On a {K_4,K_{2,2,2}}-ultrahomogeneous graph

Italo J. Dejter

Introduction

Let HH be a connected regular graph and let m,n∈ZZm,n\in{{\rm Z}\kern-2.79999pt{\rm Z}} with 1<m<n1<m<n. An {H}nm\{H\}_{n}^{m}-graph is a connected graph that: (a) is representable as an edge-disjoint union of nn induced copies of HH; (b) has exactly mm copies of HH incident to each vertex, with no two such copies sharing more than one vertex; and (c) has exactly nn copies of HH as induced subgraphs isomorphic to HH.

We remark that an {H}nm\{H\}_{n}^{m}-graph GG is {H}\{H\}-ultrahomogeneous (as in ) if every isomorphism between two copies of HH in GG extends to an automorphism of GG. Graph ultrahomogeneity is a concept that can be traced back to .

Notice that a connected graph GG is mm-regular if and only if it is a {K2}∣E(G)∣m\{K_{2}\}_{|E(G)|}^{m}-graph. In this case, GG is arc-transitive if and only if GG is {K2}\{K_{2}\}-ultrahomogeneous. Thus, {H}\{H\}-ultrahomogeneity is a notion of graph symmetry stronger than arc-transitivity.

If GG is an {Hi}nimi\{H_{i}\}_{n_{i}}^{m_{i}}-graph, where i=1,2i=1,2, and H1≠H2H_{1}\neq H_{2}, then GG is said to be an {H1}n1m1{H2}n2m2\{H_{1}\}_{n_{1}}^{m_{1}}\{H_{2}\}_{n_{2}}^{m_{2}}-graph. If, in addition, GG is {Hi}\{H_{i}\}-ultrahomogeneous, for both i=1,2i=1,2, then GG is {H1,H2}\{H_{1},H_{2}\}-ultrahomogeneous, again as in . If each edge of GG is in exactly one copy of HiH_{i}, for both i=1,2i=1,2, then GG is said to be fastened. If min(m1,m2)=m1=2(m_{1},m_{2})=m_{1}=2 and H1H_{1} is a complete graph, then GG is said to be line-graphical. For example, the line graph of the dd-cube, where 3≤d∈ZZ3\leq d\in{{\rm Z}\kern-2.79999pt{\rm Z}}, is a line-graphical fastened {Kd,K2,2}\{K_{d},K_{2,2}\}-ultrahomogeneous {Kd}2d2{K2,2}d(d−1)2d−3d−1\{K_{d}\}_{2^{d}}^{2}\{K_{2,2}\}_{d(d-1)2^{d-3}}^{d-1}-graph. The first case here, known as the cuboctahedron, is a fastened {K3,K2,2,C6}\{K_{3},K_{2,2},C_{6}\}-ultrahomogeneous {K3}82{K2,2}62{C6}42\{K_{3}\}_{8}^{2}\{K_{2,2}\}_{6}^{2}\{C_{6}\}_{4}^{2}-graph, where C6C_{6} is 6-cycle.

In Sections 3-5, a 12-regular fastened {K4,K2,2,2}\{K_{4},K_{2,2,2}\}-ultrahomogeneous {K4}424{K2,2,2}213\{K_{4}\}_{42}^{4}\{K_{2,2,2}\}_{21}^{3}-graph GG of order 42 and diameter 3 is presented. The role that dd-cliques KdK_{d} and squares K2,2K_{2,2} play in the line graph of the dd-cube is performed in GG by tetrahedra K4K_{4} and octahedra K2,2,2K_{2,2,2}, but in this case with min(m1,m2)=(m_{1},m_{2})=min(4,3)>2(4,3)>2, so GG is non-line-graphical.

The graph GG has automorphism-group order ∣A(G)∣=1008=4∣E(G)∣.|{\mathcal{A}}(G)|=1008=4|E(G)|. In Section 5, the 252 edges of GG can be seen as the left cosets of a subgroup Γ⊂A(G)\Gamma\subset{\mathcal{A}}(G) of order 4, and its vertices as the left cosets of a subgroup of A(G){\mathcal{A}}(G) of order 24.

These two equivalence classes of subgraphs of GG, i.e. tetrahedra and octahedra, allow in Section 6 to define several combinatorial configurations () related to GG, 3 of which are self-dual, with their Levi graphs as: (1) a 4-regular 2-arc-transitive graph () on 84 vertices and 1008 automorphisms, with diameter == girth == 6, reflecting a natural duality property of GG; (2) an 8-regular arc-transitive graph on 42 vertices and 2016 automorphisms, with diameter == 3 and girth == 4; and (3) a 6-regular semisymmetric graph () on 336 vertices and 1008 automorphisms, with diameter == girth == 6 and just two slightly differing distance distributions. The Menger graph and dual Menger graph associated to this Levi graph have common degree 24 and diameter == girth == 3, with 1008 and 2016 automorphisms, respectively.

Section 7 distinguishes the kk-holes (or chordless kk-cycles) of GG with the least k>4k>4, namely k=6k=6, and studies their participation in some toroidal subgraphs of GG that together with the octahedra of GG can be filled up to form a closed piecewise linear 3-manifold.

