betaFeedback

WorksMathematics

Existence of locally toroidal regular 4-polytopes

Abstract

We prove the existence half of Schulte–Weiss Problem 16 (2006) and McMullen–Schulte Conjecture 10C7 (2002): the universal regular 4-polytope {{4,4}(s,0),{4,4}(u,0)}\{\{4,4\}_{(s,0)},\{4,4\}_{(u,0)}\} exists for every pair of distinct odd integers s,u≥3s,u\geq3, with no coprimality assumption. Here {4,4}(s,0)\{4,4\}_{(s,0)} is the quotient of the square tessellation by sZ2s\mathbb Z^2. Finiteness and group classification remain open. The proof constructs four matrices over a finite field of characteristic two whose local subgroups are isomorphic to the prescribed toroidal groups. A commutator obstruction shows that these subgroups meet exactly in the subgroup on their two shared generators; this equality lifts to the universal presentation. The group theorem, the geometric identification of the toroidal maps, and the construction and universality of the resulting abstract polytope are formalized in Lean. This work has not yet been reviewed by experts. The proofs, the Lean formalization and the text were produced by AI models, working under the author's direction.

Introduction

We prove existence for an exceptional family in the classification of locally toroidal regular 4-polytopes. For an integer p≥3p \ge3, write {4,4}(p,0)\{4,4\}_{(p,0)} for the square tessellation modulo the translation lattice pZ2p\mathbb{Z}^{2}. Its vertices are the points of (Z/pZ)2(\mathbb{Z}/p\mathbb{Z})^{2}, with the edges and squares inherited from the unit square grid. For example, {4,4}(3,0)\{4,4\}_{(3,0)} has nine vertices and nine square faces. A regular 4-polytope is locally toroidal here when both its facets and its vertex-figures are such toroidal maps. The double-brace notation, for example

{{4,4}(3,0),{4,4}(5,0)},\{\{4,4\}_{(3,0)},\{4,4\}_{(5,0)}\},

denotes the universal regular polytope with the indicated facets and vertex-figures: it covers every regular polytope with those local types. The Schläfli type records the polygon sizes of rank-two sections; type {4,4,4}\{4,4,4\} means that all three kinds are squares.

Theorem 1.1. For every pair of distinct odd integers s,u≥3s,u \ge3, the universal regular 4-polytope

{{4,4}(s,0),{4,4}(u,0)}\{\{4,4\}_{(s,0)},\{4,4\}_{(u,0)}\}

exists. It has Schläfli type {4,4,4}\{4,4,4\}, facets isomorphic to {4,4}(s,0)\{4,4\}_{(s,0)}, and vertex-figures isomorphic to {4,4}(u,0)\{4,4\}_{(u,0)}.

McMullen and Schulte recorded the finite universal example with periods (3,5)(3,5) and its dual [5 pp. 375–376]. Here we establish existence for every pair of distinct odd periods, including pairs with common factors. We reuse Coxeter and Moser’s toroidal presentations and McMullen and Schulte’s reconstruction of regular polytopes from string C-groups (Definition 3.1). The hardest new step is proving that the facet and vertex-figure groups, with independently prescribed periods, meet exactly in their common dihedral subgroup (Proposition 5.3). Realizing the two local maps separately does not establish this equality; the new matrix commutator argument proves it without a coprimality assumption.

This is the existence half of Schulte–Weiss Problem 16 [10 Problem 16] and McMullen–Schulte’s 2002 Conjecture 10C7 [5 p. 376]. Finiteness and group classification remain open. There is no gcd restriction: pairs such as (3,9)(3,9), (9,15)(9,15) and (15,45)(15,45) are included. Figure 1 shows the local maps for (3,9)(3,9). The theorem does not assert finiteness or infiniteness of the universal polytope for any pair. In particular, the finite matrix group used in the proof is a quotient of the universal polytope’s automorphism group. A finite quotient does not decide finiteness of the universal member.

Figure 1. The facet map {4,4}(3,0)\{4,4\}_{(3,0)} and vertex-figure map {4,4}(9,0)\{4,4\}_{(9,0)} for the (3,9)(3,9) instance of Theorem 1.1, drawn as fundamental domains of the square tessellation. Identify the opposite sides in the directions of the matching arrows. Boundary vertices are repeated representatives of the same torus vertices. The shaded square, thick edge and marked vertex form a base flag. The panels depict the two local maps of the universal regular 4-polytope. The drawing is generated from their vertex, edge and square data, checked against the grid-quotient definition identified here with the coset map of TpT_p.

