This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

The 10 antipodal pairings of strongly involutive polyhedra on the sphere

Javier Bracho Instituto de Matemáticas, UNAM, Ciudad Universitaria, México, CP. 04510 [email protected] Luis Montejano Instituto de Matemáticas, UNAM, Campos Juriquilla, Querétaro, México, CP. 07360 [email protected] Eric Pauli Instituto de Matemáticas, UNAM, Campos Juriquilla, Querétaro, México, CP. 07360 [email protected]  and  Jorge L. Ramírez Alfonsín IMAG, Univ. Montpellier, CNRS, Montpellier, France [email protected]
Abstract.

It is known that strongly involutive polyhedra are closely related to self-dual maps where the antipodal function acts as duality isomorphism. Such a family of polyhedra appears in different combinatorial, topological and geometric contexts, and is thus attractive to be studied. In this note, we determine the 10 antipodal pairings among the classification of the 24 self-dual pairings Dual(G)Aut(G)Dual(G)\rhd Aut(G) of self-dual maps GG. We also present the orbifold associated to each antipodal pairing and describe explicitly the corresponding fundamental regions. We finally explain how to construct two infinite families of strongly involutive polyhedra (one of them new) by using their doodles and the action of the corresponding orbifolds.

1. Introduction

A polyhedron is a planar graph GG that is simple (without loops and multiple edges) and 3-connected. A face FF of an embedding of a planar graph is a region bounded by a cycle CC. We say that a vertex vv belongs to a face FF, denoted by vFv\in F, if vv is in the cycle CC. Note that every edge borders exactly two faces. The dual of GG, denoted by GG^{*}, is defined as follows: Each face in GG is a vertex in GG^{*} and two vertices in GG^{*} are adjacent if and only if the faces share an edge in GG. A polyhedron GG is said to be self-dual if there is an isomorphism of graphs τ:GG\tau:G\rightarrow G^{*}. This isomorphism is called a duality isomorphism, and it will be called a strong involution if it satisfies the following conditions:

(i)(i) For each pair u,vu,v of vertices, uτ(v)u\in\tau(v) if and only if vτ(u)v\in\tau(u) and

(ii)(ii) For every vertex vv, we have that vτ(v)v\notin\tau(v).

We say that a self-dual polyhedron is strongly involutive if it admits a strong involution. The above conditions are the combinatorial counterpart (in the 3-dimensional case) of a more general geometric object called strongly self-dual polytopes, introduced by Lovász [9], see also [5]. Strongly involutive polyhedra are relevant in different contexts: they are related to convex geometric problems such as the well-known Vázsonyi’s problem [4, 6], with ball polyhedra [1, 7, 16], the chromatic number of distance graphs on the sphere [9], the Reuleaux polyhedra [8, 11], constant width bodies [10] and more recently in relation with questions concerning the symmetry as well as the amphicheirality of knots [13] and projective links [14].

There are known infinite families of strongly involutive polyhedra [12, Propositions 1, 2 and 3 and Theorem 3]. Figure 1 illustrates two simple examples.

Refer to caption
Figure 1. (Left) 5-wheel. (Right) 4-hyperwheel.

Strongly involutive isomorphisms are closely related to the notion of antipodal isometries in the 2-sphere.

A map G=(V,E,F)G=(V,E,F) is the image of an embedding of GG into 𝕊2\mathbb{S}^{2} where the set of vertices is a collection of distinct points in 𝕊2\mathbb{S}^{2} and the set of edges is a collection of arcs joining pairs of points in VV satisfying that for any pair of arcs, say aa and aa^{\prime}, we have that aaa\cap a^{\prime} is either empty or an endpoint of one of the arcs. Any embedding of the topological realization of GG into 𝕊2\mathbb{S}^{2} partitions the 2-sphere into simply connected regions of 𝕊2G\mathbb{S}^{2}\setminus G called the faces FF of the embedding.

Recall that Aut(G)\mbox{Aut}(G) is the group formed by the set of all automorphisms of GG (i.e., the set of isomorphisms of GG into itself) and let Iso(G)\mbox{Iso}(G) be the set of all duality isomorphisms of GG into GG^{*}. We notice that Iso(G)\mbox{Iso}(G) is not a group since the composition of any two of them is an automorphism. Let us suppose that G=(V,E,F)G=(V,E,F) is a self-dual map so that there is an isomorphism ϕ:(V,E,F)(F,E,V)\phi:(V,E,F)\rightarrow(F^{*},E^{*},V^{*}). Following ϕ\phi with the correspondence * gives a permutation on VEFV\cup E\cup F which preserve incidences but reverses dimension of the elements. The collection of all such permutations generates a group Dual(G)=Aut(G)Iso(G)\mbox{Dual}(G)=\mbox{Aut}(G)\cup\mbox{Iso}(G) in which the automorphism group Aut(G)\mbox{Aut}(G) is contained as a subgroup of index 2.

It is known [15, Lemma 1] that for a given map GG there is a homeomorphism ρ\rho of 𝕊2\mathbb{S}^{2} to itself such that for every σAut(G)\sigma\in\mbox{Aut}(G) we have that ρσIsom(𝕊2)\rho\sigma\in\mbox{Isom}(\mathbb{S}^{2}) where Isom(𝕊2)\mbox{Isom}(\mathbb{S}^{2}) is the group of isometries of the 2-sphere. In other words, any planar graph GG can be drawn on the 2-sphere such that any automorphism of the resulting map acts as an isometry of the sphere. This was extended in [15] by showing that given any self-dual planar graph GG and its dual GG^{*} can be drawn on the 2-sphere so that Dual(G)\mbox{Dual}(G) is realized as a group of spherical isometries.

In [2, Theorem 9] (see also [12, Corollary 2]), it was proven that if GG is strongly involutive then the corresponding isometry is the antipodal function α:𝕊2𝕊2{\alpha:\mathbb{S}^{2}\rightarrow\mathbb{S}^{2}}, α(x)=x\alpha(x)=-x.