After some considerations on the Fano plane, we pass to define GG and study its properties.

Ordered Fano pencils

Given a point pp of F{\mathcal{F}}, the collection of lines through pp is a pencil of F{\mathcal{F}}. A linearly ordered presentation of these lines is an ordered pencil through pp. An ordered pencil vv through pp is denoted v=(p,qara,qbrb,qcrc)v=(p,q_{a}r_{a},q_{b}r_{b},q_{c}r_{c}), orderly composed, in reality, by the lines pqara,pqbrb,pqcrcpq_{a}r_{a},pq_{b}r_{b},pq_{c}r_{c}. Note that there are 3!=63!=6 ordered pencils through any point pp of F{\mathcal{F}}.

Ordered pencils constitute the vertex set of our claimed graph GG, with any two vertices v=(p,qara,qbrb,qcrc)v=(p,q_{a}r_{a},q_{b}r_{b},q_{c}r_{c}) and v′=(p′,qa′ra′,qb′rb′,qc′rc′)v^{\prime}=(p^{\prime},q^{\prime}_{a}r^{\prime}_{a},q^{\prime}_{b}r^{\prime}_{b},q^{\prime}_{c}r^{\prime}_{c}) adjacent whenever the following two conditions hold: (1) p≠p′p\neq p^{\prime}; (2) ∣piri∩pi′ri′∣=1|p_{i}r_{i}\cap p^{\prime}_{i}r^{\prime}_{i}|=1, for i=a,b,ci=a,b,c. The 3 points of intersection resulting from item (2) form a Fano line, which we consider as an ordered Fano line by taking into account the subindex order a<b<ca<b<c, and as such, set it as the strong color of the edge vv′vv^{\prime}. This provides GG with an edge-coloring.

An alternate definition of GG can be given via Φ−1\Phi^{-1}, in which the vertices of GG can be seen as the ordered Fano lines xaxbxcx_{a}x_{b}x_{c}, with any two such vertices adjacent if their associated Fano lines share the entry in F{\mathcal{F}} of exactly one of its 3 positions, either aa or bb or cc. We keep throughout, however, the ordered-pencil presentation of GG, but the first self-dual configuration of Subsection 6.1 and accompanying example show that the suggested dual presentation of GG is valid as well.

Notice that the vertices of GG with initial entry p=1p=1 appear in lexicographic order as:

which may be simplified in notation by using super-indices aa through ff to denote the shown order, that is: 1a,1b,1c,1d,1e,1f,1^{a},1^{b},1^{c},1^{d},1^{e},1^{f}, respectively. A similar lexicographic presentation may be given to the vertices of GG having p=2,…,7p=2,\ldots,7. This treatment covers the 42 vertices of GG. As an example of the adjacency of GG, the neighbors of 1a=(1,23,45,67)1^{a}=(1,23,45,67) in GG are:

or in the continuation of the simplified notation above: 2a,2b,3a,3b,4c,4e,5c,5e,6d,6f,7d,7f2^{a},2^{b},3^{a},3^{b},4^{c},4^{e},5^{c},5^{e},6^{d},6^{f},7^{d},7^{f}. The strong colors of the resulting edges are: 167167, 154154, 176176, 154154, 356356, 246246, 347347, 451451, 321321, 231231, 321321, respectively.

Given vertices v=(p,qara,qbrb,qcrc)v=(p,q_{a}r_{a},q_{b}r_{b},q_{c}r_{c}) and w=(p′,qa′ra′,qb′rb′,qc′rc′)w=(p^{\prime},q^{\prime}_{a}r^{\prime}_{a},q^{\prime}_{b}r^{\prime}_{b},q^{\prime}_{c}r^{\prime}_{c}) adjacent in GG, there exists a well-defined j∈{a,b,c}j\in\{a,b,c\} such that (1) p∈qj′rj′p\in q^{\prime}_{j}r^{\prime}_{j}; (2) p′∈qjrjp^{\prime}\in q_{j}r_{j}; (3) the lines pqjrjpq_{j}r_{j} and p′qj′rj′p^{\prime}q^{\prime}_{j}r^{\prime}_{j} intersect at either qjq_{j} or rjr_{j}, which coincides with either qj′q^{\prime}_{j} or rj′r^{\prime}_{j}. Say that these lines pqjrjpq_{j}r_{j} and p′qj′rj′p^{\prime}q^{\prime}_{j}r^{\prime}_{j} intersect at qjq_{j}. Then qjq_{j} (including the subindex jj) is taken as the weak color for the edge vwvw. This provides GG with another edge-coloring, with symbols qjq_{j}, where q∈Fq\in{\mathcal{F}} and j∈{a,b,c}j\in\{a,b,c\}. For example, the weak colors qjq_{j} corresponding to the 12 edges incident to 1a1^{a}, as cited above, are: 3a,3_{a}, 3a,3_{a}, 2a,2_{a}, 2a,2_{a}, 5b,5_{b}, 5b,5_{b}, 4b,4_{b}, 4b,4_{b}, 7c,7_{c}, 7c,7_{c}, 6c,6_{c}, 6c6_{c}, respectively.