Peter McMullen and Egon Schulte developed the classification of these polytopes in [4] and [5 Chapter 10]. Their book singles out the case of distinct odd periods as exceptional and records the finite example with periods (3,5)(3,5) and its dual, crediting G. Havas’s coset enumeration and Leytem’s group identification [5 pp. 375–376]. Egon Schulte and Asia Ivić Weiss subsequently asked for the classification of this entire family [10]. Their problem includes the group identifications and the conjecture that only (3,5)(3,5) and (5,3)(5,3) give finite universal members. The present paper addresses the existence assertion throughout the family.

The local presentations are those of H. S. M. Coxeter and W. O. J. Moser [1 Section 8.3]. The passage between groups and regular polytopes uses the theory of string C-groups (Definition 3.1) developed in [5 Section 2E]. Barry Monson and Egon Schulte constructed related locally toroidal polytopes by reduction over finite fields [6]; their type {4,4,4}\{4,4,4\} construction has equal prime periods on the two sides. Maria Elisa Fernandes and Claudio Alexandre Piedade studied faithful permutation representations of toroidal maps [2] and the related finite universal 4-polytopes [3].

Related constructions prescribe toroidal sections in other settings. Marston D. E. Conder, Yan-Quan Feng and Dong-Dong Hou constructed chiral polytopes of type {4,4,4}\{4,4,4\} [7]; Gabriel Cunningham, Feng, Hou and Schulte constructed symmetric polytopes whose automorphism groups are 2-groups [8]. Gabe Cunningham and Isabel Hubard developed criteria for polytopality of mixes, including examples with square toroidal facets [9]. These address different symmetry or parameter conditions from the regular polytopes with two prescribed odd periods considered here.

§ 2 explains the strategy. § 3 sets out the group presentation and the local maps. § 4 gives the matrices and proves the exact local orders. § 5 proves the hard intersection equality. § 6 lifts it to the presented group and constructs the universal polytope. § 7 describes the scope of the formal proof.

Proof overview

We explain how a finite matrix representation establishes existence of the universal polytope, and identify the intersection calculation needed for the proof. The universal candidate is the group on four involutions with the [4,4,4][4,4,4] Coxeter relations and the two toroidal period relations, displayed in (4). The question is whether those relations preserve the intended local maps and give the intersection property required of a regular polytope. We prove that they do by constructing a finite matrix image that is faithful on both local groups.

Choose elements λ\lambda and μ\mu of orders ss and uu in one finite field of characteristic two. Their traces λ+λ−1\lambda+ \lambda^{-1} and μ+μ−1\mu+ \mu^{-1} determine two parameters in four explicit 4×44 \times4 matrices. These matrices satisfy the required relations. The two local translation groups have orders s2s^2 and u2u^2; each local group is the product of its translation group and the common dihedral group of order eight. Thus the local groups have exactly the toroidal orders 8s28s^2 and 8u28u^2.

The difficult step is to show that the local groups meet only in that dihedral group. Here T,T′,V,V′T,T',V,V' are the translation matrices of (8): the first two are facet translations, and the last two are vertex-figure translations. The matrix TT satisfies a quadratic polynomial with two distinct roots, so every power of TT lies in the linear span of the identity matrix and TT. A particular entry of TV−VTTV-VT is nonzero. Consequently a power of TT that commutes with VV must be scalar; oddness forces that power to be the identity. Conjugating by the common dihedral group isolates the two coordinates of an arbitrary facet translation. This proves that the two translation groups intersect trivially. To pass to the full local groups, use that a vertex-figure group modulo its translation subgroup has order eight: the image of an odd-order facet translation in that quotient must be trivial.

Finally, the matrix representation is injective on each local subgroup of the universal presentation, because the toroidal presentation has order at most 8p28p^2 and the corresponding matrix image attains that bound. We can therefore pull the intersection equality back to the universal presentation. The string C-group criterion and the coset construction give the universal regular 4-polytope.

The group presentation and the toroidal maps

We describe the group whose coset polytope will prove the theorem, and recall why its local presentations specify the intended geometric maps. Throughout this section, pp is an odd integer at least three.

An abstract polytope of rank nn is a ranked partially ordered set with a least and greatest face, of ranks −1-1 and nn, whose maximal chains (flags) contain n+2n + 2 faces. It satisfies the diamond condition: exactly two intermediate faces lie between incident faces whose ranks differ by two. It is strongly flag-connected: any two flags can be joined by successive changes of one face, keeping their common faces throughout. It is regular if its automorphism group is transitive on flags. A covering is a surjective map preserving incidence, rank and adjacency of flags. These are the conventions of [10 Section 2] and [5 Chapter 2].

For example, the standard three reflections of the square grid act by

q0(x,y)=(1−x,y),q1(x,y)=(y,x),q2(x,y)=(x,−y).(1)q_0(x,y)=(1-x,y), \qquad q_1(x,y)=(y,x), \qquad q_2(x,y)=(x,-y). \tag*{(1)}

On reducing coordinates modulo pp, they act on {4,4}(p,0)\{4,4\}_{(p,0)}. The word q0q1q2q1q_0q_1q_2q_1 is translation by (1,0)(1,0), and its q1q_1-conjugate is translation by (0,1)(0,1).