The couple Dual(G)Aut(G)\mbox{Dual}(G)\rhd\mbox{Aut}(G) is called the self-dual pairing of the map GG. In [15], all self-dual maps were enumerated and classified. In the notation of [3, page 36-38] the 24 possible pairings are:

- among the infinite classes [2,q][q],[2,q]+[q]+,[2+,2q][2q],[2,q+][q]+[2,q]\rhd[q],[2,q]^{+}\rhd[q]^{+},[2^{+},2q]\rhd[2q],[2,q^{+}]\rhd[q]^{+} and [2+,2q+][2q]+[2^{+},2q^{+}]\rhd[2q]^{+} with qq positive integer, or

- among the special pairings [2][1],[2][2]+,[4][2],[2]+[1]+,[4]+[2]+,[2,2][2,2]+,[2,4][2+,4],[2,2][2,2+],[2,4][2,2],[2,4]+[2,2]+,[2+,4][2,2]+,[2+,4][2+,4+],[2,4+][2+,4+],[2,2+][2+,2+],[2,4+][2,2+],[2,2+][1],[3,4][3,3],[3,4]+[3,3]+[2]\rhd[1],[2]\rhd[2]^{+},[4]\rhd[2],[2]^{+}\rhd[1]^{+},[4]^{+}\rhd[2]^{+},[2,2]\rhd[2,2]^{+},[2,4]\rhd[2^{+},4],[2,2]\rhd[2,2^{+}],[2,4]\rhd[2,2],[2,4]^{+}\rhd[2,2]^{+},[2^{+},4]\rhd[2,2]^{+},[2^{+},4]\rhd[2^{+},4^{+}],[2,4^{+}]\rhd[2^{+},4^{+}],[2,2^{+}]\rhd[2^{+},2^{+}],[2,4^{+}]\rhd[2,2^{+}],[2,2^{+}]\rhd[1],[3,4]\rhd[3,3],[3,4]^{+}\rhd[3,3]^{+} and [3+,4][3,3]+[3^{+},4]\rhd[3,3]^{+}.

In Section 2, we classify the self-dual pairings Dual(G)Aut(G)\mbox{Dual}(G)\rhd\mbox{Aut}(G) that correspond to antipodal isometries, that is, self-dual pairings that give rise to strongly involutive polyhedra.

Theorem 1.

Every strongly involutive map corresponds to one of the following self-dual pairings: [q][2,q],[q]+[2,q+][q]\triangleleft[2,q],[q]^{+}\triangleleft[2,q^{+}] for qq even; [q][2+,2q],[q]+[2+,2q+]{[q]\triangleleft[2^{+},2q]},{[q]^{+}\triangleleft[2^{+},2q^{+}]} for qq odd, [2,2]+[2,2],[2+,4][2,4],[2+,4+][2,4+],[1][2,2+],[3,3][3,4] or [3,3]+[3+,4][2,2]^{+}\triangleleft[2,2],[2^{+},4]\triangleleft[2,4],[2^{+},4^{+}]\triangleleft[2,4^{+}],[1]\triangleleft[2,2^{+}],[3,3]\triangleleft[3,4]\mbox{ or }[3,3]^{+}\triangleleft[3^{+},4].

The self-dual pairings given in Theorem 1 will be called antipodal pairings. We shall also present a short description of the orbifold associated to each antipodal pairing. We do so by explicitly giving the fundamental region for each group and its generators.

In Section 3, we study the doodle (the map restricted to its fundamental region) of two special families of strongly involutive polyhedra: the qq-multi hyperwheels (a new family of strongly involutive polyhedra) and the qq-multi wheels. We explain how these families can be constructed from their doodle via the action of the corresponding orbifold. We finally end with some concluding remarks.

2. Antipodal pairings

2.1. Spherical isometries

We quickly recall some notions on the group of spherical isometries (we refer the reader to [3] for further details and where our notation is taken from). An isometry of 𝕊2\mathbb{S}^{2} is either a rotation or a reflection or a composition of these two. Let Isom(𝕊2)\mbox{Isom}(\mathbb{S}^{2}) be the group of isometries of the 2-sphere. We remark that any orientation reversing isometry is either a reflection or a product of three reflections. Also, it can be checked that Isom(𝕊2)\mbox{Isom}(\mathbb{S}^{2}) is generated by only reflections. Moreover, any isometry is a product of at most three reflections:

\bullet [one reflection] Let HH be a plane passing through the center of 𝕊2\mathbb{S}^{2}. The reflection γ\gamma w.r.t. HH is an involutive isometry and therefore of order 2. This group is denoted by [1][1].

\bullet [two reflections] Let H1H_{1} and H2H_{2} be two planes with angle (H1,H2)=πq\angle(H_{1},H_{2})=\frac{\pi}{q}. The reflection γ1\gamma_{1} and γ2\gamma_{2} w.r.t. these planes generate the group [q][q] (isomorphic to the dihedral group DqD_{q}) having presentation γ12=γ22=(γ1γ2)q=id\gamma_{1}^{2}=\gamma_{2}^{2}=(\gamma_{1}\gamma_{2})^{q}=id.

\bullet [three reflections] Let H1,H2H_{1},H_{2} and H3H_{3} be three planes inducing a spheric triangle with angles πp,πq\frac{\pi}{p},\frac{\pi}{q} and πr\frac{\pi}{r}, denoted by [p,q][p,q].

Refer to caption
Figure 2. Reflection groups.

It is known that the sum of the angles of a spheric triangle is strictly larger that π\pi, therefore 1p+1q+1r>1\frac{1}{p}+\frac{1}{q}+\frac{1}{r}>1 implying that min{p,q,r}=2\min\{p,q,r\}=2. We take r=2r=2, obtaining that 1p+1q>12\frac{1}{p}+\frac{1}{q}>\frac{1}{2} yielding to (p2)(q2)<4(p-2)(q-2)<4. By analyzing the latter, it can be found that the possible cases are :

- [3,3][3,3] (corresponding to the tetrahedron),

- [3,4][3,4] (corresponding to the cube and octahedron),

- [3,5][3,5] (corresponding to the dodecahedron and icosahedron),

- [2,q][2,q] (corresponding to a diamond polyhedron which dual is a prism with a qq-polygon as a base).