The 12 neighbors of 1a1^{a} displayed above induce a subgraph NG(1a)N_{G}(1^{a}) of GG, called the open neighborhood of 1a1^{a} in GG, which is isomorphic to the graph Λ\Lambda of the hemi-rhombicuboctahedron (obtained from the rhombicuboctahedron by identification of antipodal vertices and edges). This is a 4-regular vertex-transitive graph on 12 vertices embedded in the projective plane with 13 faces realized by 4 disjoint triangles and 9 additional 4-holes. The 4-holes are of two types: (1) 6 have two opposite sides adjacent each to a triangle; (2) the other 3 have only their vertices in common with the 4 triangles. We also have the graph homomorphism f:Λ→K4f:\Lambda\rightarrow K_{4} of Figure 1, where f(ji)=if(j_{i})=i for i∈{0,1,2,3}i\in\{0,1,2,3\}, j∈{a,b,c}j\in\{a,b,c\} and Λ\Lambda is depicted in two different ways inside (dotted) fundamental polygons of the real projective plane. Moreover, we may identify Λ\Lambda with NG(1a)N_{G}(1^{a}) via a graph isomorphism g:Λ→NG(1a)g:\Lambda\rightarrow N_{G}(1^{a}) given by:

Moreover, the graph homomorphism ff induces, at the level of automorphism groups of graphs, a group isomorphism f∗:A(Λ)→A(K4)=S4f^{*}:{\mathcal{A}}(\Lambda)\rightarrow{\mathcal{A}}(K_{4})=S_{4}. In fact, f∗f^{*} is given by sending the following generators of A(Λ){\mathcal{A}}(\Lambda) into corresponding generators of S4S_{4} (that can be better visualized from the leftmost Λ\Lambda to the rightmost K4K_{4} depicted in Figure 1):

Thus, A(NG(1a))=A(Λ)=S4{\mathcal{A}}(N_{G}(1^{a}))={\mathcal{A}}(\Lambda)=S_{4} has 24 elements, which is consistent with the size of a vertex stabilizer of GG. Furthermore, since GG has 42 vertices that behave exactly in the same geometric way as ordered pencils in F\mathcal{F}, we conclude that ∣A(G)∣=42×24=1008|{\mathcal{A}}(G)|=42\times 24=1008.

Notice that ff maps bijectively the 3 4-cycles and 4 triangles of K4K_{4} respectively onto the 3 4-holes of Λ\Lambda of type (2) above and the 4 triangles of Λ\Lambda. Notice also that these 7 holes form a cycle-decomposition of Λ\Lambda. Inside the closed neighborhood NG[w]N_{G}[w] of each vertex ww of GG (induced in GG by ww and the open neighborhood NG(w)N_{G}(w)), we obtain 3 copies of K2,2,2K_{2,2,2} and 4 copies of K4K_{4}, which are induced by ww together respectively with the mentioned 3 4-holes and 4 triangles. Observe that these 7 induced subgraphs of GG have intersection formed solely by ww. The rest of this section is dedicated to the study of these polyhedral subgraphs.

First, notice that the inverse image f−1f^{-1} of each edge of K4K_{4} is one of the 6 4-holes of Λ\Lambda of type (1) above. This yields another cycle-decomposition of Λ\Lambda, which in turn makes explicit the remaining 4-holes of GG, apart from the 4-holes contained in the copies of K2,2,2K_{2,2,2} of GG. However, these new 4-holes are not contained in any copy of K2,2,2K_{2,2,2} in GG.

Each vertex of GG belongs to 3 induced copies of K2,2,2K_{2,2,2} in GG. For example, the sets of weak colors qjq_{j} of the edges of such copies for the vertex 1a1^{a}, that contain the 4-holes g(c0,c2,c1,c3)g(c_{0},c_{2},c_{1},c_{3}), g(a0,a1,a2,a3)g(a_{0},a_{1},a_{2},a_{3}) and g(b0,b1,b3,b2)g(b_{0},b_{1},b_{3},b_{2}) arising in Subsection 3.1, are respectively: {1a,2a,3a},{1b,4b,5b},{1c,6c,7c}\{1_{a},2_{a},3_{a}\},\{1_{b},4_{b},5_{b}\},\{1_{c},6_{c},7_{c}\}.

Each qjq_{j} colors the edges of a specific 4-hole in its corresponding copy of K2,2,2K_{2,2,2}. The 3 weak colors appearing in each copy of K2,2,2K_{2,2,2} correspond bijectively with its 3 4-holes, the edges of each 4-hole bearing a common weak color of its own.