Write TpT_p for the toroidal group below; the subscript distinguishes it from the translation matrix TT used later. Its presentation is

Tp=⟨q0,q1,q2∣qi2=1, (q0q1)4=(q1q2)4=(q0q2)2=1, (q0q1q2q1)p=1⟩.(2)T_p=\langle q_0,q_1,q_2 \mid q_i^2=1,\ (q_0q_1)^4=(q_1q_2)^4=(q_0q_2)^2=1,\ (q_0q_1q_2q_1)^p=1\rangle. \tag*{(2)}

The familiar toroidal presentation (2) is the full reflection-group presentation, not just the rotation subgroup [1 Section 8.3]; see also [2 Section 2]. To see its size directly, put x=q0q1q2q1x=q_0q_1q_2q_1 and y=q1xq1y=q_1xq_1. The Coxeter relations imply that x,yx,y commute and that

q1xq1=y,q1yq1=x,q2xq2=x,q2yq2=y−1.q_1xq_1=y,\qquad q_1yq_1=x,\qquad q_2xq_2=x,\qquad q_2yq_2=y^{-1}.

Every element has a form xiyjdx^iy^jd, where 0≤i,j<p0 \leq i,j < p and d∈⟨q1,q2⟩d \in\langle q_1,q_2\rangle. The latter group has at most eight elements, so ∣Tp∣≤8p2|T_p| \leq8p^2. Conversely, the affine action (1) contains all p2p^2 translations and the eight signed coordinate permutations. For p≥3p \geq3 these are distinct, so ∣Tp∣=8p2|T_p| = 8p^2 and the action is faithful.

The stabilizers of the base vertex (0,0)(0,0), the base edge from (0,0)(0,0) to (1,0)(1,0), and the base unit square are respectively ⟨q1,q2⟩\langle q_1,q_2\rangle, ⟨q0,q2⟩\langle q_0,q_2\rangle and ⟨q0,q1⟩\langle q_0,q_1\rangle, as the normal form verifies. The induced coset-to-face bijections preserve incidence and rank, so the coset polytope of TpT_p is the square tessellation quotient {4,4}(p,0)\{4,4\}_{(p,0)}.

For a group generated by g0,…,gn−1g_0,\ldots,g_{n-1}, put GI=⟨gi:i∈I⟩G_I=\langle g_i:i\in I\rangle for each subset II of the index set. An involution is a nonidentity group element whose square is the identity; in particular, the distinguished involutions below satisfy gi2=1g_i^2=1.

Definition 3.1. A group G=⟨g0,…,gn−1⟩G=\langle g_0,\ldots,g_{n-1}\rangle with distinguished involutions is a string C-group if gigj=gjgig_i g_j=g_j g_i whenever ∣i−j∣≥2|i-j|\geq2, and

GI∩GJ=GI∩J(I,J⊆{0,…,n−1}).(3)G_I \cap G_J = G_{I\cap J} \qquad(I,J \subseteq\{0,\ldots,n-1\}). \tag*{(3)}

The generators of TpT_p have this property: the square torus is a regular polytope, and the distinguished generators of any regular polytope satisfy the string relations and intersection property [10 Section 2]. Here their affine action was identified with the full torus group above. String C-groups reconstruct regular polytopes [5 Theorem 2E11]. We will use the rank-four criterion: if the subgroups on g0,g1,g2g_0,g_1,g_2 and on g1,g2,g3g_1,g_2,g_3 are string C-groups and meet in ⟨g1,g2⟩\langle g_1,g_2\rangle, the string relations imply (3) for all subsets [5 Proposition 2E16].

Define the universal candidate by

Γ(s,u)=⟨r0,r1,r2,r3∣ri2=1 (0≤i≤3), (r0r1)4=(r1r2)4=(r2r3)4=1,\Gamma(s,u)=\langle r_0,r_1,r_2,r_3 \mid r_i^2=1\ (0\le i\le3),\ (r_0r_1)^4=(r_1r_2)^4=(r_2r_3)^4=1,
(r0r2)2=(r0r3)2=(r1r3)2=1, (r0r1r2r1)s=(r1r2r3r2)u=1⟩.(4)(r_0r_2)^2=(r_0r_3)^2=(r_1r_3)^2=1,\ (r_0r_1r_2r_1)^s=(r_1r_2r_3r_2)^u=1\rangle. \tag*{(4)}

We must prove that its local groups are Ts,TuT_s,T_u and that it is a string C-group; neither conclusion follows merely by writing down the relators.

