Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Stability of pair graphs

Stability of pair graphs STABILITY OF GRAPH PAIRS YAN-LI QIN, BINZHOU XIA, JIN-XIN ZHOU, AND SANMING ZHOU Abstract. We start up the study of the stability of general graph pairs. This notion is a generalization of the concept of the stability of graphs. We say that a pair of graphs (Γ, Σ) is stable if Aut(Γ × Σ) Aut(Γ) × Aut(Σ) and unstable otherwise, where Γ × Σ is the direct product of Γ and Σ. An unstable graph pair (Γ, Σ) is said to be a nontrivially unstable graph pair if Γ and Σ are connected coprime graphs, at least one of them is non-bipartite, and each of them has the property that different vertices have distinct neighbourhoods. We obtain necessary conditions for a pair of graphs to be stable. We also give a characterization of a pair of graphs (Γ, Σ) to be nontrivially unstable in the case when both graphs are connected and regular with coprime valencies and Σ is vertex-transitive. This characterization is given in terms of the Σ-automorphisms of Γ, which are a new concept introduced in this paper as a generalization of both automorphisms and two-fold automorphisms of a graph. Key words: stable graph; stable graph pair; direct product of graphs 1. Introduction All graphs considered in this paper are finite, undirected and simple, unless stated otherwise. As usual, for a graph Γ we use V (Γ), E(Γ) and Aut(Γ) to denote its vertex set, edge set and full automorphism group, respectively, and we use val(Γ) to denote the valency of Γ if Γ is regular. For a vertex u of Γ, the neighborhood of u in Γ, denoted by N (u), is the set of vertices adjacent to u in Γ. For two adjacent vertices u, v in a graph, the edge between them is denoted by the unordered pair {u, v}. Let Γ and Σ be graphs. The direct product of Γ and Σ, denoted by Γ × Σ, is the graph with vertex set V (Γ)×V (Σ) such that two vertices (u, x), (v, y) ∈ V (Γ)×V (Σ) are adjacent if and only if u and v are adjacent in Γ and x and y are adjacent in Σ. Clearly, Aut(Γ) × Aut(Σ) . Aut(Γ × Σ). (1) Herein and in the sequel we use X . Y to indicate that X is isomorphic to a subgroup of Y , and × on the left-hand side denotes the direct product of groups. In the literature much attention has been paid to the automorphism group Aut(Γ×Σ) of Γ × Σ. In particular, the question of when the equality in (1) holds has attracted considerable interest (see, for example, [3, 4]). In line with this we introduce the following definition. Definition 1.1. A graph pair (Γ, Σ) is called stable if Aut(Γ×Σ) Aut(Γ)×Aut(Σ) and unstable otherwise. arXiv:2010.16137v1 [math.CO] 30 Oct 2020 2 QIN, XIA, ZHOU, AND ZHOU Note that (Γ, Σ) is stable if and only if (Σ, Γ) is stable. The notion of the stability of graph pairs generalizes the concept of the stability of graphs [6], in the sense that a graph Γ is stable if and only if the graph pair (Γ, K ) is stable, where K is the 2 2 complete graph with two vertices. Introduced by Maruˇsiˇc et al. [6] in the language of symmetric (0, 1) matrices, the stability of graphs has been studied extensively (see, for example, [9, 11, 12, 14]) owing to its close connections with regular embeddings of canonical