\lambda = \tfrac{1}{2}(d^{-1} - d)\cos\alpha > 0\) depends on the specific curve and Platonic solid. In polar form \(v = re^{i\phi}\), this equation factors (after dividing out~\(r > 0\)) into a quadratic in~\(r\), namely \(\cos(\phi - \alpha)\, r^2 + 2\lambda\, r - \cos(\phi - \alpha) = 0\), so that each boundary curve can be expressed explicitly as\N\[\Nr = \sqrt{\lambda^2 \sec^2(\phi - \alpha) + 1}\;-\; \lambda \sec(\phi - \alpha).\N\] \NOn the sphere, the equation is quadratic in the ambient coordinates~\((\xi_1, \xi_2, \xi_3)\), which directly yields the quadratic-cone interpretation. Full explicit formulae are provided for all three curves, in all three preferred stereographic projections (\(A\)-, \(B\)-, and \(M\)-projections), for each of the three Platonic solids. To obtain the visually most informative \(M\)-projection formulae, the authors apply explicit Möbius transforms; the Cartesian equations in this projection are degree-4 curves, the degree having increased from~3 due to the rational substitution.\N\NIn fact, the proof that these curves constitute the full boundary of the moduli space requires a two-part argument. The first part is a\Nregion-by-region exhaustion (Section~1): the sphere is divided into \(6n\) triangular regions, and for most of them the pentagon is shown to be either always simple or never simple, by tracking how the arcs \(a_1, a_2, b_1, b_2, c\) pass through the regions. The authors employ a geometric lemma (Lemma~1), which gives angle-based sufficient conditions for two arcs to intersect on the sphere, to rule out several additional regions. This leaves a small set of ``ambiguous regions'' -- four for the tetrahedron and octahedron (\(\Omega_4, \Omega_5, \Omega_8, \Omega_{13}\)), and six for the icosahedron (additionally \(\Omega_{14}\) and \(\Omega_{19}\)) -- through which the moduli space boundary passes. The second part establishes a monotonicity property (Lemma~2): for any fixed angular direction from the projection centre, the critical point on the boundary curve is unique, and the moduli space on that ray is precisely the open segment between the origin and the critical point. Once these two ingredients are in place, the calculation of the boundary reduces to the algebraic condition described above.\N\NThe proof strategy derives strong inspiration from the systematic approach to spherical tilings developed in the earlier papers of the\Nseries, particularly [\textit{E. Wang} and \textit{M. Yan}, Adv. Math. 394, Article ID 107866, 36 p. (2022; Zbl 1484.52016); \textit{Ho Man Cheung} et al., ``Tilings of the sphere by congruent quadrilaterals or triangles'', Preprint, \url{arXiv:2204.02736}]. What is genuinely new is the complete and explicit parametrisation of the moduli space boundary in closed algebraic form, together with the use of Möbius transforms to unify the pictures in the \(M\)-projection.\N\NBeyond the main theorem, the paper establishes several additional results that flesh out the geometry of the moduli space.\N\N\textit{Edge-length bounds} (Theorem~3). The suprema of the three edge lengths \(a\), \(b\), \(c\) are computed to full algebraic precision. For the tetrahedron, \(\sup a = \sup b \approx 0.6259\pi\) (by a symmetry that exchanges \(A\) and~\(B\)) and \(\sup c = \pi\); for the octahedron, \(\sup b = \tfrac{1}{2}\pi\) exactly; for the icosahedron, \(\sup b =\arccos(1/\!\sqrt{5})\). In general, \(\sup a\) and~\(\sup b\) are realised by the maximum spherical distance from~\(A\) to~\(\gamma_B\) and from~\(B\) to~\(\gamma_A\) respectively, whose computation reduces to finding critical points of~\(R(\theta)\) via \(R'(\theta) = 0\) and yields algebraic expressions for \(\tan\theta_{AB}\) involving cube roots. The supremum of~\(c\) is twice the maximum spherical distance from the midpoint~\(M\) to~\(\gamma_C\); for \(n = 3\) this gives \(\sup c = \pi\) (since \(R_C\) vanishes at \(\theta_{MC} = \tfrac{5}{4}\pi\)), while for \(n = 4\) and \(n = 5\) it yields \(\sup c \approx 0.6882\pi\) and~\(0.4358\pi\) respectively.