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 exists for every pair of distinct odd integers , with no coprimality assumption. Here is the quotient of the square tessellation by . 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 , write for the square tessellation modulo the translation lattice . Its vertices are the points of , with the edges and squares inherited from the unit square grid. For example, 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
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 means that all three kinds are squares.
Theorem 1.1. For every pair of distinct odd integers , the universal regular 4-polytope
exists. It has Schläfli type , facets isomorphic to , and vertex-figures isomorphic to .
McMullen and Schulte recorded the finite universal example with periods 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 , and are included. Figure 1 shows the local maps for . 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 and vertex-figure map for the 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 .
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 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 and 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 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 [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 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 and of orders and in one finite field of characteristic two. Their traces and determine two parameters in four explicit matrices. These matrices satisfy the required relations. The two local translation groups have orders and ; 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 and .
The difficult step is to show that the local groups meet only in that dihedral group. Here are the translation matrices of (8): the first two are facet translations, and the last two are vertex-figure translations. The matrix satisfies a quadratic polynomial with two distinct roots, so every power of lies in the linear span of the identity matrix and . A particular entry of is nonzero. Consequently a power of that commutes with 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 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, is an odd integer at least three.
An abstract polytope of rank is a ranked partially ordered set with a least and greatest face, of ranks and , whose maximal chains (flags) contain 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
On reducing coordinates modulo , they act on . The word is translation by , and its -conjugate is translation by .
Write for the toroidal group below; the subscript distinguishes it from the translation matrix used later. Its presentation is
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 and . The Coxeter relations imply that commute and that
Every element has a form , where and . The latter group has at most eight elements, so . Conversely, the affine action (1) contains all translations and the eight signed coordinate permutations. For these are distinct, so and the action is faithful.
The stabilizers of the base vertex , the base edge from to , and the base unit square are respectively , and , as the normal form verifies. The induced coset-to-face bijections preserve incidence and rank, so the coset polytope of is the square tessellation quotient .
For a group generated by , put for each subset of the index set. An involution is a nonidentity group element whose square is the identity; in particular, the distinguished involutions below satisfy .
Definition 3.1. A group with distinguished involutions is a string C-group if whenever , and
The generators of 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 and on are string C-groups and meet in , the string relations imply (3) for all subsets [5 Proposition 2E16].
Define the universal candidate by
We must prove that its local groups are 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 for the remainder of the proof. All matrix entries in this section belong to a finite field of characteristic two. The notation denotes the identity matrix; the letters 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 , where is Euler’s totient function, and choose . Since is odd, Euler’s theorem gives . The cyclic group therefore contains elements of exact orders . Set
These parameters satisfy
Indeed, either trace is zero only if the corresponding element has square one. Equality of the two traces would give , contradicting their distinct orders. Hence is nonzero too. This choice of field uses no coprimality between and .
For define , and then the four generators, by
They are invertible because their squares are . Put . Direct multiplication gives the string relations and the three adjacent fourth-power relations. Each adjacent product has order exactly four: using indices , the entries of and are , while the entry of is . 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
For the fixed trace parameter from (5), put . Evaluating this polynomial at a matrix gives . The polynomial is distinct notation from the reflections 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 have order , the elements have order , and
where denotes a cyclic group of order $p.
Proof. The polynomial has distinct roots . The matrix is nonscalar: its entry is one. Thus its minimal polynomial is this quadratic. It follows that has order , and conjugation gives the same order for . Likewise is annihilated by the squarefree polynomial
Since , its minimal polynomial has at least one of the roots as well as, possibly, . Its order is therefore . Conjugation gives the same order for .
To prove independence of the two facet directions, observe that are linearly independent. Indeed, the , and entries of give successively , and . Every power of belongs to , by its quadratic relation; similarly powers of belong to . If , then lies in the intersection of these two planes and is scalar. Evaluating a scalar power at both roots gives . Since is odd, this forces , and then . The commuting generators consequently give .