double covers [8], two-fold automorphisms of graphs [5], and general- ized Cayley graphs [7]. For example, the stability of circulant graphs was studied by Wilson in [14], and an open question in [14] about the stability of arc-transitive circulant graphs was answered and an infinite family of counterexamples to a con- jecture of Maruˇsiˇc et al. [6] was constructed by Qin et al. in [9]. A conjecture of Wilson [14] about the stability of generalized Petersen graphs was recently proved by Qin et al. in [10]. A graph Γ is said to be R-thick [4] if there exist distinct vertices u, v of Γ such that N (u) = N (v). Graphs that are not R-thick are said to be R-thin [4] or Γ Γ vertex-determining [9, 14]. A graph is said to be prime (with respect to the direct product) if it has order greater than 1 and cannot be represented as a direct product of two graphs of smaller orders, where the order of a graph is defined as its number of vertices. Since the direct product of graphs is an associative and commutative operation, the direct product of more than two graphs is well defined up to isomor- phism. An expression Γ = Γ × Γ × · · · × Γ with each Γ prime is called a prime 1 2 k i factorization of Γ (with respect to the direct product). It is well known [4, Theorem 8.17] that up to permutation of factors any non-bipartite graph with order greater than 1 has a unique prime factorization. Two graphs are called coprime (with re- spect to the direct product) if they do not have any common factor of order greater than 1. In particular, any two graphs of coprime orders must be coprime. Remark 1.2. If two graphs have a common factor with respect to the direct product, then their valencies must have a common divisor greater than 1. Thus regular graphs with coprime valencies must be coprime. Our first main result in this paper gives necessary conditions for a graph pair to be stable. Theorem 1.3. Let (Γ, Σ) be a stable pair of graphs. Then Γ and Σ are coprime R- thin graphs. Moreover, if in addition both Aut(Γ) and Aut(Σ) are nontrivial groups, then both Γ and Σ are connected and at least one of them is non-bipartite. In this paper we are only interested in graphs with nontrivial automorphism groups. Under this assumption Theorem 1.3 implies that in studying the stability of graph pairs we can focus on those pairs (Γ, Σ) such that Γ and Σ are connected coprime R-thin graphs and at least one of them is non-bipartite. This fact motivates the following definition. STABILITY OF GRAPH PAIRS 3 Definition 1.4. An unstable graph pair (Γ, Σ) is said to be nontrivially unstable if Γ and Σ are connected coprime R-thin graphs and at least one of them is non-bipartite. Since a graph not coprime to K is necessarily bipartite, we see that (Γ, K ) is 2 2 nontrivially unstable if and only if Γ is a non-bipartite connected R-thin unstable graph. Such a graph Γ is called nontrivially unstable by Wilson in [14] in a study of the stability of graphs. So the above definition of nontrivially unstable graph pairs generalizes the concept of nontrivially unstable graphs. Remark 1.5. It can be easily shown that Γ × Σ is R-thin if and only if both Γ and Σ are R-thin ([9, Lemma 2.3], stated as