\N\N\emph{Size of the moduli space} (Section~4.5). The areas of the three moduli spaces are expressed in closed form using elliptic\Nintegrals of the first kind \(F(t, k)\) with purely imaginary modulus~\(k = \lambda^{-1}i\). Numerically, the moduli spaces occupy\Napproximately \(21.5\%\), \(11.5\%\), and \(4.9\%\) of the total area of~\(S^2\) for the tetrahedron, octahedron, and icosahedron\Nrespectively.\N\N\textit{Reduction curves} (Section~4.4). When edge lengths coincide (\(a = b\), \(a = c\), or \(b = c\)), the edge combination \(a^2b^2c\) degenerates. The locus \(a = b\) is the intersection of~\(S^2\) with a plane (a great circle for \(n = 3\)); the loci \(a = c\) and \(b = c\) are intersections with parabolic cylinders. Complete equations are given. What is noteworthy in this context is that for \(n = 4\), the curve \(b = c\) is tangent to the boundary curve~\(\gamma_A\), confirmed by Taylor expansion. The uniqueness of the equilateral pentagonal subdivision (\(a = b = c\)) for each Platonic solid -- rigorously established in [\textit{Y. Akama} et al., Adv. Math. 394, Article ID 107881, 41 p. (2022; Zbl 1484.52014)] -- is consistent with the intersection behaviour of these\Nreduction curves.\N\N\textit{Companion and boundary location} (Sections~4.1 and~4.2). The companion construction, introduced in [\textit{E. Wang} and \textit{M. Yan}, Adv. Math. 394, Article ID 107866, 36 p. (2022; Zbl 1484.52016), Section~3.1], associates to each anchor\Npoint~\(V\) its reflection~\(V'\) across the arc~\(BB'\). The paper proves that both \(V\) and~\(V'\) lie in the moduli space if and only if \(V\) lies in the region bounded by the arc~\(AA'\), the curve~\(\gamma_B\), and the reflected curve~\(\gamma_B'\); the key inequality needed is \(R_B(2\pi - \rho) \ge R_C(\pi + \rho)\) for \(\rho \in [0, \tfrac{1}{2}\pi]\). In addition, refined locations of the boundary curves are given: for all three Platonic solids, \(\gamma_A\) is shown to lie inside a specific sub-triangle \(\triangle AB'X \subset \Omega_5\), and for \(n = 4, 5\) the arc~\(M'C_1\) is tangential to~\(\gamma_C\) at~\(C\). It is also worth noting that \(\gamma_C\) is shown to be a single algebraic curve spanning both~\(\Omega_4\) and~\(\Omega_8\): the two portions, computed via the \(A\)-projection and \(B\)-projection respectively, yield the same curve because the corresponding parameters satisfy \(\lambda_C = \tfrac{1}{2}(d_{AM}^{-1} - d_{AM}) \cos \tfrac{1}{3}\pi = \tfrac{1}{2}(d_{BM}^{-1} - d_{BM})\cos \tfrac{1}{n}\pi\).\N\NThe paper is carefully written and the geometric arguments are complemented by real 3D pictures and perspective stereographic\Nprojection diagrams. All non-decimal values are stated to be exact, and all computations are carried out by symbolic algebra software. On the other hand, three minor typographical errors should be noted: (i)~on page~2, ``relatively few is known'' should read ``relatively little is known''; (ii)~on page~3, the definition ``Let \(c_1\) be the arc between \(A\) and the end of~\(a_2\) other than~\(A\)'' should read ``between \(C\) and the end of~\(a_2\) other than~\(A\),'' since as written \(c_1 = a_2\), contradicting both Figure~1 and the parallel definition of~\(c_2\); and (iii)~on page~3, ``mean'' should be ``means'' in the sentence defining the decimal convention for Table~3. None of these affect the mathematical content, since the correct reading of~(ii) is unambiguous from the figure and surrounding context.\N\NThe main takeaway of this work is the complete and explicit description of the moduli spaces of the three pentagonal subdivision\Nfamilies, which are the most intricate two-parameter families among all edge-to-edge tilings of the sphere by congruent polygons. The results effectively close the geometric chapter of the classification programme for spherical pentagonal tilings. The techniques are, for the most part, natural extensions of the methods developed across the series -- stereographic projections, region-by-region analysis, concyclicity conditions -- but the synthesis into explicit algebraic formulae and the area computations via elliptic integrals are carried out with considerable care and completeness. The connection to Tao's open problem on MathOverflow, asking whether the sphere can be partitioned into arbitrarily small-diameter congruent cells, is mentioned as motivation, though the paper does not make direct progress on this question.