The group [p,q][p,q] or [q,p][q,p] is defined by γ12=γ22=γ32=(γ1γ2)p=(γ2γ3)q=(γ1γ3)2=id\gamma_{1}^{2}=\gamma_{2}^{2}=\gamma_{3}^{2}=(\gamma_{1}\gamma_{2})^{p}=(\gamma_{2}\gamma_{3})^{q}=(\gamma_{1}\gamma_{3})^{2}=id having as a fundamental region the spherical triangle with angles πp,πq\frac{\pi}{p},\frac{\pi}{q} and π2\frac{\pi}{2}. Each of these groups has a rotational group consisting only of the possible rotations generated by pairs of reflections:

- The rotational subgroup of [q][q], consisting of the powers of γ1γ2\gamma_{1}\gamma_{2}, is denoted by [q]+[q]^{+} (clearly [1]+[1]^{+} is the trivial group),

- the rotational subgroup of [p,q][p,q], generated by the rotations γ1γ2,γ2γ3\gamma_{1}\gamma_{2},\gamma_{2}\gamma_{3} and γ3γ1\gamma_{3}\gamma_{1}, is denoted by [p,q]+[p,q]^{+} (these subgroups are of order 2),

- the rotational subgroup of [p,q][p,q] when qq is even, generated by γ1γ2\gamma_{1}\gamma_{2} which is a rotation of order pp followed by the reflection γ3\gamma_{3}, is denoted by [p+,q][p^{+},q] or [q,p+][q,p^{+}] (these subgroups are of order 2),

- the rotational subgroup of [2+,q][2^{+},q], generated by γ1γ2γ3\gamma_{1}\gamma_{2}\gamma_{3}, is denoted by [2+,q+][2^{+},q^{+}] (which turns out to be a cyclic group of index 2) generated by the rotatory reflection ργ3\rho\gamma_{3}.

2.2. Detecting antipodal pairings

We first present the following lemma that identifies which groups contain the antipodal function.

Lemma 1.

Let α:𝕊2𝕊2\alpha:\mathbb{S}^{2}\rightarrow\mathbb{S}^{2} be the antipodal function. Then,

  1. (1)

    α[q]\alpha\notin[q] for any positive integer qq.

  2. (2)

    α[q]+\alpha\notin[q]^{+} for any positive integer qq.

  3. (3)

    α[p,q]\alpha\in[p,q] if and only if p=3,q=4p=3,q=4 or p=2p=2 and qq is even.

  4. (4)

    α[p,q]+\alpha\notin[p,q]^{+} for any positive integers pp and qq.

  5. (5)

    α[p+,q]\alpha\in[p^{+},q] or [q,p+][q,p^{+}] if and only if p=3,q=4p=3,q=4 or p=2p=2, qq is even and q2\frac{q}{2} is odd.

  6. (6)

    α[p+,q+]\alpha\in[p^{+},q^{+}] if and only if p=2p=2, qq is even and q2\frac{q}{2} is odd.

Proof. Since α\alpha is an isometry without fixed points then (1) is verified. Moreover, since α\alpha inverts orientation then (2) and (4) hold. Notice that (2) also follows since [q]+[q]^{+} is a subgroup of [q][q]. We can establish (3) by observing that any prism has central symmetry if and only if it has a polygon with an even number of sides as a base and similarly we see that the cube and the octahedron are the only platonic solids on the list with central symmetry.

In order to illustrate the proof of (5) and (6), let us see that α[2+,6]\alpha\in[2^{+},6]. The group [2,6][2,6] is generated by the reflections γ1,γ2\gamma_{1},\gamma_{2} and γ3\gamma_{3} such that γ12=γ22=γ32=(γ1γ2)2=(γ3γ1)2=(γ2γ3)2=id\gamma_{1}^{2}=\gamma_{2}^{2}=\gamma_{3}^{2}=(\gamma_{1}\gamma_{2})^{2}=(\gamma_{3}\gamma_{1})^{2}=(\gamma_{2}\gamma_{3})^{2}=id. The subgroup [2+,6][2^{+},6] is generated by the reflection γ=γ1\gamma=\gamma_{1} and the rotation ρ=γ2γ3\rho=\gamma_{2}\gamma_{3}. The element ργ\rho\gamma is a rotatory reflection or a step, satisfying (ργ)6=id(\rho\gamma)^{6}=id and (ργ)3=α(\rho\gamma)^{3}=\alpha (See Figure 3).

Refer to caption
Figure 3. Zig-zag described by (ργ)i(\rho\gamma)^{i} in [2+,6][2^{+},6]. The exterior dotted circle represents one of the poles, the central black circle the other pole and the middle dotted circle the equator.

In the general case [2+,q][2^{+},q], if qq is even and q2\frac{q}{2} is odd, then we have that (ργ)q=id(\rho\gamma)^{q}=id and (ργ)q2=α(\rho\gamma)^{\frac{q}{2}}=\alpha. Conversely, if (ργ)q2=α(\rho\gamma)^{\frac{q}{2}}=\alpha, then qq is even and q2\frac{q}{2} must be odd in order to express α\alpha as a product of generators with an odd number of reflections, otherwise α\alpha would preserve orientation, which is not the case. From this fact we can conclude also 6 since [2+,q+][2^{+},q^{+}] is precisely the cyclic group generated by the step ργ\rho\gamma. Finally, by considering (3), we just need to check the subgroup [3+,4][3^{+},4], which is generated by a reflection γ\gamma and a rotation ρ\rho. In this case, it can be verified that α=(ργ)3\alpha=(\rho\gamma)^{3}. \square

We may now prove Theorem 1.