Lemma 3.1(b) in the present paper), and Γ × Σ is connected if and only if both Γ and Σ are connected and at least one of them is non-bipartite ([4, Theorem 5.9], stated as Lemma 3.1(a) in the present paper). Therefore, our definition of graph pairs (Γ, Σ) being nontrivially unstable is equivalent to requiring that Γ and Σ are coprime graphs with Γ × Σ connected and R-thin. Needless to say, the following problem is of central importance to the study of the stability of graph pairs. Problem 1.6. Characterize nontrivially unstable pairs of graphs (Γ, Σ) with both Aut(Σ) and Aut(Γ) nontrivial. We will study this problem in the case when Γ and Σ are regular graphs of coprime valencies. Our study is motivated by orientably regular embeddings of the canonical double cover Γ×K of a given graph Γ. More precisely, it was shown by Nedela and Skoviera [8] that for any stable graph Γ (that is, for any stable graph pair (Γ, K )), all orientably regular embeddings of Γ × K can be described in terms of orientably regular embeddings of Γ. As a natural extension, one would expect that for a stable pair (Γ, Σ) of graphs we may be able to describe all orientably regular embeddings of Γ × Σ in terms of orientably regular embeddings of Γ and Σ. In this regard it has been proved by Chen [2] that, if (Γ, Σ) is a stable pair of regular graphs such that Γ × Σ has an orientably regular embedding, then the valencies of Γ and Σ must be coprime. It is thus natural to impose the extra condition gcd(val(Γ), val(Σ)) = 1 when studying Problem 1.6, and we will do so in this paper. Note that this condition is satisfied by (Γ, K ) for any regular graph Γ. Let Γ be a graph. A pair of permutations (α, β) of V (Γ) is called a two-fold auto- α β morphism of Γ if for all u, v ∈ V (Γ), {u, v} ∈ E(Γ) if and only if {u , v } ∈ E(Γ). A two-fold automorphism (α, β) is said to be nontrivial if in addition α 6= β. It is proved in [5, Theorem 3.2] that a graph is unstable if and only if it has a nontrivial two-fold automorphism. The second main result in our paper, Theorem 1.8 below, generalizes this result to the setting of nontrvially unstable graph pairs when Γ is regular. To present our result we need the following definition. 4 QIN, XIA, ZHOU, AND ZHOU Definition 1.7. Let Γ and Σ be graphs with V (Σ) = {1, . . . , n}, and let α , . . . , α 1 n be permutations of V (Γ). We say that the n-tuple (α , . . . , α ) is a Σ-automorphism 1 n α α i j of Γ if for all u, v ∈ V (Γ), {u, v} ∈ E(Γ) if and only if {u , v } ∈ E(Γ) for all i, j ∈ V (Σ) with {i, j} ∈ E(Σ). Such a Σ-automorphism (α , . . . , α ) of Γ is said 1 n to be nondiagonal if there exists at least one pair of vertices i, j ∈ V (Σ) such that α 6= α . i j It is readily seen that a two-fold automorphism of a graph Γ is exactly a K - automorphism of Γ, and a nontrivial two-fold automorphism of Γ is precisely a nondiagonal K -automorphism of Γ. As a side note, we mention that the definition above applies when Γ and Σ are pseudographs. (A pseudograph is a graph in which both loops and multiple edges are permitted.) And we observe that, if Σ is the ◦ ◦ pseudograph K with only one vertex and one self-loop, then a K -automorphism 1 1 of Γ is an