An explicit matrix construction

We now construct the finite image of (4) and calculate its local groups. Fix distinct odd integers s,u≥3s,u\ge3 for the remainder of the proof. All matrix entries in this section belong to a finite field KK of characteristic two. The notation 11 denotes the 4×44\times4 identity matrix; the letters I,JI,J in Definition 3.1 denote index sets only.

Parameters and generators

We choose a common finite field and parameters encoding the two prescribed periods, then define the four generators. Let k=φ(su)k=\varphi(su), where φ\varphi is Euler’s totient function, and choose K=F2kK=\mathbb{F}_{2^k}. Since susu is odd, Euler’s theorem gives su∣2k−1su\mid2^k-1. The cyclic group K×K^\times therefore contains elements λ,μ\lambda,\mu of exact orders s,us,u. Set

σ=λ+λ−1,τ=μ+μ−1,a=σ2τ,b=σ+a.(5)\sigma=\lambda+\lambda^{-1},\qquad\tau=\mu+\mu^{-1},\qquad a=\frac{\sigma^2}{\tau},\qquad b=\sigma+a. \tag*{(5)}

These parameters satisfy

στab≠0,a+b=σ,(a+b)2a=τ.(6)\sigma\tau ab\ne0,\qquad a+b=\sigma,\qquad\frac{(a+b)^2}{a}=\tau. \tag*{(6)}

Indeed, either trace is zero only if the corresponding element has square one. Equality of the two traces would give (μ−λ)(μ−λ−1)=0(\mu-\lambda)(\mu-\lambda^{-1})=0, contradicting their distinct orders. Hence b=σ(τ+σ)/τb=\sigma(\tau+\sigma)/\tau is nonzero too. This choice of field uses no coprimality between ss and uu.

For c∈Kc\in K define R(c)R(c), and then the four generators, by

R(c)=(11000100c0110c01),ρ0=R(a),ρ2=R(b),R(c)= \begin{pmatrix} 1&1&0&0\\ 0&1&0&0\\ c&0&1&1\\ 0&c&0&1 \end{pmatrix}, \qquad\rho_0=R(a),\qquad\rho_2=R(b),
ρ1=(1000011000100001),ρ3=(000a−101000010a000).(7)\rho_1= \begin{pmatrix} 1&0&0&0\\ 0&1&1&0\\ 0&0&1&0\\ 0&0&0&1 \end{pmatrix}, \qquad \rho_3= \begin{pmatrix} 0&0&0&a^{-1}\\ 0&1&0&0\\ 0&0&1&0\\ a&0&0&0 \end{pmatrix}. \tag*{(7)}

They are invertible because their squares are 11. Put Λ=⟨ρ0,ρ1,ρ2,ρ3⟩≤GL⁡4(K)\Lambda=\langle\rho_0,\rho_1,\rho_2,\rho_3\rangle\leq\operatorname{GL}_4(K). Direct multiplication gives the string relations and the three adjacent fourth-power relations. Each adjacent product has order exactly four: using indices 1,2,3,41,2,3,4, the (1,1)(1,1) entries of (ρ0ρ1)2−1(\rho_0\rho_1)^2-1 and (ρ1ρ2)2−1(\rho_1\rho_2)^2-1 are a,ba,b, while the (1,2)(1,2) entry of (ρ2ρ3)2−1(\rho_2\rho_3)^2-1 is σ/a\sigma/a. All three are nonzero.

Translation orders and independent directions

We prove that each local group has two independent translation directions of the prescribed order. Define the translations and their subgroups by

T=ρ0ρ1ρ2ρ1,T′=ρ1Tρ1,V=ρ1ρ2ρ3ρ2,V′=ρ2Vρ2,A=⟨T,T′⟩,B=⟨V,V′⟩,D=⟨ρ1,ρ2⟩,F=⟨ρ0,ρ1,ρ2⟩,E=⟨ρ1,ρ2,ρ3⟩.(8)\begin{aligned} T &= \rho_0\rho_1\rho_2\rho_1, \qquad T' = \rho_1T\rho_1, \qquad V = \rho_1\rho_2\rho_3\rho_2, \qquad V' = \rho_2V\rho_2,\\ A &= \langle T,T'\rangle, \qquad B = \langle V,V'\rangle, \qquad D = \langle\rho_1,\rho_2\rangle,\\ F &= \langle\rho_0,\rho_1,\rho_2\rangle, \qquad E = \langle\rho_1,\rho_2,\rho_3\rangle. \tag*{(8)} \end{aligned}

For the fixed trace parameter τ\tau from (5), put qτ(X)=X2+τX+1q_\tau(X)=X^2+\tau X+1. Evaluating this polynomial at a matrix MM gives qτ(M)=M2+τM+1q_\tau(M)=M^2+\tau M+1. The polynomial qτq_\tau is distinct notation from the reflections q0,q1,q2q_0,q_1,q_2 of § 3. The identities in Appendix § 8, together with the explicit matrices there, give the calculations needed for the following proof.