A similar situation holds for any other vertex of GG. There is a copy of K2,2,2K_{2,2,2} in GG whose set of weak colors of edges is {xj,yj,zj}\{x_{j},y_{j},z_{j}\}, for each line xyzxyz of F{\mathcal{F}} and index j∈{a,b,c}j\in\{a,b,c\}. We denote this copy of K2,2,2K_{2,2,2} by [xyz]j[xyz]_{j}. As a result, there is a total of 21=7×321=7\times 3 copies of K2,2,2K_{2,2,2} in GG. In fact, triangles with weak colors qjq_{j} sharing a common jj (but qq varying) are only present in the said copies of K2,2,2K_{2,2,2} in GG. Each 4-hole in a copy of K2,2,2K_{2,2,2} in GG have: (1) edges sharing a common weak color qjq_{j} and (2) opposite vertices representing ordered pencils through a common point of F{\mathcal{F}}, which yields a total of two such points per 4-hole.

For example, the strong colors of the triangles [xyz]j[xyz]_{j} composing the copies of K2,2,2K_{2,2,2} incident to 1a1_{a} conform triples of strong colors having:

for a_{a}, aa-entries covering line 123, and another fixed entry equal to each one of 4,5,6,7:

for b_{b}, bb-entries covering line 145, and another fixed entry equal to each one of 2,3,6,7:

for c_{c}, cc-entries covering line 167, and another fixed entry equal to each one of 2,3,4,5:

In fact, these triangles are respectively:

The sets of strong colors for the respective composing 4-holes are:

There is one copy of K4K_{4} in GG for each ordered Fano line xyzxyz. Such a copy, denoted ⟨xyz⟩\langle xyz\rangle, is formed by 3 pairs of equally weakly-colored opposite edges, with weak colors xax_{a}, yby_{b} and zcz_{c}. For each p∈F∖{x,y,z}p\in{\mathcal{F}}\setminus\{x,y,z\}, there is exactly one vertex (p,qara,qbrb,qcrc)(p,q_{a}r_{a},q_{b}r_{b},q_{c}r_{c}) of ⟨xyz⟩\langle xyz\rangle, with x∈qarax\in q_{a}r_{a}, y∈qbrby\in q_{b}r_{b}, z∈qcrcz\in q_{c}r_{c}. The strong colors of the edges of ⟨xyz⟩\langle xyz\rangle are precisely xyzxyz. For example, the triangles g(c1,a2,b2)g(c_{1},a_{2},b_{2}), g(c3,a0,b1)g(c_{3},a_{0},b_{1}), g(c2,a1,b0)g(c_{2},a_{1},b_{0}) and g(c0,a3,b2)g(c_{0},a_{3},b_{2}) from Subsection 3.1 are contained respectively in ⟨347⟩\langle 347\rangle, ⟨246⟩\langle 246\rangle, ⟨257⟩\langle 257\rangle and ⟨356⟩\langle 356\rangle. Since there are 42 such copies of K4K_{4} in GG, we arrive at the following result.

The graph GG is a 12-regular {K4}424{K2,2,2}213\{K_{4}\}_{42}^{4}\{K_{2,2,2}\}_{21}^{3}-graph of order 42 and diameter 3. Each vertex of GG is incident to exactly 3 copies of K2,2,2K_{2,2,2} and 4 copies of K4K_{4}.

Proof: Let G′G^{\prime} be the graph defined by the same rules that define GG on the unordered Fano lines. Then it is not hard to prove that G′G^{\prime} is isomorphic to the graph 2K72K_{7}, the complete graph on 7 vertices with each edge doubled. The graph GG is then a 6-fold covering graph over G′G^{\prime}. Also, the lexicographically smallest path realizing the diameter of GG is the 3-path (1a(1^{a}, 2a2^{a}, 4a4^{a}, 1d).1^{d}). The statement follows.

Each point pp of F{\mathcal{F}} determines a Pasch configuration PC(p)PC(p), formed by the 4 lines of F{\mathcal{F}} that do not contain pp. This PC(p)PC(p) may be denoted also pc(qara,qbrb,qcrc),pc(q_{a}r_{a},q_{b}r_{b},q_{c}r_{c}), where pqara,pqbrb,pqcrcpq_{a}r_{a},pq_{b}r_{b},pq_{c}r_{c} are the lines of F{\mathcal{F}} containing pp. None of the lines of PC(p)PC(p) contains either qaraq_{a}r_{a} or qbrbq_{b}r_{b} or qcrcq_{c}r_{c}. The 7 possible Pasch configurations here are:

Figure 2 shows the disposition of the induced copies of K2,2,2K_{2,2,2} and K4K_{4} incident to the vertex 7f7^{f} in GG, represented by 3 octahedra and 4 tetrahedra, respectively, with vertices and edges accompanied by their respective simplified notations and weak colors. The 4 tetrahedra in the figure are also shown as separate entities, for better distinction, while the 3 octahedra are integrated in the central drawing as an upper-left, an upper-right and a lower-central octahedron, radiated from the central vertex, 7f7^{f}. This 7 polyhedra can be blown up to 3-space without more intersections than those of the vertices and edges shown in the figure. Starting from the right upper corner in the figure and shown counterclockwise, the 3 octahedra have respective composing 4-holes, each sub-indexed with its common weak color, as follows:

The triangles in each octahedron here differ from those in the copies of K4K_{4} in GG in the way their edges are weakly colored. For example, the copies of K4K_{4} in Figure 2, namely those denoted ⟨321⟩\langle 321\rangle, ⟨426⟩\langle 426\rangle, ⟨356⟩\langle 356\rangle, ⟨451⟩\langle 451\rangle, have their corresponding sets of constituent triangles with the clockwise sequences of simplified notations and weak colors of respective alternate incident vertices and edges, as follows:

This reflects the fact that the vertex 7f=(7,34,25,16)=(p,qa,ra,qbrb,qcrc)7^{f}=(7,34,25,16)=(p,q_{a},r_{a},q_{b}r_{b},q_{c}r_{c}) is associated with the Pasch configuration pc(34,25,16)=pc(qara,qbrb,qcrc)pc(34,25,16)=pc(q_{a}r_{a},q_{b}r_{b},q_{c}r_{c}) given with its triples ordered according to the presence of the different symbols ≠7\neq 7 at the 3 pair positions a,b,ca,b,c, which is shown in the ordered Fano lines 321=qaqbqc321=q_{a}q_{b}q_{c}, 426=raqbrc426=r_{a}q_{b}r_{c}, 356=qarbrc356=q_{a}r_{b}r_{c}, 451=rarbqc451=r_{a}r_{b}q_{c}, or in their respectively associated tetrahedra: ⟨321⟩,⟨426⟩,⟨356⟩,⟨451⟩\langle 321\rangle,\langle 426\rangle,\langle 356\rangle,\langle 451\rangle. These ordered lines form the ordered Pasch configuration pc‾(7f)={321,426,356,451}\overline{pc}(7^{f})=\{321,426,356,451\}. Similarly, an ordered Pasch configuration is associated to the set of copies of K4K_{4} incident to any other vertex of GG. Moreover, the following two results are readily checked.

Any vertex v=(p,qara,qbrb,qcrc)v=(p,q_{a}r_{a},q_{b}r_{b},q_{c}r_{c}) of GG can be expressed in such a way that ⟨qaqbqc⟩,\langle q_{a}q_{b}q_{c}\rangle, ⟨qarbrc⟩,\langle q_{a}r_{b}r_{c}\rangle, ⟨raqbrc⟩,\langle r_{a}q_{b}r_{c}\rangle, ⟨rarbqc⟩\langle r_{a}r_{b}q_{c}\rangle are its 4 incident copies of K4K_{4}, reflecting their notation and that of its 3 incident octahedra.

Proof: The ordered Pasch configuration pc‾(v)\overline{pc}(v) associated to vv determines the ordered lines qaqbqc,q_{a}q_{b}q_{c}, qarbrc,q_{a}r_{b}r_{c}, raqbrc,r_{a}q_{b}r_{c}, rarbqcr_{a}r_{b}q_{c} associated to the copies of K4K_{4}, while the 3 remaining triples of F{\mathcal{F}} provide the data for the octahedra incident to vv: [pqara]a,[pqbrb]b,[pqcrc]c[pq_{a}r_{a}]_{a},[pq_{b}r_{b}]_{b},[pq_{c}r_{c}]_{c}.

For any edge ee of GG, there exists exactly one copy of K2,2,2K_{2,2,2} and one of K4K_{4} in GG that intersect at ee. Moreover, ee is the only edge at which those copies intersect. Thus, GG is fastened.

Proof: Let e=vv′e=vv^{\prime} have weak color qjq_{j}, where v=(p,qara,qbrb,qcrc)v=(p,q_{a}r_{a},q_{b}r_{b},q_{c}r_{c}) and v′=(p′,qa′ra′,qb′rb′,v^{\prime}=(p^{\prime},q^{\prime}_{a}r^{\prime}_{a},q^{\prime}_{b}r^{\prime}_{b}, qc′rc′)q^{\prime}_{c}r^{\prime}_{c}). Then, the octahedron [pp′p′′]j[pp^{\prime}p^{\prime\prime}]_{j} and the tetrahedron ⟨xyz⟩\langle xyz\rangle are the copies of K2,2,2K_{2,2,2} and K4K_{4} in the statement, where: (a) pp′p′′pp^{\prime}p^{\prime\prime} is the Fano line containing pp and p′p^{\prime}, (b) j∈{a,b,c}j\in\{a,b,c\} is such that pp′′=qj′rj′pp^{\prime\prime}=q^{\prime}_{j}r^{\prime}_{j} and pp′=qjrjpp^{\prime}=q_{j}r_{j} and (c) xyzxyz, one of the 4 ordered lines cited in Theorem 4.2 with respect to vv, is the strong color of ee.

For example, the edge 7f5a7^{f}5^{a} has weak color 2b2_{b} and strong color 426. This is the only edge shared by the octahedron b_{b} and the tetrahedron ⟨426⟩\langle 426\rangle.