Proof of Theorem 1. We may use Lemma 1 to identify those self-dual pairings Dual(G)Aut(G)\mbox{Dual}(G)\rhd\mbox{Aut}(G) such that αDual(G)\alpha\in\mbox{Dual}(G) and αAut(G)\alpha\notin\mbox{Aut}(G). In Table 1 we examine each case. \square

Dual(G)(G)\rhdAut(G)(G) α\alpha\in Dual(G(G)? α\alpha\in Aut(G)(G)?
[𝟐,𝐪][𝐪]\mathbf{[2,q]\rhd[q]}, qq-even α[2,q]\alpha\in[2,q] (3) α[q]\alpha\not\in[q] (1)
[2,q]+[q]+[2,q]^{+}\rhd[q]^{+} α[2,q]+\alpha\not\in[2,q]^{+} (4) α[q]+\alpha\not\in[q]^{+} (2)
[𝟐+,𝟐𝐪][𝟐𝐪]\mathbf{[2^{+},2q]\rhd[2q]}, qq-odd α[2+,2q]\alpha\in[2^{+},2q] (5) α[2q]\alpha\not\in[2q] (1)
[𝟐,𝐪+][𝐪]+\mathbf{[2,q^{+}]\rhd[q]^{+}}, qq-even α[2,q+]\alpha\in[2,q^{+}] (5) α[q]+\alpha\not\in[q]^{+} (2)
[𝟐+,𝟐𝐪+][𝟐𝐪]+\mathbf{[2^{+},2q^{+}]\rhd[2q]^{+}}, qq-odd α[2+,2q+]\alpha\in[2^{+},2q^{+}] (6) α[2q]+\alpha\not\in[2q]^{+} (2)
[2][1][2]\rhd[1] α[2]\alpha\not\in[2] (1) α[1]\alpha\not\in[1] (1)
[2][2]+[2]\rhd[2]^{+} α[2]\alpha\not\in[2] (1) α[2]+\alpha\not\in[2]^{+} (2)
[4][2][4]\rhd[2] α[4]\alpha\not\in[4] (1) α[2]\alpha\not\in[2] (1)
[2]+[1]+[2]^{+}\rhd[1]^{+} α[2]+\alpha\not\in[2]^{+} (2) α[1]+\alpha\not\in[1]^{+} (1)
[4]+[2]+[4]^{+}\rhd[2]^{+} α[4]+\alpha\not\in[4]^{+} (2) α[2]+\alpha\not\in[2]^{+} (2)
[𝟐,𝟐][𝟐,𝟐]+\mathbf{[2,2]\rhd[2,2]^{+}} α[2,2]\alpha\in[2,2] (3) α[2,2]+\alpha\not\in[2,2]^{+} (4)
[𝟐,𝟒][𝟐+,𝟒]\mathbf{[2,4]\rhd[2^{+},4]} α[2,4]\alpha\in[2,4] (3) α[2+,4]\alpha\not\in[2^{+},4] (5)
[2,2][2,2+][2,2]\rhd[2,2^{+}] α[2,2]\alpha\in[2,2] (3) α[2,2+]\alpha\in[2,2^{+}] (5)
[2,4][2,2][2,4]\rhd[2,2] α[2,4]\alpha\in[2,4] (3) α[2,2]\alpha\in[2,2] (3)
[2,4]+[2,2]+[2,4]^{+}\rhd[2,2]^{+} α[2,4]+\alpha\not\in[2,4]^{+} (4) α[2,2]+\alpha\not\in[2,2]^{+} (4)
[2+,4][2,2]+[2^{+},4]\rhd[2,2]^{+} α[2+,4]\alpha\not\in[2^{+},4] (5) α[2,2]+\alpha\not\in[2,2]^{+} (4)
[2+,4][2+,4+][2^{+},4]\rhd[2^{+},4^{+}] α[2+,4]\alpha\not\in[2^{+},4] (5) α[2+,4+]\alpha\not\in[2^{+},4^{+}] (6)
[𝟐,𝟒+][𝟐+,𝟒+]\mathbf{[2,4^{+}]\rhd[2^{+},4^{+}]} α[2,4+]\alpha\in[2,4^{+}] (5) α[2+,4+]\alpha\not\in[2^{+},4^{+}] (6)
[2,2+][2+,2+][2,2^{+}]\rhd[2^{+},2^{+}] α[2,2+]\alpha\in[2,2^{+}] (5) α[2+,2+]\alpha\in[2^{+},2^{+}] (6)
[2,4+][2,2+][2,4^{+}]\rhd[2,2^{+}] α[2,4+]\alpha\in[2,4^{+}] (5) α[2,2+]\alpha\in[2,2^{+}] (5)
[𝟐,𝟐+][𝟏]\mathbf{[2,2^{+}]\rhd[1]} α[2,2+]\alpha\in[2,2^{+}] (5) α[1]\alpha\not\in[1] (1)
[𝟑,𝟒][𝟑,𝟑]\mathbf{[3,4]\rhd[3,3]} α[3,4]\alpha\in[3,4] (3) α[3,3]\alpha\not\in[3,3] (3)
[3,4]+[3,3]+[3,4]^{+}\rhd[3,3]^{+} α[3,4]+\alpha\not\in[3,4]^{+} (4) α[3,3]+\alpha\not\in[3,3]^{+} (4)
[𝟑+,𝟒][𝟑,𝟑]+\mathbf{[3^{+},4]\rhd[3,3]^{+}} α[3+,4]\alpha\in[3^{+},4] (5) α[3,3]+\alpha\not\in[3,3]^{+} (4)
Table 1. Detecting if the antipodal map α\alpha belongs to Aut(G)\mbox{Aut}(G) and/or Dual(G)\mbox{Dual}(G). In parenthesis the condition of Lemma 1 that is applied. In bold the self-dual pairings with αDual(G)\alpha\in\mbox{Dual}(G) and αAut(G)\alpha\notin\mbox{Aut}(G).

2.3. Orbifolds

Given σDual(G)\sigma\in\mbox{Dual}(G) and a point p𝕊2p\in\mathbb{S}^{2} we have that σp=σ(p)\sigma\cdot p=\sigma(p), that is, Dual(G)\mbox{Dual}(G) acts as an evaluation function in 𝕊2\mathbb{S}^{2}. The orbit of RR is the set of regions congruent to RR covering the sphere (the number of such regions is as many as the elements in Dual(G)\mbox{Dual}(G). The stabilizer of pp consist of all the elements in Dual(G)\mbox{Dual}(G) fixing pp. An orbifold is defined as the fundamental region RR together with the points with nontrivial stabilizer.