Lemma 4.1. For the matrices (7) with parameters (5), the elements T,T′T,T' have order ss, the elements V,V′V,V' have order uu, and

A≅Cs×Cs,B≅Cu×Cu,A \cong C_s \times C_s,\qquad B \cong C_u \times C_u,

where CpC_p denotes a cyclic group of order $p.

Proof. The polynomial X2+σX+1X^2+\sigma X+1 has distinct roots λ,λ−1\lambda,\lambda^{-1}. The matrix TT is nonscalar: its (1,3)(1,3) entry is one. Thus its minimal polynomial is this quadratic. It follows that TT has order ss, and conjugation gives the same order for T′T'. Likewise VV is annihilated by the squarefree polynomial

(X−1)(X−μ)(X−μ−1).(X-1)(X-\mu)(X-\mu^{-1}).

Since V≠1V\ne1, its minimal polynomial has at least one of the roots μ,μ−1\mu,\mu^{-1} as well as, possibly, 11. Its order is therefore uu. Conjugation gives the same order for V′V'.

To prove independence of the two facet directions, observe that 1,T,T′1,T,T' are linearly independent. Indeed, the (1,3)(1,3), (2,1)(2,1) and (1,1)(1,1) entries of α1+βT+γT′=0\alpha1+\beta T+\gamma T'=0 give successively γ=β\gamma=\beta, σβ=0\sigma\beta=0 and α=0\alpha=0. Every power of TT belongs to K1+KTK1+KT, by its quadratic relation; similarly powers of T′T' belong to K1+KT′K1+KT'. If Ti(T′)j=1T^i(T')^j=1, then Ti=(T′)−jT^i=(T')^{-j} lies in the intersection of these two planes and is scalar. Evaluating a scalar power at both roots gives λi=λ−i\lambda^i=\lambda^{-i}. Since ss is odd, this forces s∣is\mid i, and then s∣js\mid j. The commuting generators consequently give A≅Cs×CsA\cong C_s\times C_s.

For the vertex directions, put W=qτ(V)W=q_\tau(V). The identities in (12) give W≠0W\ne0, VW=WVW=W and qτ(V′)W=0q_\tau(V')W=0. If Vi(V′)j=1V^i(V')^j=1, commutation implies (V′)jW=W(V')^jW=W. The polynomial qτ(X)q_\tau(X) has roots μ,μ−1\mu,\mu^{-1}. If u∤ju\nmid j, then qτ(X)q_\tau(X) and Xj−1X^j-1 are coprime. Evaluating a Bézout identity at V′V' and multiplying by WW would give W=0W=0, a contradiction: both polynomial summands annihilate WW by the two identities above. Thus u∣ju\mid j, and the order of VV then gives u∣iu\mid i. This proves B≅Cu×CuB\cong C_u\times C_u.

Proposition 4.2. For the matrices (7), the assignments qi↦ρiq_i\mapsto\rho_i and qi↦ρi+1q_i\mapsto\rho_{i+1} give isomorphisms Ts≅FT_s\cong F and Tu≅ET_u\cong E, respectively. Both preserve the distinguished generators, and ∣F∣=8s2|F|=8s^2, ∣E∣=8u2|E|=8u^2. Moreover F=A×DF=A\times D and E=B×DE=B\times D, with ∣D∣=8|D|=8.

Proof. The product ρ1ρ2\rho_1\rho_2 has order four, so DD is dihedral of order eight. The conjugation identities are

TT′VV′ρ1(⋅)ρ1T′TV−1V′ρ2(⋅)ρ2T(T′)−1V′V(9)\begin{array}{c|cccc} & T & T' & V & V'\\ \hline \rho_1(\mathord\cdot)\rho_1 & T' & T & V^{-1} & V'\\ \rho_2(\mathord\cdot)\rho_2 & T & (T')^{-1} & V' & V \tag*{(9)} \end{array}

and show that DD normalizes AA and BB. The identities

ρ0=Tρ1ρ2ρ1,ρ3=ρ2ρ1Vρ2\rho_{0}=T\rho_{1}\rho_{2}\rho_{1},\qquad\rho_{3}=\rho_{2}\rho_{1}V\rho_{2}

show that F=ADF=AD and E=BDE=BD. The odd orders ∣A∣=s2|A|=s^{2} and ∣B∣=u2|B|=u^{2} imply A∩D=B∩D=1A\cap D=B\cap D=1. This gives the asserted orders and semidirect products. Finally the local relations induce surjections Ts→FT_{s}\to F and Tu→ET_{u}\to E. The bound ∣Tp∣≤8p2|T_{p}|\leq8p^{2} from § 3, together with the image orders just proved, makes both maps isomorphisms.

The local intersection

We prove F∩E=DF\cap E=D. The local groups have the required sizes, but that does not by itself control their intersection. The following commutator obstruction is the reason that shared divisors of ss and uu cause no problem.