Symmetric properties of G𝐺G

Each automorphism τ∈A(G)\tau\in{\mathcal{A}}(G), is the composition of a permutation ϕτ\phi^{\tau} of F{\mathcal{F}} with a permutation ψτ\psi^{\tau} of {a,b,c}\{a,b,c\}. A set of 16 generators τi\tau_{i} of A(G){\mathcal{A}}(G), (i=1…16i=1\ldots 16), is given by τi=ψi∘ϕi=ϕi∘ψi\tau_{i}=\psi_{i}\circ\phi_{i}=\phi_{i}\circ\psi_{i}, where we denote ϕi=ϕτi\phi_{i}=\phi^{\tau_{i}}, ψi=ψτi\psi_{i}=\psi^{\tau_{i}}, and

with ψi\psi_{i} and ϕj\phi_{j} taken as the identity maps of F{\mathcal{F}} and {a,b,c}\{a,b,c\}, respectively, for 1≤i≤141\leq i\leq 14 and j=15,16j=15,16.

The subgroup Γ⊂A(G)\Gamma\subset{\mathcal{A}}(G) that sends the lexicographically smallest arc (1a,2a)(1^{a},2^{a}) onto itself, either directly or inversely oriented, includes exchanging, or not, its incident triangles (2a,3a,1a)(2^{a},3^{a},1^{a}) and (2a,3b,1a)(2^{a},3^{b},1^{a}) in a_{a}, or (1a,2a,6f)(1^{a},2^{a},6^{f}) and (1a,2a,5e)(1^{a},2^{a},5^{e}) in ⟨347⟩\langle 347\rangle. Thus, Γ\Gamma contains 4 elements and has generating set {τ6∘τ16,τ5}\{\tau_{6}\circ\tau_{16},\tau_{5}\}. Moreover, Γ\Gamma is a subgroup of A(a){\mathcal{A}}(_{a}), which has generating set {τ1,τ2,τ5,τ6,τ16}\{\tau_{1},\tau_{2},\tau_{5},\tau_{6},\tau_{16}\}. Furthermore, {τ1,τ2,τ5,τ6,τ15,τ16}\{\tau_{1},\tau_{2},\tau_{5},\tau_{6},\tau_{15},\tau_{16}\} is a generating set for {\mathcal{A}}(\cup_{j=a}^{c}_{j}). The remaining automorphisms τi\tau_{i} map {\mathcal{A}}(\cup_{j=a}^{c}_{j}) onto its nontrivial cosets in A(G){\mathcal{A}}(G) by left multiplication. The subgroup of A(G){\mathcal{A}}(G) that fixes 1a1^{a} has order 24 and generating set {ϕ1,  ϕ2,  ϕ4∘ϕ16,  ϕ6∘ϕ16,  ϕ8∘ϕ15,  ϕ10∘ϕ15,  ϕ12∘ϕ15∘ϕ16,  ϕ14∘ϕ15∘ϕ16}.\{\phi_{1},\,\,\phi_{2},\,\,\phi_{4}\circ\phi_{16},\,\,\phi_{6}\circ\phi_{16},\,\,\phi_{8}\circ\phi_{15},\,\,\phi_{10}\circ\phi_{15},\,\,\phi_{12}\circ\phi_{15}\circ\phi_{16},\,\,\phi_{14}\circ\phi_{15}\circ\phi_{16}\}.

GG is a fastened {K4,K2,2,2}\{K_{4},K_{2,2,2}\}-ultrahomogeneous {K4}424{K2,2,2}213\{K_{4}\}_{42}^{4}\{K_{2,2,2}\}_{21}^{3}-graph which is non-line-graphical, with

The edges of GG can be seen as the left cosets of a subgroup Γ⊂A(G)\Gamma\subset{\mathcal{A}}(G) of order 4, and its vertices as the left cosets of a subgroup of A(G){\mathcal{A}}(G) of order 24.

Proof: Recall from Subsection 3.1 that ∣A(G)∣=1008|{\mathcal{A}}(G)|=1008.

Notice that A(⟨347⟩)=S4{\mathcal{A}}(\langle 347\rangle)=S_{4} is formed by 24 automorphisms. Since ∣A(G)∣∣A(⟨347⟩)∣=100824=42\frac{|{\mathcal{A}}(G)|}{|{\mathcal{A}}(\langle 347\rangle)|}=\frac{1008}{24}=42, then ⟨347⟩\langle 347\rangle is sent by an automorphism of GG onto any other copy of K4K_{4} in GG, from which it is not difficult to see that GG is K4K_{4}-ultrahomogeneous.

Similarly, A(a){\mathcal{A}}(_{a}) is formed by 48 automorphisms. Since ∣A(G)∣∣A(a)∣=100848=21\frac{|{\mathcal{A}}(G)|}{|{\mathcal{A}}(_{a})|}=\frac{1008}{48}=21, then a_{a} is sent by an automorphism of GG onto any other copy of K2,2,2K_{2,2,2} in GG, from which it is not difficult to see that GG is K2,2,2K_{2,2,2}-ultrahomogeneous.