automorphism of Γ in the usual sense. Nevertheless, in this paper we only consider the case when both Γ and Σ are simple graphs with order at least two. The second main result in this paper, presented below, settles Problem 1.6 in the case when Γ and Σ are regular with coprime valencies and Σ is vertex-transitive. Theorem 1.8. Let Γ be a connected regular graph and Σ a connected vertex-transitive graph such that val(Γ) and val(Σ) are coprime. Suppose that both Γ and Σ are R- thin and at least one of them is non-bipartite. Then (Γ, Σ) is nontrvially unstable if and only if at least one Σ-automorphism of Γ is nondiagonal. The rest of this paper is organized as follows. In the next section we will introduce the Σ-automorphism group of Γ and two subgroups (see Definitions 2.1 and 2.3 in the next section) of Aut(Γ×Σ), and study connections between these groups and the stability of (Γ, Σ). The proof of Theorem 1.3 will be given in Section 3. In Section 4, we will prove a number of lemmas concerning pairs of graphs with coprime valencies. These lemmas will be used to prove Theorem 1.8 in Section 5. We will conclude the paper with some remarks and questions in Section 6. 2. Σ-automorphism group of Γ Throughout this section Γ and Σ are graphs with V (Σ) = {1, . . . , n} and n > 1. 2.1. Aut (Γ), P (Γ, Σ) and Q(Γ, Σ). It is not difficult to see that the set of all Σ-automorphisms of Γ with multiplication defined by (α , . . . , α )(β , . . . , β ) = (α β , . . . , α β ) 1 n 1 n 1 1 n n is a group. We call this group the Σ-automorphism group of Γ and denote it by Aut (Γ). Note that (α, α, . . . , α) ∈ Aut (Γ) if and only if α ∈ Aut(Γ). Hence Σ Σ Aut(Γ) . Aut (Γ). (2) Moreover, Aut(Γ) = Aut (Γ) ⇔ α = · · · = α for each (α , . . . , α ) ∈ Aut (Γ). (3) Σ 1 n 1 n Σ STABILITY OF GRAPH PAIRS 5 Definition 2.1. Define Q(Γ, Σ) = {σ ∈ Aut(Γ × Σ) : (V (Γ) × {i}) = V (Γ) × {i} for each i ∈ V (Σ)}. Note that Q(Γ, Σ) is a subgroup of Aut(Γ × Σ). Denote the projections from V (Γ × Σ) to V (Γ) and V (Σ) by π and π , respectively. In other words, Γ Σ π π Γ Σ (u, i) = u and (u, i) = i, for (u, i) ∈ V (Γ × Σ). Lemma 2.2. Let Γ and Σ be graphs. Then Q(Γ, Σ) Aut (Γ). Proof. Set V (Σ) = {1, . . . , n}. Define f : Q(Γ, Σ) → Sym(V (Γ)) × · · · × Sym(V (Γ)), σ 7→ (α , . . . , α ) 1 n σπ π α Γ Γ i such that (u, i) = (u, i) for each (u, i) ∈ V (Γ × Σ). For any σ, τ ∈ Q(Γ, Σ), let f(σ) = (α , . . . , α ) and f(τ) = (β , . . . , β ). It is straightforward to verify that 1 n 1 n στπ π (α β ) Γ Γ i i (u, i) = (u, i) for each (u, i) ∈ V (Γ × Σ). Hence f(σ)f(τ) = (α , . . . , α )(β , . . . , β ) 1 n 1 n = (α β , . . . , α β ) 1 1 n n = f(στ). Thus f is a group homomorphism. Since f is injective, we have Q(Γ, Σ) = f(Q(Γ, Σ)). σ α Let (α , . . . , α ) ∈ Aut (Γ). Define σ by (u, i) = (u , i) for all i ∈ V (Σ). Then 1 n Σ σ ∈ Q(Γ, Σ) and f(σ) = (α , . . . , α ). Thus Aut (Γ) ⊆ f(Q(Γ, Σ)). 