Lemma 5.1. For the matrices (7) with parameters (5), if n≥0n\geq0 and TnV=VTnT^{n}V=VT^{n}, then s∣ns\mid n.

Proof. The quadratic relation for TT gives scalars α,β∈K\alpha,\beta\in K with

Tn=α1+βT,λn=α+βλ,λ−n=α+βλ−1.(10)T^{n}=\alpha1+\beta T,\qquad\lambda^{n}=\alpha+\beta\lambda,\qquad\lambda^{-n}=\alpha+\beta\lambda^{-1}. \tag*{(10)}

These formulas follow simultaneously by induction, using the same quadratic relation for T,λ,λ−1T,\lambda,\lambda^{-1}. Commuting the first formula with VV gives β(TV−VT)=0\beta(TV-VT)=0. But direct multiplication gives

(TV−VT)12=σ/a≠0.(11)(TV-VT)_{12}=\sigma/a\ne0. \tag*{(11)}

Hence β=0\beta=0. The remaining formulas in (10) give λn=λ−n\lambda^{n}=\lambda^{-n}, so s∣2ns\mid2n. Since ss is odd, s∣ns\mid n, as required.

Lemma 5.2. For the translation subgroups A,BA,B in (8), with parameters (5), one has A∩B=1A\cap B=1.

Proof. Let z∈A∩Bz\in A\cap B and write z=Ti(T′)jz=T^{i}(T')^{j}, with 0≤i,j<s0\leq i,j<s. Conjugation by ρ2\rho_{2} fixes TT and inverts T′T', so

z(ρ2zρ2)=T2i,z(ρ2zρ2)−1=(T′)2j.z(\rho_{2}z\rho_{2})=T^{2i},\qquad z(\rho_{2}z\rho_{2})^{-1}=(T')^{2j}.

Both elements belong to BB, since DD normalizes BB. Conjugating the second one by ρ1\rho_{1} shows that T2j∈BT^{2j}\in B too. The group BB is abelian and contains VV. Thus T2iT^{2i} and T2jT^{2j} commute with VV. Lemma 5.1 gives s∣2is\mid2i and s∣2js\mid2j. Oddness of ss gives s∣i,js\mid i,j, whence z=1z=1.

Proposition 5.3. For the local matrix groups F,E,DF,E,D in (8), with parameters (5), one has

F∩E=D.F\cap E=D.

Proof. The inclusion D⊆F∩ED\subseteq F\cap E is immediate. For the reverse inclusion, let z∈F∩Ez\in F\cap E. Write z=tdz=td with t∈At\in A and d∈Dd\in D, using Proposition 4.2. Then t=zd−1∈Et=zd^{-1}\in E. Its order is odd, because ∣A∣=s2|A|=s^{2} is odd. But BB is normal in EE, and ∣E/B∣=8|E/B|=8. The image of tt in E/BE/B has order dividing both eight and an odd number; it is therefore trivial. Thus t∈Bt\in B. Lemma 5.2 now gives t=1t=1, so z=d∈Dz=d\in D.

This argument never compares divisors of ss with divisors of uu. The only cancellations are by two in the facet period, and by a power of two in a quotient of the vertex-figure group. Coprimality of the two periods is unnecessary.

The local isomorphisms of Proposition 4.2 and the rank-four criterion also show that Λ\Lambda is a string C-group. Because its matrices are over a finite field, its coset polytope is a finite regular 4-polytope with facets {4,4}(s,0)\{4,4\}_{(s,0)} and vertex-figures {4,4}(u,0)\{4,4\}_{(u,0)}. This gives a finite realization for every pair in the theorem. The next section concerns the universal member, whose group maps onto Λ\Lambda and need not be identified with it.

The universal group and polytope

We transfer the intersection calculation from the finite matrix image to the full presentation (4), then recover its faces and universal covering property.

Theorem 6.1. Let s,u≥3s,u \ge3 be distinct odd integers, and let Γ(s,u)\Gamma(s,u) be the group with exactly the presentation (4). Its distinguished generators form a string C-group, and all three adjacent products have order exactly four. The homomorphisms from (2) give distinguished-generator-preserving isomorphisms

Ts≅⟨r0,r1,r2⟩,Tu≅⟨r1,r2,r3⟩.T_s \cong\langle r_0,r_1,r_2\rangle,\qquad T_u \cong\langle r_1,r_2,r_3\rangle.

These subgroups have orders 8s28s^2 and 8u28u^2, respectively.