On the other hand, ∣A(K2,2,2)∣=48|{\mathcal{A}}(K_{2,2,2})|=48 and ∣E(K2,2,2)∣=12|E(K_{2,2,2})|=12 agree with the fact that ∣Γ∣=∣A(K2,2,2)∣∣E(K2,2,2)∣=4|\Gamma|=\frac{|{\mathcal{A}}(K_{2,2,2})|}{|E(K_{2,2,2})|}=4. Since GG is the edge-disjoint union of 21 copies of K2,2,2K_{2,2,2}, it contains a total of 21∣E(K2,2,2)∣=21×12=25221|E(K_{2,2,2})|=21\times 12=252 edges. Now, ∣A(G)∣=1008=21×48=21∣A(K2,2,2)∣|{\mathcal{A}}(G)|=1008=21\times 48=21|{\mathcal{A}}(K_{2,2,2})|. This is 4 times the number 252 of edges of GG. These edges correspond to the left cosets of Γ\Gamma in A(G){\mathcal{A}}(G) and its vertices to the left cosets of the stabilizer of 1a=(1,23,45,67)1^{a}=(1,23,45,67) in A(G){\mathcal{A}}(G), whose order is 24.

Configurations associated with G𝐺G

The symmetrical disposition of objects in GG gives place to several combinatorial point-line configurations and to their associated Levi, Menger, and dual Menger graphs.

We present the points and lines of 3 self-dual configurations obtained from GG, and their incidence relations:

the 42 vertices and 42 tetrahedra of GG, and incidence given by inclusion of a vertex in a tetrahedron; this is a self-dual (424)(42_{4})-configuration with 2-arc-transitive Levi graph of diameter == girth =6=6, automorphism-group order 2016, stabilizer order 24, distance distribution vector (1,4,12,24,27,14,2)(1,4,12,24,27,14,2) and isomorphic arc-transitive Menger graphs of diameter == girth =3=3, degree 12 and automorphism-group order 1008;

the 168 tetrahedral triangles and 168 octahedral triangles in GG and their sharing of an edge; this is a self-dual (1686)(168_{6})-configuration with semisymmetric Levi graph of diameter == girth =6=6, automorphism-group order 1008, common stabilizer order 6, distance distribution vectors (1,6,24,60,111,102,32)(1,6,24,60,111,102,32) and (1,6,24,60,108,102,35)(1,6,24,60,108,102,35), (just differing at distances 4 and 6 by 3 vertices) and vertex-transitive Menger graphs of common degree 24, diameter == girth =3=3 and automorphism-group orders 1008 and 2016, respectively.

the 168 tetrahedral triangles and 168 octahedral triangles in GG and their sharing of an edge; this is a self-dual (1686)(168_{6})-configuration with semisymmetric Levi graph of diameter == girth =6=6, automorphism-group order 1008, common stabilizer order 6, distance distribution vectors (1,6,24,60,111,102,32)(1,6,24,60,111,102,32) and (1,6,24,60,108,102,35)(1,6,24,60,108,102,35), (just differing at distances 4 and 6 by 3 vertices) and vertex-transitive Menger graphs of common degree 24, diameter == girth =3=3 and automorphism-group orders 1008 and 2016, respectively.

Another interesting configuration associated to GG is formed by the 42 tetrahedra and 21 octahedra of GG, and their sharing of an edge; this is a flag-transitive (426,2112(42_{6},21_{12})-configuration.

Example. Let L\mathcal{L} be the Levi graph of the (424)(42_{4})-configuration in item 1 above. Then

are the lexicographically smallest paths realizing the diameter of L\mathcal{L} and departing from each one of the two vertex parts of L\mathcal{L}. The second lexicographically smallest paths are

We reach this way to the only two vertices realizing the diameter of L\mathcal{L} starting from (1, 23, 45, 67), namely (1,67,23,45)(1,67,23,45) and (1,45,67,23)(1,45,67,23); respectively: starting from ⟨123⟩\langle 123\rangle, namely ⟨312⟩\langle 312\rangle and ⟨231⟩\langle 231\rangle. Those two pairs of paths reflect the correspondence between both parts of L\mathcal{L} induced by the map Φ\Phi in Section 2.

On 6-holes and other subgraphs of G𝐺G

GG contains 84 6-holes obtainable from the octahedral triangles of GG based on these depictions of F\mathcal{F}. This is exemplified on the left side of Figure 4, where the upper-left triangle in Figure 3 appears as the bottom triangle, sharing its edge of weak color 2a2_{a} with the central 6-hole. The 6-cycle of weak colors associated to this 6-hole is (2a3b1a2b3a1b)(2_{a}3_{b}1_{a}2_{b}3_{a}1_{b}).