1 n Σ Conversely, let σ ∈ Q(Γ, Σ) and f(σ) = (α , . . . , α ). Then for any i, j ∈ V (Σ) 1 n with {i, j} ∈ E(Σ), we have {u, v} ∈ E(Γ) ⇔ {(u, i), (v, j)} ∈ E(Γ × Σ) σ σ α α i j ⇔ {(u, i) , (v, j) } = {(u , i), (v , j)} ∈ E(Γ × Σ) α α i j ⇔ {u , v } ∈ E(Γ), α α i j whence {u, v} ∈ E(Γ) if and only if {u , v } ∈ E(Γ). Thus f(σ) = (α , . . . , α ) ∈ 1 n Aut (Γ), and so f(Q(Γ, Σ)) ⊆ Aut (Γ). Therefore, Q(Γ, Σ) f(Q(Γ, Σ)) = Σ Σ Aut (Γ), completing the proof. Obviously, {V (Γ) × {i} | i ∈ V (Σ)} is a partition of V (Γ × Σ). Definition 2.3. Define P (Γ, Σ) to be the set of elements of Aut(Γ × Σ) that leave the partition {V (Γ) × {i} | i ∈ V (Σ)} invariant. That is, P (Γ, Σ) is the setwise stabilizer of {V (Γ) × {i} | i ∈ V (Σ)} under Aut(Γ × Σ). Hence it is a subgroup of Aut(Γ × Σ). Of course P (Γ, Σ) induces an action on {V (Γ) × {i} | i ∈ V (Σ)}, and the kernel of this action is exactly Q(Γ, Σ). Hence P (Γ, Σ)/Q(Γ, Σ) = Aut(Σ). This together with Lemma 2.2 implies the following result. 6 QIN, XIA, ZHOU, AND ZHOU Lemma 2.4. Let Γ and Σ be graphs. Then P (Γ, Σ) Aut (Γ) ⋊ Aut(Σ). 2.2. P (Γ, Σ) and the stability of (Γ, Σ). Lemma 2.5. Let Γ and Σ be graphs. If (Γ, Σ) is stable, then Aut(Γ × Σ) = P (Γ, Σ). Proof. By (2) and Lemma 2.4, we have Aut(Γ) × Aut(Σ) . Aut (Γ) ⋊ Aut(Σ) P (Γ, Σ) 6 Aut(Γ × Σ). Thus, if (Γ, Σ) is stable, then Aut(Γ × Σ) = P (Γ, Σ). Lemma 2.6. Let Γ and Σ be graphs, where V (Σ) = {1, . . . , n}. (a) If at least one Σ-automorphism of Γ is nondiagonal, then (Γ, Σ) is unstable. (b) If Aut(Γ × Σ) = P (Γ, Σ), then (Γ, Σ) is unstable if and only if at least one Σ-automorphism of Γ is nondiagonal. Proof. (a) Suppose that there is at least one nondiagonal Σ-automorphism of Γ. Then we see from (2) and (3) that Aut(Γ) is isomorphic to a proper subgroup of Aut (Γ). Combining this with Lemma 2.4, we obtain |Aut(Γ) × Aut(Σ)| < |Aut (Γ) ⋊ Aut(Σ)| = |P (Γ, Σ)| 6 |Aut(Γ × Σ)|. Thus Aut(Γ × Σ) ≇ Aut(Γ) × Aut(Σ) and so (Γ, Σ) is unstable. (b) Suppose that Aut(Γ×Σ) = P (Γ, Σ). Then Aut(Γ×Σ) Aut (Γ)⋊Aut(Σ) by Lemma 2.4. Thus (Γ, Σ) is unstable if and only if Aut (Γ) ≇ Aut(Γ), which, by (2) and (3), is true if and only if at least one Σ-automorphism of Γ is nondiagonal. 3. Proof of Theorem 1.3 3.1. Preparation. We need a few lemmas before we can prove Theorem 1.3. First, the following known results will be used in our proofs of Theorems 1.3 and 1.8. Lemma 3.1. Let Γ and Σ be graphs. (a) If Γ and Σ are connected with order at least 2, then Γ×Σ is connected if at least one of Γ and Σ is non-bipartite, and Γ × Σ has exactly two components if both Γ and Σ are bipartite ([13]; see also [4, Theorem 5.9]). (b) Γ × Σ is R-thin if and only if both Γ and Σ are R-thin ([9, Lemma 2.3]). (c) Γ × Σ is non-bipartite if and only if both Γ and Σ are non-bipartite ([4, Exercise 8.13]). Lemma 3.2. Let Γ and Σ be graphs. If (Γ, Σ) is stable, then both Γ and Σ are R-thin. STABILITY OF GRAPH PAIRS 7 Proof. We prove the contrapositive of the statement. Suppose that at least one of Γ and Σ is R-thick. Without loss of generality we may assume that Γ is R-thick. Then there exist two distinct vertices u and v of Γ such that N (u) = N (v). Let α be Γ Γ the permutation of V (Γ) which swaps u and v and fixes each vertex in V (Γ)\{u, v}, and let τ = (α, 1, 1, . . ., 1). Then τ is a Σ-automorphism of Γ. Moreover, τ is nondiagonal as α 6= 1. Thus, by Lemma 2.6(a), (Γ, Σ) is unstable. Lemma 3.3. Let Γ and Σ be graphs. If one of them is disconnected and the other has a nontrivial automorphism