Let ΓΔ\Gamma\triangleleft\Delta be a self-duality pairing where Γ\Gamma and Δ\Delta are finite groups of isometries of the sphere, such that Γ\Gamma is an index 2 subgroup of Δ\Delta. In order to describe each orbifold we need fundamental regions R1R_{1} and R2R_{2} corresponding to the groups Γ\Gamma and Δ\Delta, respectively as well as their corresponding generators. We observe that since Δ\Delta is an extension of index 2, region R2R_{2} is contained in R1R_{1} and it will have half of its area.

In what follows we describe the fundamental regions for each antipodal self-dual pairing. We have drawn with :

- a thick line the singular elements of Γ\Gamma, i.e., sets of fixed points under the action of Γ\Gamma

- a double line the singular elements of Δ\Delta

- continuous lines denote reflection lines

- dotted lines delimit the regions without considering a reflection and

- points marked with a circle denote a center of rotation by the marked angle.

[𝐪][𝟐,𝐪]\mathbf{[q]\triangleleft[2,q]} [q]=γ1,γ2[q]=\langle\gamma_{1},\gamma_{2}\rangle is a group generated by reflections γ1\gamma_{1} and γ2\gamma_{2} through the thick lines. The fundamental region R1R_{1} is a bigon formed by these lines. [2,q]=γ1,γ2,γ3[2,q]=\langle\gamma_{1},\gamma_{2},\gamma_{3}\rangle is generated by reflections γ1,γ2\gamma_{1},\gamma_{2} and γ3\gamma_{3} and its fundamental region R2R_{2} is a triangle with angles πq,π2,π2\frac{\pi}{q},\frac{\pi}{2},\frac{\pi}{2}. [Uncaptioned image]
[𝐪]+[𝟐,𝐪+]\mathbf{[q]^{+}\triangleleft[2,q^{+}]} The fundamental region R1R_{1} is a bigon formed by lines with an angle 2πq\frac{2\pi}{q} and [q]+=ρ[q]^{+}=\langle\rho\rangle, where ρ\rho is a rotation for 2πq\frac{2\pi}{q}. The extension [2,q+]=ρ,γ[2,q^{+}]=\langle\rho,\gamma\rangle , being γ\gamma the reflection through the equator. [Uncaptioned image]
[𝐪][𝟐+,𝟐𝐪]\mathbf{[q]\triangleleft[2^{+},2q]} The fundamental region R1R_{1} is a bigon with angle πq\frac{\pi}{q} and [q]=γ1,γ2[q]=\langle\gamma_{1},\gamma_{2}\rangle. The region R2R_{2} is a triangle with angles πq,π2,π2\frac{\pi}{q},\frac{\pi}{2},\frac{\pi}{2} and [2+,2q]=ρ,γ1=ρ,γ2[2^{+},2q]=\langle\rho,\gamma_{1}\rangle=\langle\rho,\gamma_{2}\rangle, where ρ\rho is a rotation by π\pi. [Uncaptioned image]
[𝐪]+[𝟐+,𝟐𝐪+]\mathbf{[q]^{+}\triangleleft[2^{+},2q^{+}]} The fundamental region R1R_{1} consists of two triangles disposed in a way one can be obtained from the other by a rotatory reflection. This rotatory reflection is the composition of the reflection through the “I” mark and the rotation around the “C”. Then [q]+=τ2[q]^{+}=\langle\tau^{2}\rangle where τ\tau is the rotatory reflection with order 2q2q, and [2+,2q+]=τ[2^{+},2q^{+}]=\langle\tau\rangle. [Uncaptioned image]
[𝟐,𝟐]+[𝟐,𝟐]\mathbf{[2,2]^{+}\triangleleft[2,2]} In this case R1R_{1} is an equilateral triangle with internal angles of 2π3\frac{2\pi}{3}; [2,2]+=ρ12,ρ23,ρ31[2,2]^{+}=\langle\rho_{12},\rho_{23},\rho_{31}\rangle, ρij\rho_{ij} is a rotation by π\pi in the midpoint of each of the sides and [2,2]=γ1,γ2,γ3[2,2]=\langle\gamma_{1},\gamma_{2},\gamma_{3}\rangle where γi\gamma_{i} is the reflection through a line lil_{i} and ρij=γjγi\rho_{ij}=\gamma_{j}\gamma_{i}. [Uncaptioned image]
[𝟐+,𝟒][𝟐,𝟒]\mathbf{[2^{+},4]\triangleleft[2,4]} [2+,4]=ρ,γ1[2^{+},4]=\langle\rho,\gamma_{1}\rangle and [2,4]=γ1,γ2,γ3[2,4]=\langle\gamma_{1},\gamma_{2},\gamma_{3}\rangle. Here γi\gamma_{i} is a reflection through a line lil_{i} and ρ=γ3γ2\rho=\gamma_{3}\gamma_{2}. [Uncaptioned image]
[𝟐+,𝟒+][𝟐,𝟒+]\mathbf{[2^{+},4^{+}]\triangleleft[2,4^{+}]} [2+,4+]=σ[2^{+},4^{+}]=\langle\sigma\rangle, where σ\sigma is a rotatory reflection of order 4. [2,4+]=γ,ρ[2,4^{+}]=\langle\gamma,\rho\rangle, with γ\gamma a reflection through the equator and ρ\rho a rotation by π2\frac{\pi}{2} in the north pole. Note that σ=ργ\sigma=\rho\gamma. [Uncaptioned image]
[𝟏][𝟐,𝟐+]\mathbf{[1]\triangleleft[2,2^{+}]} In this case R1R_{1} is one hemisphere and [1]=γ[1]=\langle\gamma\rangle, where γ\gamma is the reflection through the equator, and [2,2+]=γ,ρ[2,2^{+}]=\langle\gamma,\rho\rangle, where ρ\rho is the rotation by π\pi through the marked point (north or south pole). [Uncaptioned image]
[𝟑,𝟑][𝟑,𝟒]\mathbf{[3,3]\triangleleft[3,4]} [3,3]=γ1,γ2,γ3[3,3]=\langle\gamma_{1},\gamma_{2},\gamma_{3}\rangle and [3,4]=γ1,γ2,γ4[3,4]=\langle\gamma_{1},\gamma_{2},\gamma_{4}\rangle. Here γ4\gamma_{4} is the reflection through the bisector or the right angle in the first triangle. [Uncaptioned image]
[𝟑,𝟑+][𝟑+,𝟒]\mathbf{[3,3^{+}]\triangleleft[3^{+},4]} [3,3+]=γ1,ρ1[3,3^{+}]=\langle\gamma_{1},\rho_{1}\rangle and [3+,4]=γ2,ρ2[3^{+},4]=\langle\gamma_{2},\rho_{2}\rangle. In this case, R1R_{1} is a quadrilateral. γ1\gamma_{1} is a reflection through one of the thick lines (the other one can be obtained by rotation), and ρ1\rho_{1} is the rotation around the thick circle. R2R_{2} is a triangle and γ2\gamma_{2} is the reflection in the “horizontal” line and ρ2\rho_{2} is the rotation around the circle with double line. [Uncaptioned image]

