total graph coherent configuration new graphs from moore
play

Total graph coherent configuration: New graphs from Moore graphs - PowerPoint PPT Presentation

Total graph coherent configuration: New graphs from Moore graphs Leif K. Jrgensen Aalborg University Denmark Joint work with M. Klin and M. Ziv-Av, BGU, Israel Our goal is to construct nice graphs from complements of Moore graphs, i.e.,


  1. Total graph coherent configuration: New graphs from Moore graphs Leif K. Jørgensen Aalborg University Denmark Joint work with M. Klin and M. Ziv-Av, BGU, Israel

  2. Our goal is to construct nice graphs from complements of Moore graphs, i.e., regular graphs with • large automorphism group, or • relation of a homogeneous coherent configuration with low rank.

  3. A Moore graph (of diameter 2) is a graph M with diameter 2 and girth 5. It is known that M is regular of some valency k and that (Hoffman and Singleton 1960) either • k = 2 and M is C 5 , (Automorphism group: Dihedral group of order 10) • k = 3 and M is the Petersen graph, P (Automorphism group: S 5 of order 120) • k = 7 and M is the Hoffman-Singleton graph or (Automorphism group: P Σ U ( 3, 5 2 ) of order 252 000) • k = 57 where existence is unknown.

  4. For each of the three known Moore graphs, the automorphism group G has rank 3, i.e., G acts transitively of vertices and sta- bilizer of a vertex v has 3 orbits: { v } , the set of neighbours of v , the set of non-neighbours of v . For a Moore graph of valency 57 the aut. group G satisfies: • (Aschbacher 1971) G is not rank 3 • (G. Higman, see Cameron 1999) G is not vertex-transitive • (Makhnev and Paduchikh 2001) caj and ˇ • (Maˇ Sir´ aˇ n 2010) G has order at most 375.

  5. For a graph Γ , the total graph of Γ , denoted by T ( Γ ) , has vertex set V ( Γ ) ∪ E ( Γ ) and two vertices in T ( Γ ) are adjacent if they are adjacent or incident in Γ . We assume that Γ is regular of valency k . Then T ( Γ ) is regular of valency 2 k .

  6. v 1 t v 2 t v k t v 1 t v 2 t v k t ❏ ✡ ❏ ✡ ❏ ✡ ❏ ✡ ❏ ✡ ❏ ✡ t ❏ ✡ ❏ ✡ v ❏ ✡ ❏ ✡ ❏ t ✡ ✡ ❏ ❏ ✡ ✡ ❏ v ✡ ❏ ✡ ❏ t ✡ t ❏ t ✡ ❏ vv 1 vv 2 vv k N Γ ( v ) and N T ( Γ ) ( v ) for v ∈ V ( Γ ) v 1 v k − 1 vv 1 vv k − 1 t t t t ✥✥✥✥✥✥✥✥✥✥ ✟✟✟✟ t ❆ � ❇ ✡ v ❇ ✡ ❆ � ◗◗◗◗◗◗ ❆ t � ❇ ✡ v ❆ � ❇ ✡ ❇ t ✡ ❇ ✡ uv ✑✑✑✑✑✑ t ✂ ❏ u ✂ ❏ ✁ ❅ t ✂ ❏ ✁ ❅ u ✂ ❏ ❵❵❵❵❵❵❵❵❵❵ ❍❍❍❍ t ✁ ❅ t t ✂ ❏ t ✁ ❅ ✂ ❏ u 1 u k − 1 uu 1 uu k − 1 N Γ ( uv ) and N T ( Γ ) ( uv ) for uv ∈ E ( Γ )

  7. If Γ is connected and regular of valency k ≥ 3 and Γ is not a complete graph then the automorphism groups satisfy Aut ( T ( Γ )) ≃ Aut ( Γ ) . For a complete graph we have T ( K n ) ≃ L ( K n + 1 ) and so Aut ( T ( K n )) ≃ Aut ( L ( K n + 1 )) ≃ Aut ( K n + 1 ) ≃ S n + 1 �≃ Aut ( K n ) ≃ S n .