Proof. The matrix relations and Lemma 4.1 give an epimorphism π:Γ(s,u)→Λ\pi:\Gamma(s,u)\to\Lambda with π(ri)=ρi\pi(r_i)=\rho_i. Put F^=⟨r0,r1,r2⟩\widehat{F}=\langle r_0,r_1,r_2\rangle, E^=⟨r1,r2,r3⟩\widehat{E}=\langle r_1,r_2,r_3\rangle and D^=⟨r1,r2⟩\widehat{D}=\langle r_1,r_2\rangle. There are surjections Ts→F^→FT_s\to\widehat{F}\to F and Tu→E^→ET_u\to\widehat{E}\to E. The composed maps Ts→FT_s\to F and Tu→ET_u\to E are exactly the generator-preserving isomorphisms of Proposition 4.2. Hence each map in the two chains is an isomorphism, and π\pi is injective on both F^\widehat{F} and E^\widehat{E}. In particular, both are string C-groups with the displayed generators.

If z∈F^∩E^z\in\widehat{F}\cap\widehat{E}, then π(z)∈F∩E=D=π(D^)\pi(z)\in F\cap E=D=\pi(\widehat{D}). Choose d∈D^d\in\widehat{D} with π(z)=π(d)\pi(z)=\pi(d). Both z,dz,d lie in F^\widehat{F}, where π\pi is injective. Thus z=dz=d, proving F^∩E^=D^\widehat{F}\cap\widehat{E}=\widehat{D}. The rank-four criterion recalled in § 3 now proves the intersection property for every pair of index subsets. The string relations already hold in the presentation. Finally the adjacent products have fourth power one in Γ(s,u)\Gamma(s,u) and images of order four in Λ\Lambda, so their orders are exactly four. Each generator is consequently a nonidentity involution. □

Proof of Theorem 1.1. Let Γ=Γ(s,u)\Gamma=\Gamma(s,u). For i=0,1,2,3i=0,1,2,3 put Hi=⟨rj:j≠i⟩H_i=\langle r_j:j\ne i\rangle. The proper faces of rank ii are the left cosets gHigH_i. Two proper faces of different ranks are incident if their cosets meet; orient this incidence by rank, and adjoin a least face of rank −1-1 and a greatest face of rank 4. This is the coset construction P(Γ)P(\Gamma) of [5 Section 2E]. Theorem 6.1 supplies the string C-group hypotheses, so the construction is a regular 4-polytope.

By the reconstruction theorem [5 Theorem 2E11], left multiplication identifies Γ\Gamma with the full automorphism group, acting simply transitively on flags.

The facets are the coset polytopes of F^\widehat{F}, and the vertex-figures those of E^\widehat{E}. Their generator-preserving identifications with Ts,TuT_s,T_u show that they are {4,4}(s,0),{4,4}(u,0)\{4,4\}_{(s,0)},\{4,4\}_{(u,0)}, with the geometric meanings established in § 3. Exact adjacent orders give Schläfli type {4,4,4}\{4,4,4\}.

It remains to check universality. Let QQ be any regular 4-polytope with these facets and vertex-figures, and choose a base flag. Its distinguished involutions satisfy the Coxeter relations of (4). The two local types imply the two additional period relations. Thus rir_i may be sent to the corresponding distinguished involution of QQ, giving a surjection Γ→Aut⁡(Q)\Gamma\to\operatorname{Aut}(Q), where Aut⁡(Q)\operatorname{Aut}(Q) is the group of order automorphisms of QQ, meaning bijections preserving its partial order in both directions. On cosets, this sends gHigH_i to the image under gg of the base face of rank ii of QQ. It is well-defined because every generator of HiH_i fixes that face. The map is surjective and preserves rank, incidence and flag adjacency, so it is a covering. Therefore P(Γ)P(\Gamma) is the universal member {{4,4}(s,0),{4,4}(u,0)}\{\{4,4\}_{(s,0)},\{4,4\}_{(u,0)}\}, as asserted. This is the universal construction of [5 Theorem 4A2] in this particular class.

Formalization

The published Lean proof (linked below the title) establishes the group and geometric polytope conclusions from their definitions. It includes the finite-field choice, the local orders, the full intersection property, and the transfer to the universal presentation. The square torus is defined independently of a group as the vertex sets of grid faces modulo pZ2p\mathbb{Z}^2, ordered by inclusion; the proof identifies it with the local coset polytope. Regularity, the full automorphism group, and the universal covering property are also included.

The formal group theorem asserts the exact presentation, the full intersection property, and the two generator-preserving local isomorphisms. The formal geometric theorem constructs the regular polytope with the specified square-torus sections and its covering onto every regular polytope with those local sections.1 Trust rests on the Lean kernel and the stated foundational definitions.