3. Constructing strongly involutive maps

Let us consider the orbifold of a map GG consisting of a fundamental region RR with the singular points. By construction, we have that the orbit of RR is a set of congruent regions covering the 2-sphere. Given a map GG, we may consider the part of the map GG which is drawn in the fundamental region and does not have any other symmetry. We call this restriction the doodle of GG.

Next we describe how two involutive polyhedra are constructed from their doodle by means of the action of the duality group Dual(G)Aut(G)\mbox{Dual}(G)\rhd\mbox{Aut}(G).

3.1. The qq-multi hyperwheel

Let qq and ll be natural numbers. The qq-multi hyperwheel with ll levels is the graph 𝒪ql\mathcal{O}_{q}^{l} consisting of the cycles:

(a11,,aq1),,(a1l,,aql),(b12,,bq2),,(b1l,,bql) and (a11,b11,a21,b21,,aq1,bn1)(a_{1}^{1},\ldots,a_{q}^{1}),\cdots,(a_{1}^{l},\ldots,a_{q}^{l}),(b_{1}^{2},\ldots,b_{q}^{2}),\cdots,(b_{1}^{l},\ldots,b_{q}^{l})\text{ and }(a_{1}^{1},b_{1}^{1},a_{2}^{1},b_{2}^{1},\ldots,a_{q}^{1},b_{n}^{1})

as well as a vertex c=bil+1c=b_{i}^{l+1} called the cusp and the edges aijaij+1a_{i}^{j}a_{i}^{j+1} and bijbij+1b_{i}^{j}b_{i}^{j+1} for j=1,,lj=1,\ldots,l; see Figure 4.

Refer to caption
Figure 4. The 4-multi hyperwheel with 3 levels 𝒪43\mathcal{O}_{4}^{3}.

If q=2k4q=2k\geq 4 then 𝒪ql\mathcal{O}_{q}^{l} is a strongly involutive self-dual polyhedron. Indeed, let us check this for 𝒪ql\mathcal{O}_{q}^{l} where the strong involution α\alpha is given as follows:

α(ail)=Ail=(bi+k1bi+k11c),α(aij)=Aij=(bi+kjbi+k1jbi+k1j+1bi+kj+1), for each j=2,,l1,α(ai1)=Ai1=(bi+k2bi+k12bi+k11ai+k1bi+k1),α(bi1)=Bi1=(ai+k1bi+k1bi+k11),α(bij)=Bij=(ai+kj1ai+k1jai+k1jai+k1j1), for each j=2,,l,α(c)=C=(a1laql).\begin{array}[]{lll}\alpha(a_{i}^{l})&=A_{i}^{l}&=(b_{i+k}^{1}b_{i+k-1}^{1}c),\\ \alpha(a_{i}^{j})&=A_{i}^{j}&=(b_{i+k}^{j}b_{i+k-1}^{j}b_{i+k-1}^{j+1}b_{i+k}^{j+1}),\ \text{ for each }j=2,\ldots,l-1,\\ \alpha(a_{i}^{1})&=A_{i}^{1}&=(b_{i+k}^{2}b_{i+k-1}^{2}b_{i+k-1}^{1}a_{i+k}^{1}b_{i+k}^{1}),\\ \alpha(b_{i}^{1})&=B_{i}^{1}&=(a_{i+k}^{1}b_{i+k}^{1}b_{i+k-1}^{1}),\\ \alpha(b_{i}^{j})&=B_{i}^{j}&=(a_{i+k}^{j-1}a_{i+k-1}^{j}a_{i+k-1}^{j}a_{i+k-1}^{j-1}),\ \text{ for each }j=2,\ldots,l,\\ \alpha(c)&=C&=(a_{1}^{l}\cdots a_{q}^{l}).\\ \end{array}

Note that α\alpha, as duality isomorphism, associates the vertex set of the dual face to each point pp, while, as an isometry of the sphere, it associates the antipodal point p-p to each point pp.

We may now construct the 4-multi hyperwheel G=𝒪41G=\mathcal{O}_{4}^{1} which is just the 4-hyperwheel as illustrated in Figure 1 (right) by means of the action of its antipodal pairing. Figure 5 gives the fundamental region RR and its doodle. This doodle can generate a family of self-dual polyhedra whose pairing is [q][2,q][q]\triangleleft[2,q].

Refer to caption
Figure 5. (Left) A fundamental region (shaded) of [4][2,4][4]\triangleleft[2,4]. (Right) The doodle of the 4-hyperwheel.