For the vertex directions, put . The identities in (12) give , and . If , commutation implies . The polynomial has roots . If , then and are coprime. Evaluating a Bézout identity at and multiplying by would give , a contradiction: both polynomial summands annihilate by the two identities above. Thus , and the order of then gives . This proves .
Proposition 4.2. For the matrices (7), the assignments and give isomorphisms and , respectively. Both preserve the distinguished generators, and , . Moreover and , with .
Proof. The product has order four, so is dihedral of order eight. The conjugation identities are
and show that normalizes and . The identities
show that and . The odd orders and imply . This gives the asserted orders and semidirect products. Finally the local relations induce surjections and . The bound from § 3, together with the image orders just proved, makes both maps isomorphisms.
The local intersection
We prove . 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 and cause no problem.
Lemma 5.1. For the matrices (7) with parameters (5), if and , then .
Proof. The quadratic relation for gives scalars with
These formulas follow simultaneously by induction, using the same quadratic relation for . Commuting the first formula with gives . But direct multiplication gives
Hence . The remaining formulas in (10) give , so . Since is odd, , as required.
Lemma 5.2. For the translation subgroups in (8), with parameters (5), one has .
Proof. Let and write , with . Conjugation by fixes and inverts , so
Both elements belong to , since normalizes . Conjugating the second one by shows that too. The group is abelian and contains . Thus and commute with . Lemma 5.1 gives and . Oddness of gives , whence .
Proposition 5.3. For the local matrix groups in (8), with parameters (5), one has
Proof. The inclusion is immediate. For the reverse inclusion, let . Write with and , using Proposition 4.2. Then . Its order is odd, because is odd. But is normal in , and . The image of in has order dividing both eight and an odd number; it is therefore trivial. Thus . Lemma 5.2 now gives , so .
This argument never compares divisors of with divisors of . 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 is a string C-group. Because its matrices are over a finite field, its coset polytope is a finite regular 4-polytope with facets and vertex-figures . This gives a finite realization for every pair in the theorem. The next section concerns the universal member, whose group maps onto 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 be distinct odd integers, and let 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
These subgroups have orders and , respectively.
Proof. The matrix relations and Lemma 4.1 give an epimorphism with . Put , and . There are surjections and . The composed maps and are exactly the generator-preserving isomorphisms of Proposition 4.2. Hence each map in the two chains is an isomorphism, and is injective on both and . In particular, both are string C-groups with the displayed generators.
If , then . Choose with . Both lie in , where is injective. Thus , proving . 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 and images of order four in , so their orders are exactly four. Each generator is consequently a nonidentity involution. □
Proof of Theorem 1.1. Let . For put . The proper faces of rank are the left cosets . Two proper faces of different ranks are incident if their cosets meet; orient this incidence by rank, and adjoin a least face of rank and a greatest face of rank 4. This is the coset construction 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 with the full automorphism group, acting simply transitively on flags.
The facets are the coset polytopes of , and the vertex-figures those of . Their generator-preserving identifications with show that they are , with the geometric meanings established in § 3. Exact adjacent orders give Schläfli type .
It remains to check universality. Let 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 may be sent to the corresponding distinguished involution of , giving a surjection , where is the group of order automorphisms of , meaning bijections preserving its partial order in both directions. On cosets, this sends to the image under of the base face of rank of . It is well-defined because every generator of fixes that face. The map is surjective and preserves rank, incidence and flag adjacency, so it is a covering. Therefore is the universal member , 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 , 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, and . Expanding the products in (8) gives
The quadratic expression used to separate the vertex translations is
With and , multiplication gives the entries , and used to check independence in Lemma 4.1, and the identities
These are identities in characteristic two with the parameter relations (6). The displayed matrices exhibit , and , which are the nonvanishing entries used in Lemma 4.1.
References
- [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]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]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]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]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]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]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]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]G. Cunningham and I. Hubard, Polytopality criteria for the mix of polytopes and manifolds, arXiv:2506.19334.arxiv.org/abs/2506.19334
- [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