group, then (Γ, Σ) is unstable. Proof. Without loss of generality we may assume that Γ is disconnected and Aut(Σ) 6= 1. Take a connected component Γ of Γ and an element α 6= 1 of Aut(Σ). Let σ be the permutation of V (Γ × Σ) which fixes each vertex in V (Γ × Σ) \ V (Γ × Σ) and permutes the vertices in V (Γ × Σ) in the following way: σ α (u, i) = (u, i ) for each (u, i) ∈ V (Γ × Σ). It is straightforward to verify that σ ∈ Aut(Γ × Σ) \ P (Γ, Σ). So (Γ, Σ) is unstable by Lemma 2.5. In regard to Lemma 3.3, there exist both stable pairs (Γ, Σ) and unstable pairs (Γ, Σ) with Γ disconnected and Aut(Σ) = 1, as illustrated by the following example. Example 3.4. Let Γ be the graph with V (Γ ) = {1, 2, 3, 4, 5} and 1 1 E(Γ ) = {{1, 2}, {3, 4}, {4, 5}, {3, 5}}. Let Γ be the graph with V (Γ ) = {1, 2, 3, 4, 5, 6, 7, 8} and 2 2 E(Γ ) = {{1, 2}, {2, 3}, {3, 4}, {4, 5}, {3, 6}, {4, 6}, {7, 8}}. Let Σ be the graph with V (Σ) = {1, 2, 3, 4, 5, 6} and E(Σ) = {{1, 2}, {2, 3}, {3, 4}, {4, 5}, {3, 6}, {4, 6}}. Then both Γ and Γ are disconnected and Aut(Σ) = 1. Computation in Magma [1] 1 2 shows that |Aut(Γ × Σ)| = |Aut(Γ )| = 12, |Aut(Γ )| = 2, and |Aut(Γ × Σ)| = 4. 1 1 2 2 Hence (Γ , Σ) is stable but (Γ , Σ) is unstable. 1 2 As usual, for a graph Γ and a subset U ⊆ V (Γ), we use hUi to denote the subgraph of Γ induced by U, namely the graph with vertex set U in which two vertices u, v ∈ U are adjacent if and only if they are adjacent in Γ. Lemma 3.5. Let Γ and Σ be connected bipartite graphs. If both Aut(Γ) and Aut(Σ) are nontrivial groups, then (Γ, Σ) is unstable. Proof. Let {B , B } be the bipartition of Γ and {C , C } the bipartition of Σ. Set 1 2 1 2 U = (B × C ) ∪ (B × C ) and U = (B × C ) ∪ (B × C ). 1 1 1 2 2 2 1 2 2 1 8 QIN, XIA, ZHOU, AND ZHOU Then hU i and hU i are two connected components of Γ×Σ = hU ∪U i. Since both 1 2 1 2 Aut(Γ) and Aut(Σ) are nontrivial, we have |Aut(Γ × Σ)| > |Aut(Γ)| · |Aut(Σ)| > 2 · 2 = 4. (4) If both Aut(hU i) and Aut(hU i) are trivial, then 1 2 1, if hU i ≇ hU i 1 2 Aut(Γ × Σ) = Aut(hU ∪ U i) = 1 2 Z , if hU i hU i, 2 1 2 which contradicts (4). Hence at least one of Aut(hU i) and Aut(hU i) is nontriv- 1 2 ial. Without loss generality we may assume that Aut(hU i) is nontrivial. Take an element 1 6= α ∈ Aut(hU i). Let σ be the permutation of V (Γ × Σ) defined by (u, i), if (u, i) ∈ U (u, i) = (5) (u, i) , if (u, i) ∈ U for (u, i) ∈ V (Γ × Σ). Then σ ∈ Aut(Γ × Σ). Suppose to the contrary that (Γ, Σ) is stable. Then by Lemma 2.5, Aut(Γ × Σ) = P (Γ, Σ), and so σ ∈ P (Γ, Σ). Set V (Σ) = {1, . . . , n} and let α , . . . , α be 1 n α σπ i Γ permutations of V (Γ) such that u = (u, i) for u ∈ V (Γ) and i ∈ V (Σ). Since σ ∈ P (Γ, Σ), by (5) we have for (u, i) ∈ V (Γ × Σ), (u, i), if (u, i) ∈ U (u, i) = (6) (u , i), if (u, i) ∈ U . Hence σ σ {u, v} ∈ E(Γ) ⇔ {(u, i) , (v, j) } ∈ E(Γ × Σ) for any i, j ∈ V (Σ) with {i, j} ∈ E(Σ) α α i j ⇔ {(u , i), (v , j)} ∈ E(Γ × Σ) for any i, j ∈ V (Σ) with {i, j} ∈ E(Σ) α α i j ⇔ {u , v } ∈ E(Γ) for any i, j ∈ V (Σ) with {i, j} ∈ E(Σ). In other words, (α , . . . , α ) is a Σ-automorphism of Γ. Since (Γ, Σ) is stable, by 1 n Lemma 2.6(a) we deduce that α = · · · = α . This together with (6) implies that σ 1 n fixes