The external circle with dotted line represents the north pole, the circle with double line represents the equator and the central point is the south pole. The graphs GG and GG^{*} appear in blue and red, respectively and there is a black vertex in each intersection of an edge in GG with an edge in GG^{*}. The join of these graphs (including the black vertices) is the graph of squares GG_{\square} whose all faces are quadrilaterals of the form (vafb)(vafb) for some vV(G),fV(G)v\in V(G),f\in V(G^{*}) and a,bE(G)=E(G)a,b\in E(G)=E(G^{*}).

Remark 1.

Let ϕ\phi be any element in Dual(G)\mbox{Dual}(G). We have that ϕ\phi preserves squares. Furthermore, ϕAut(G)\phi\in\mbox{Aut}(G) if and only if ϕ\phi is color preserving; ϕIso(G)\phi\in\mbox{Iso}(G) if and only if ϕ\phi is color reversing. In any case, black vertices are mapped to black vertices.

We suppose that ϕ\phi is the reflection in a spherical line ll which intersects the square (vafb)(vafb). If ϕIso(G)\phi\in\mbox{Iso}(G) then ll must pass through aa and bb and must interchange vv and ff. In this case, the reflection in the equator is a duality isomorphism since it interchanges vertices blue and red. Each automorphism can be obtained as a composition of reflections in four lines through the poles. The reflections through these lines generates the group [4][4] and each one of them works as an automorphism, so we can write Aut(G)=[4]\mbox{Aut}(G)=[4]. The group Dual(G)\mbox{Dual}(G) is [2,4][2,4] and it can be obtained by adding the reflection through the equator. In this way, Dual(G)\mbox{Dual}(G) is generated by the reflections in three planes H1,H2H_{1},H_{2} and H3H_{3}. Planes H1H_{1} and H2H_{2} intersect with angle π4\frac{\pi}{4} and we can assume they are vertical planes through the poles, while H3H_{3} is the equatorial plane and it forms angles of π2\frac{\pi}{2} with each H1,H2H_{1},H_{2}. The fundamental region is the spherical triangle with angles π2,π2,π4\frac{\pi}{2},\frac{\pi}{2},\frac{\pi}{4}. The self-dual pairing for this graph is [4][2,4][4]\triangleleft[2,4] and it is determined by this triangle with three isometries: two reflections in Aut(G)\mbox{Aut}(G) (black lines) and one in Iso(G)\mbox{Iso}(G) (double line).

Refer to caption
Figure 6. The doodle on a fundamental region RR of 𝒪ql\mathcal{O}_{q}^{l} and the resulting graphs for q=4q=4.

The above construction can be extended for each even qq and any ll: For q=4q=4 and l=1l=1 we have the doodle of G=𝒪41G=\mathcal{O}_{4}^{1} (Figure 5). Then for l=2l=2 and 33 we have the doodles and the graphs in Figure 6. Here the doodle is drawn on the region between two planes forming an angle πq\frac{\pi}{q}: the fundamental region of the group [q][q].

In Figure 7, we consider the embedding of both: the graph in blue and its dual graph in red. Here the reflection through the double line interchanges colors blue and red since it is a duality isomorphism of [q][2,q][q]\triangleleft[2,q]. If we let the group [q][q] act on the sphere, we will have a generalization of the qq-multi hyperwheel. Figure 7 illustrates the case q=4q=4 and l=3l=3.

Refer to caption
Figure 7. (Left) The doodle and fundamental region RR of 𝒪43\mathcal{O}_{4}^{3} (Right) Resulting graph embedded on 𝕊2\mathbb{S}^{2}.

3.2. The qq-multi wheel

Let qq and ll be natural numbers. The qq-multi wheel with ll levels is the graph 𝒫ql\mathcal{P}_{q}^{l} consisting of the ll cycles

(a11,,aq1),,(a1l,,aql),(a_{1}^{1},\ldots,a_{q}^{1}),\cdots,(a_{1}^{l},\ldots,a_{q}^{l}),

a central vertex c=ai0c=a_{i}^{0} called the cusp and edges aijaij+1a_{i}^{j}a_{i}^{j+1}. Figure 9 illustrates 𝒫53\mathcal{P}_{5}^{3}.

Refer to caption
Figure 8. The 5-multi wheel with 3 levels.

We notice that 𝒫q1\mathcal{P}_{q}^{1} is actually the qq-wheel (a graph consisting of a qq-cycle with a center joined to each vertex of the cycle).

If q=2k+13q=2k+1\geq 3 then 𝒫ql\mathcal{P}_{q}^{l} is a strongly involutive self-dual polyhedron. Indeed, let us check this for 𝒫ql\mathcal{P}_{q}^{l} where the strong involution α\alpha is given as follows:

α(aij)=(ai+kljai+klj+1aikljaiklj+1), for each i=1,,q and j=1,,l1,α(ail)=(ai+k1aik1c),α(c)=(a1laql).\begin{array}[]{ll}\alpha(a_{i}^{j})&=(a_{i+k}^{l-j}a_{i+k}^{l-j+1}a_{i-k}^{l-j}a_{i-k}^{l-j+1}),\text{ for each }i=1,\ldots,q\text{ and }j=1,\ldots,l-1,\\ \alpha(a_{i}^{l})&=(a_{i+k}^{1}a_{i-k}^{1}c),\\ \alpha(c)&=(a_{1}^{l}\cdots a_{q}^{l}).\end{array}

We may now construct the 3-multi wheel G=𝒫31G=\mathcal{P}_{3}^{1} by means of the action of its antipodal pairing. Figure 5 gives the fundamental region RR and its doodle. This doodle can be naturally extended to generate a family of self-dual polyhedra whose pairing is [q][2,q][q]\triangleleft[2,q].

Let us consider an isometric embedding of the graph G=𝒫3G=\mathcal{P}_{3} given in Figure 9.

Refer to caption
Figure 9. (Left) A fundamental region of [3][2+,6][3]\triangleleft[2^{+},6]. (Right) The doodle of 𝒫31\mathcal{P}_{3}^{1}.