Matrix calculations

We give the explicit translation matrices and collect the multiplication identities used in § 4. The notation is that of (5) and (8); in particular, σ=a+b\sigma=a+b and τ=(a+b)2/a\tau=(a+b)^2/a. Expanding the products in (8) gives

T=(1+b011b101σσ1+σ0abσb1+a),V=(0σ/a0a−1σ1+τ1σ/aστ1σ/aaσ00).T=\begin{pmatrix}1+b&0&1&1\\b&1&0&1\\\sigma&\sigma&1+\sigma&0\\ab&\sigma&b&1+a\end{pmatrix},\qquad V=\begin{pmatrix}0&\sigma/a&0&a^{-1}\\\sigma&1+\tau&1&\sigma/a\\\sigma&\tau&1&\sigma/a\\a&\sigma&0&0\end{pmatrix}.

The quadratic expression used to separate the vertex translations is

qτ(V)=V2+τV+1=(τ0σ/a0σ00σ/a000000στ).q_{\tau}(V)=V^2+\tau V+1=\begin{pmatrix}\tau&0&\sigma/a&0\\\sigma&0&0&\sigma/a\\0&0&0&0\\0&0&\sigma&\tau\end{pmatrix}.

With T′=ρ1Tρ1T'=\rho_1T\rho_1 and V′=ρ2Vρ2V'=\rho_2V\rho_2, multiplication gives the entries (T′)13=1(T')_{13}=1, (T′)21=a(T')_{21}=a and (T′)11=1+b(T')_{11}=1+b used to check independence in Lemma 4.1, and the identities

T2=σT+1,(T′)2=σT′+1,TT′=T′T,VV′=V′V,(V+1)qτ(V)=0,(V′+1)qτ(V′)=0,qτ(V′)qτ(V)=0,(qτ(V))11=τ,V≠1.(12)\begin{aligned} T^2&=\sigma T+1, & (T')^2&=\sigma T'+1, & TT'&=T'T,\\ VV'&=V'V, & (V+1)q_{\tau}(V)&=0, & (V'+1)q_{\tau}(V')&=0,\\ q_{\tau}(V')q_{\tau}(V)&=0, & (q_{\tau}(V))_{11}&=\tau, & V&\ne1. \tag*{(12)} \end{aligned}

These are identities in characteristic two with the parameter relations (6). The displayed matrices exhibit T13=1T_{13}=1, V11=0V_{11}=0 and (qτ(V))11=τ≠0(q_{\tau}(V))_{11}=\tau\ne0, which are the nonvanishing entries used in Lemma 4.1.

References

  1. [1]H. S. M. Coxeter and W. O. J. Moser, Generators and Relations for Discrete Groups, 4th ed., Ergebnisse der Mathematik und ihrer Grenzgebiete 14, Springer, 1980.
  2. [2]M. E. Fernandes and C. A. Piedade, Faithful permutation representations of toroidal regular maps, J. Algebraic Combin. 52 (2020), 317–337. doi:10.1007/s10801-019-00904-8; arXiv:1808.09705. Correction: J. Algebraic Combin. 54 (2021), 733–738, doi:10.1007/s10801-020-00985-w.DOI
  3. [3]M. E. Fernandes and C. A. Piedade, The degrees of regular polytopes of type {4, 4, 4}, SIAM J. Discrete Math. 36 (2022), 1143–1155. doi:10.1137/20M1375012.DOI
  4. [4]P. McMullen and E. Schulte, Regular polytopes of type {4, 4, 3} and {4, 4, 4}, Combinatorica 12 (1992), 203–220. doi:10.1007/BF01204723.DOI
  5. [5]P. McMullen and E. Schulte, Abstract Regular Polytopes, Encyclopedia of Mathematics and its Applications 92, Cambridge University Press, 2002. doi:10.1017/CBO9780511546686.DOI
  6. [6]B. Monson and E. Schulte, Reflection groups and polytopes over finite fields, II, Adv. Appl. Math. 38 (2007), 327–356. doi:10.1016/j.aam.2005.12.001; arXiv:math/0601502.DOI
  7. [7]M. D. E. Conder, Y.-Q. Feng and D.-D. Hou, Two infinite families of chiral polytopes of type {4, 4, 4} with solvable automorphism groups, doi:10.1016/j.jalgebra.2020.11.002; arXiv:1912.03398.DOI
  8. [8]G. Cunningham, Y.-Q. Feng, D.-D. Hou and E. Schulte, Symmetric polytopes whose automorphism groups are 2-groups, doi:10.1016/j.jalgebra.2026.03.017; arXiv:2512.15511.DOI
  9. [9]G. Cunningham and I. Hubard, Polytopality criteria for the mix of polytopes and manifolds, arXiv:2506.19334.arxiv.org/abs/2506.19334
  10. [10]E. Schulte and A. Ivić Weiss, Problems on polytopes, their groups, and realizations, Period. Math. Hungar. 53 (2006), 231–255. doi:10.1007/s10998-006-0035-y; arXiv:math/0608397.DOI

Paper details

vlnc:W_ydgkcwwa
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

AI models

View statement
Looking for related papers…
Contents