Since the central color of each octahedral triangle having an edge in common with this 6-hole is 5c5_{c}, and the line 123 has its points composing the weak colors of its edges, we denote this 6-hole by 123c5123_{c}^{5}, as indicated in the figure. Similar denominations are given to the 6-holes neighboring this 123c5123_{c}^{5} in the figure. We observe that the 6-holes neighboring 123c5123_{c}^{5} on left and right coincide with 246c5246_{c}^{5}; on upper-left and lower-right with 347c5347_{c}^{5}; on lower-left and upper-right with 167c5167_{c}^{5}. These 4 6-holes form, together with the 8 octahedral triangles that appear in their generation (i.e. the 6 in the figure plus (1b,2a,3a)(1^{b},2^{a},3^{a}) and (1e,2c,3c)(1^{e},2^{c},3^{c})), a non-induced toroidal subgraph of GG that we may denote c_{c}. In the same way, a non-induced subgraph [w]d[w]_{d} is obtained, for each w∈Fw\in\mathcal{F} and d∈{a,b,c}d\in\{a,b,c\}. The subgraph [[w]]d[[w]]_{d} induced by [w]d[w]_{d} in GG is formed by its union with the copies of K4K_{4} in GG of the form ⟨xaxbxc⟩\langle x_{a}x_{b}x_{c}\rangle, with xd=wx_{d}=w, a total of 6 copies of K4K_{4} sharing each a 4-cycle with [w]d[w]_{d}. Each [x1xbxc][x_{1}x_{b}x_{c}] here is the union of such a 4-cycle plus two additional edges of weak color wdw_{d}. We get 21 subgraphs [[w]]d[[w]]_{d} of GG.

The 6-holes of the form xyzdwxyz_{d}^{w}, where xyzxyz is a fixed line of F\mathcal{F}, dd varies in {a,b,c}\{a,b,c\} and ww in F\mathcal{F}, are 12 in number and conform a subgraph [xyz][xyz] of GG isomorphic to the star Cayley graph ST4ST_{4}, that can be defined as the graph with vertex set S4S_{4} and each vertex (a0,a1,a2,a3)∈S4(a_{0},a_{1},a_{2},a_{3})\in S_{4} adjacent solely to (a1,a0,a2,a3)(a_{1},a_{0},a_{2},a_{3}), (a2,a1,a0,a3)(a_{2},a_{1},a_{0},a_{3}) and (a3,a1,a2,a0)(a_{3},a_{1},a_{2},a_{0}); see . For example, the right side of Figure 4 depicts a (dotted) fundamental polygon of the torus whose convex hull contains a representation of the subgraph $ofofG.Thissubgraphs. This subgraphs[xyz]arenotinducedinare not induced inG.However,thegraph. However, the graph[[xyz]]inducedininduced inGbyeachby each[xyz]istheedge−disjointunionofis the edge-disjoint union of[xyz]withtheedge−disjointunionofsixcopiesofwith the edge-disjoint union of six copies ofK_{4}ininG,namely:, namely:\langle xyz\rangle,\langle xzy\rangle,\langle yxz\rangle,\langle yzx\rangle,\langle zxy\rangle,\langle zyx\rangle.Figure4has4verticespaintedblack,whichspanacopyof. Figure 4 has 4 vertices painted black, which span a copy ofK_{4}ininGbutnotinbut not in.Weget7subgraphs. We get 7 subgraphs[xyz]ofofG$.

Now, two new flag-transitive configurations associated to GG (apart from those cited in Section 6) are given by: (a) the 42 tetrahedra and 21 subgraphs [wd][w_{d}] in GG, and inclusion of a tetrahedron in a copy of T‾\overline{T}; this is a (423,216)(42_{3},21_{6})-configuration; (b) the 21 subgraphs [wd][w_{d}] and 7 copies of ST4ST_{4} in GG, and their sharing of a 6-hole; this is a (214,712)(21_{4},7_{12})-configuration.

When considered immersed in 3-space, the 21 octahedra and 28 toroidal subgraphs of GG presented above have faces that appear in canceling pairs, allowing the visualization of a closed piecewise-linear 3-manifold. We ask: which are the properties of this manifold?

Open problems

It remains to see whether GG is a Cayley graph or not. On the other hand, the definition of GG may be extended by means of projective planes, like the Fano plane, but over larger fields than GF(2)GF(2), starting with GF(3)GF(3). Moreover, the two conditions of the definition of GG in Section 3 may be taken to 3 conditions, replacing F{\mathcal{F}} by a binary projective space P(r−1,2)P(r-1,2) of dimension r−1r-1, and the Fano lines by subspaces of dimension σ<r−1\sigma<r-1, where 2<r∈ZZ2<r\in{{\rm Z}\kern-2.79999pt{\rm Z}} and σ∈(0,r−1)∩ZZ\sigma\in(0,r-1)\cap{{\rm Z}\kern-2.79999pt{\rm Z}}, and requiring, as a third condition, that the points of intersection of a modified condition (b) form a projective hyperplane in P(r−1,2)P(r-1,2), (which was not required for GG, since it was a ready conclusion). The resulting graph, that appears in place of GG, may not be even connected, but the study of the component containing the lexicographically smallest vertex could still be interesting. Another step would be taking the study over other fields, starting with the ternary one.

Acknowledgement: The author is grateful to Josep Rifà and Jaume Pujol, for their friendship and support, and to the referee, for his helpful observations, among which was suggesting the hemi-rhombicuboctahedron as the open neighborhood of each vertex of GG.

References