In this embedding, the vertices a,ba,b and cc are in a circle while d,ed,e and ff are in another circle. These circles are concentric to the center oo, which we can assume to be the south pole. The red vertices correspond to the dual graph and they are also in concentric circles, in this case D,ED,E and FF are connected with the central vertex OO located in the north pole (dotted external line in the figure).

We can observe that the reflection on the equator does not work as duality but the rotation by π\pi around the circle in double lines in the shaded region is a duality isomorphism. This rotation is color reversing: ee goes to FF, the red point CC goes to bb and the central vertex oo goes to OO. If we call this rotation ρ\rho, then the group Dual(𝒫31)\mbox{Dual}(\mathcal{P}_{3}^{1}) is generated by γ1,γ2,ρ\gamma_{1},\gamma_{2},\rho and in Coxeter notation it corresponds to [2+,6][2^{+},6]. The duality pairing is [3][2+,6][3]\triangleleft[2^{+},6].

The above construction can be extended for each odd qq and any ll. We can start by restricting the graph P51P_{5}^{1}, which is the 5-wheel, to a fundamental region and then we can add more levels in order to obtain a simple, planar and 3-connected graph. See Figure 10.

Refer to caption
Figure 10. The doodle on a fundamental region RR of 𝒫ql\mathcal{P}_{q}^{l} and the resulting graphs for q=5q=5.

In Figure 11 we have a fundamental region RR for the group [q][q]. It is between two vertical planes forming an angle of πq\frac{\pi}{q}. The circle in the center of the region denotes a rotation by π\pi. This rotation sends the blue lines in the doodle to the red lines in the doodle and viceversa since it is a duality isomorphism in the pairing [q][2+,2q][q]\triangleleft[2^{+},2q]. If we let [q][q] act on 𝕊2\mathbb{S}^{2} we may obtain 𝒫ql\mathcal{P}_{q}^{l}.

Refer to caption
Figure 11. (Left) The doodle in a fundamental region RR of 𝒫53\mathcal{P}_{5}^{3}. (Right) Resulting graph embedded on 𝕊2\mathbb{S}^{2}.

4. Concluding Remarks

In [2, Theorem 6], the authors showed that if GG is a strongly involutive polyhedron then there always exists an edge eE(G)e\in E(G) such that G=G/{e}{τ(e)}G^{\prime}=G/\{e\}\setminus\{\tau(e)\} is also an strongly involutive polyhedron where τ\tau is the strong involution and G{f}G\setminus\{f\} (resp. G/{f}G/\{f\}) denotes the deletion (resp. contraction) of an edge ff in GG. It can be checked that the inverse of the delete-contraction operation corresponds to the diagonalization faces of the graph and its dual simultaneously. The latter can be settled as an add-expansion operation from GG^{\prime} to obtain GG.

The above implies that all strongly involutive polyhedra can be reduced to a wheel (with an odd number of vertices in the main cycle) by a finite sequence of delete-contraction operations (applied simultaneously each time), see [2, Corollary 7]. We end with the following

Question 1.

Let GG be a strongly involutive polyhedron and let GG^{\prime} be the strongly involutive polyhedron obtained from GG by applying a delete-contraction operation. Can the antipodal pairing of GG be determined from the antipodal pairing of GG^{\prime}? What about the corresponding orbifolds and doodles?

References

  • [1] K. Bezdek, Z. Lángi, M. Naszódi and P. Papez, Ball-polyhedra, Disc.& Comp.Geom. 38 (2007), 201–230.
  • [2] J. Bracho, L. Montejano, E. Pauli and J.L. Ramírez Alfonsín, Strongly involutive self-dual polyhedra, Ars Math. Contemp. 20 (2021), 143–149.
  • [3] H.S.M. Coxeter and W.O.J. Moser, Generators and Relations for Discrete Groups, Ergebnisse der Mathematik und ihrer Grenzgebiete, Bd. 13, Springer Verlag, 1972.
  • [4] P. Erdös, On sets of distances of nn points, Amer. Math. Monthly 53 (1946) 248–250.
  • [5] B. Grünbaum and G.C. Shepard, Is selfduality involutory?, Amer. Math. Monthly 95 (1985), 729–733.
  • [6] B. Grünbaum, A proof of Vázsonyi’s conjecture, Bull. Res. Council Israel, Sect. A, 6 (1956), 77–78.
  • [7] Y. S. Kupitz, H. Martini, and M. A. Perles, Ball polytopes and the Vázsonyi problem, Acta Math. Hungarica 126 (2010), 99–163.
  • [8] G. López-Campos, D. Oliveros and J.L. Ramírez Alfonsín, Borsuk and Vázsonyi problems through Reuleaux polyhedra, (arXiv:2308.03889).
  • [9] L. Lovàsz, Self-dual polytopes and the chromatic number of distance graphs on the sphere, Acta Sci. Math. 45 (1983), 317–323.
  • [10] H. Martini, L. Montejano and D. Oliveros, Bodies of constant width, Birkhäuser (2019).
  • [11] L. Montejano, E. Pauli, M. Raggi and E. Roldán-Pensado, The graphs behind Reuleaux polyhedra, Disc.& Comp.Geom. 64(3) (2020), 1013–1022.
  • [12] L. Montejano, J.L. Ramírez Alfonsín and I. Rasskin, Self-dual maps I: antipodality, SIAM Disc. Math., 36(3) (2022), 1551–1566.
  • [13] L. Montejano, J.L. Ramírez Alfonsín and I. Rasskin, Self-dual maps II : links and symmetry, SIAM Disc. Math., 37(1) (2023), 191–220.
  • [14] L. Montejano, J.L. Ramírez Alfonsín and I. Rasskin, Self-dual maps III: projective links, J. Knot Th. and its Ramifications, 32(10) (2023) No 2350066, 22 pp.
  • [15] B. Servatius and H. Servatius, The 24 symmetry pairings of self-dual maps on the sphere, Disc. Math. 140 (1995), 167–183.
  • [16] K.J. Swanepoel, A new proof of Vázsonyi’s conjecture, J. Comb. Th. Ser. A 115 (2008), 888–892.