  8. ✈ ✈ v 5 v 1 v 1 v 2 ✈ ✈ v 1 v 1 ✈ ✈ ✈ ✈ v 2 v 2 v 5 v 5 ✈ ✈ v 4 v 5 v 2 v 3 ✈ v 4 v 3 ✈ ✈ v 4 v 3 ✈ ✈ v 3 v 4 C 5 and T ( C 5 ) This picture is easily generalized to the total graph of C n .

  9. We want to consider the total graph of the complement of a Moore graph. First, let M be the Moore graph of valency k = 2 and M be its complement, i.e., M = C 5 . On the next page we will construct a new graph with the same set of vertices as T ( C 5 ) , shown on the previous page, but with another construction of the edge set.

  10. ✈ ✈ v 5 v 1 v 1 v 2 v 1 ✈ ✈ ✈ ✈ v 1 v 5 v 2 ✈ ✈ ✈ ✈ v 2 v 5 ✈ ✈ v 4 v 5 v 2 v 3 ✈ v 4 v 3 ✈ ✈ ✈ v 4 v 3 M = C 5 Γ ✈ v 3 v 4 Adjacent pairs in new graph Γ isomorphic to Petersen graph: { v i , v j } where v i and v j are adjacent in M . R 2 : R 15 : { v i v j , v ℓ v h } where v i = v ℓ and v j and v h are adjacent in M . . R 5 : ( v ℓ , v i v j ) and . R 10 : ( v i v j , v ℓ ) where v ℓ is adjacent to v i and v j in M .

  11. The automorphism group of T ( C 5 ) is vertex transitive and has order 20 . Consider the subgroup G of Aut ( T ( C 5 )) mapping (blue) vertices of C 5 to (blue) vertices of C 5 . The above are 4 orbits under the action of G on ordered pairs of vertices. However, we will now focus on the combinatorial description of types of adjacency.

  12. A coherent configuration ( X , { R 1 , . . . , R r } consists of a finite set X (of points or vertices) and a partition of X × X in relations R i , i ∈ I = { 1, . . . , r } satisfying that 1. There is a subset I ′ ⊂ I such that � R i = ∆ : = { ( x , x ) | x ∈ X } . i ∈ I ′ 2. For each i ∈ I there is i ′ ∈ I such that R T = R i ′ , where i R T i = { ( y , x ) | ( x , y ) ∈ R i } . 3. For all h , i , j ∈ I there is a constant p h ij such that for every x , y ∈ X where ( x , y ) ∈ R h the number of points z ∈ X with ( x , z ) ∈ R i and ( z , y ) ∈ R j is exactly p h ij .

  13. Example: From a Moore graph M of diameter 2 (or any strongly regular graph), we get a coherent configuration with rank r = 3 : Let X = V ( M ) and let R 1 = ∆ , R 2 = E ( M ) , R 3 = E ( M ) . This coherent configuration is • homogeneous , i.e., ∆ is one the relations, • symmetric , i.e., i ′ = i for all i ∈ I . A homogeneous (and symmetric) coherent configuration is also called an association scheme.

  14. Example: For any finite set X there is a coherent configuration with rank r = n 2 , n = | X | , where each R i consists of one element from X × X . For a graph Γ with V ( Γ ) = X the partition of X × X in R 1 = ∆ , R 2 = E ( Γ ) , R 3 = E ( Γ ) is in general not a coherent configuration. But there is a refinement { R ′ 1 , . . . , R ′ r } of { R 1 , R 2 , R 3 } such that ( X , { R ′ 1 , . . . , R ′ r } ) is a coherent configuration. The coarsest such partition (i.e., with least possible r ) is called the coherent configuration generated by Γ . A polynomial-time algorithm for computing this is Weisfeiler- Leman stabilization (1968).

  15. ③ ③ ③ ③ ❅ � ❅ � ❅ � ❅ � ❅ � ❅ � ❅ � ❅ � ❅ � ❅ � ❅ ③ � ③ ❅ � � ❅ � ❅ � ❅ � ❅ � ❅ � ❅ � ❅ � ❅ � ❅ � ❅ ③ ③ ③ ③ � ❅ � ❅ Graph Γ and the coherent configuration generated by Γ . The coherent configuration generated by Γ has rank 6 and 2 fibers. ∆ = R 1 ∪ R 2 , R 3 T = R 4 , R 5 , R 6 .

  16. Our goal is to find new homogeneous coherent configurations from a given coherent configuration by merging some of the relations. Such merging can be computed by COCO (Faradˇ zev and Klin 1992).

  17. For a graph Γ the coherent configuration generated by T ( Γ ) is called the total graph coherent configuration of Γ . For a given graph Γ we can perform the following computations 1. Compute T ( Γ ) 2. Compute coherent configuration genrated by T ( Γ ) , using Weisfeiler- Leman stabilization 3. Compute homogeneous coherent configurations by merging relations using COCO.

  18. We are interested in the case when Γ is a strongly regular graph and in particular when Γ is the complement of a Moore graph (of diameter 2). Theorem Ziv-Av, 2009 If Γ = L ( K n ) (triangular graph) then interesting mergings exist only for n = 5 and n = 7 . . For n = 7 : Zara graph. . L ( K 5 ) is the complement of the Petersen graph, see later. If Γ = L ( K n , n ) (square lattice graph) then no homogenous merging exist. Proof is based on clever use of computers. But when possible we want to describe the results without use of computer.

  19. Let M denote a Moore graph of diameter 2 and valency k . We will consider T ( M ) . We describe 33 relations that are naturally obtained from T ( M ) . For a particular Moore graph these 33 relations need not all appear. For k = 2 , i.e., M = C 5 , 12 of the relations appear. For k = 3 , i.e., M = Petersen graph, 24 of the relations appear. For k = 7 , i.e., M = Hoffman-Singleton graph, 31 relations appear. For k = 57 it may be that some further refinement is needed in order to get a coherent configuration.

  20. A vertex in T ( M ) will be denoted by v , where v ∈ V ( M ) or vw where v and w are non-adjacent vertices in M . First consider relations of the form ( v , x ) , for v , x ∈ V ( M ) . Description of x so that ( v , x ) ∈ R i Relation valency of v in R i Restrictions x = v 1 R 1 x ∼ v R 2 k k ( k − 1 ) R 3 x ≁ v

  21. Next consider relations of the form ( v , xy ) , where v ∈ V ( M ) and xy is a non-edge in M . Let z denote the unique common neighbour of x and y in M Description of xy Relation so that ( v , xy ) ∈ R i valency of v in R i Restrictions R 4 = R T v ∈ { x , y } k ( k − 1 ) 9 R 5 = R T 1 x ∼ v and y ∼ v 2 k ( k − 1 ) 10 R 6 = R T k ( k − 1 ) 2 x ∼ v ≁ y or x ≁ v ∼ y 12 R 7 = R T 1 x ≁ v , y ≁ v and z ∼ v 2 k ( k − 1 )( k − 2 ) k ≥ 3 11 R 8 = R T 1 2 k ( k − 1 ) 2 ( k − 2 ) x ≁ v , y ≁ v and z ≁ v k ≥ 3 13

  22. Let v 1 and v 2 be non-adjacent vertices in M . Let w be the common neighbour of v 1 and v 2 . Let A = N ( w ) \ { v 1 , v 2 } , B 1 = N ( v 1 ) \ { w } , B 2 = N ( v 2 ) \ { w } and C = V ( M ) \ ( { v 1 , v 2 , w } ∪ A ∪ B 1 ∪ B 2 ) . Then | A | = k − 2 , | B 1 | = | B 2 | = k − 1 , and | C | = k 2 + 1 − ( 3 + ( k − 2 ) + ( k − 1 ) + ( k − 1 )) = k 2 − 3 k + 2 = ( k − 1 )( k − 2 ) .

  23. w ① ✘ ❳❳❳❳❳❳❳❳❳❳❳ ✘ ✘ ✘ ✘ ✘ ① ✘ ① ✬ ✩ ✘ ✘ ✘ ✘ ✘ ❳ v 1 v 2 A ✬ ✩ ✬ ✩ ✫ ✪ B 1 B 2 ✫ ✪ ✫ ✪ ❆ ✁ ❆ ✁ ❆ ✁ ✬ ✩ ❆ ✁ ❆ ✁ ❆ ✁ ❆ ✁ ❆ ✁ ❆ ✁ C ✫ ✪

Recommend


More recommend