each vertex of Γ × Σ, whence σ = 1. It then follows from (5) that α = 1. This contradiction shows that (Γ, Σ) is unstable and the proof is complete. The following example shows that the condition that both Aut(Γ) and Aut(Σ) are nontrivial cannot be removed from Lemma 3.5 for otherwise the result may not be true. Example 3.6. Let Γ be the graph with V (Γ) = {1, 2, 3, 4, 5, 6, 7} and E(Γ) = {{1, 2}, {2, 3}, {3, 4}, {4, 5}, {2, 6}, {3, 7}, {6, 7}}. Let Σ = K with V (Σ) = {a, b}. Then both Γ and Σ are bipartite and Aut(Γ) = 1. Let σ be the permutation of V (Γ × Σ) which interchanges (i, a) and (i, b) for each i ∈ V (Γ). Then Aut(Γ × Σ) = ∼ ∼ hσi Z . Hence Aut(Γ × Σ) Aut(Γ) × Aut(Σ) and (Γ, Σ) is stable. = = 2 STABILITY OF GRAPH PAIRS 9 Let Σ and Γ be graphs each with a nontrivial automorphism group. If one of them is R-thick or disconnected, then by Lemma 3.2 or 3.3 we know that (Γ, Σ) is unstable. Part (b) of the following lemma determines the stability of (Γ, Σ) when both Σ and Γ are connected, R-thin and non-bipartite. Lemma 3.7. Let Γ and Σ be graphs. (a) If (Γ, Σ) is stable, then Γ and Σ are coprime. (b) If both Γ and Σ are connected, R-thin and non-bipartite, then (Γ, Σ) is stable if and only if Γ and Σ are coprime. Proof. (a) We prove the contrapositive of this statement. Suppose that Γ and Σ are not coprime. Then there exist graphs Γ , Σ and Δ such that Γ = Γ × Δ, 1 1 1 Σ = Σ × Δ and |V (Δ)| > 1. So Γ × Σ = Γ × Δ × Σ × Δ. Let σ be the 1 1 1 permutation of V (Γ× Σ) = V (Γ × Δ× Σ × Δ) defined by (x, u, y, v) = (x, v, y, u) 1 1 for (x, u, y, v) ∈ V (Γ × Δ × Σ × Δ). Since |V (Δ)| > 1, it is straightforward to 1 1 verify that σ ∈ Aut(Γ × Σ) \ P (Γ, Σ). This together with Lemma 2.5 implies that (Γ, Σ) is unstable. (b) The “only if” part follows from (a), so it remains to prove the “if” part. Since both Γ and Σ are connected, R-thin and non-bipartite, by Lemma 3.1, so is Γ × Σ. If Γ and Σ are coprime, then by [4, Theorem 8.18], we have Aut(Γ × Σ) = Aut(Γ) × Aut(Σ) and hence (Γ, Σ) is stable. 3.2. Proof of Theorem 1.3. We are now ready to prove Theorem 1.3. Proof. Suppose that (Γ, Σ) is stable. Then by Lemmas 3.2 and 3.7(a) we know that Γ and Σ are coprime R-thin graphs. Moreover, if both Aut(Γ) and Aut(Σ) are nontrivial, then by Lemmas 3.3 and 3.5 both Γ and Σ are connected and at least one of them is non-bipartite. 4. Pairs of regular graphs with coprime valencies As a preparation for our proof of Theorem 1.8, we study pairs of regular graphs with coprime valencies in this section. As before, throughout this section Γ and Σ are graphs with V (Σ) = {1, . . . , n} and n > 1. As usual, the set of positive integers is denoted by N. A key concept used in this section is the Boolean square [4] of a graph Δ, denoted by B(Δ), which is the graph with vertex set V (Δ) and edge set {{u, v} | u, v ∈ V (Δ), u 6= v, N (u) ∩ N (v) 6= ∅}. Δ Δ 4.1. Notation. For (u, i), (v, j) ∈ V (Γ × Σ), define |N ((u, i)) ∩ N ((v, j))| Γ×Σ Γ×Σ f ((u, i), (v, j)) = . Γ,Σ |N ((u, i))| Γ×Σ Define X (u, i) = {(v, j) ∈ V (Γ × Σ) \ {(u, i)} | val(Σ) · f ((u, i), (v, j)) ∈ N} Γ,Σ Γ,Σ 10 QIN, XIA, ZHOU, AND ZHOU and Y (u, i) = {(v, j) ∈ X (u, i) | f ((u, i), (v, j)) > f ((u, i), (w, j)) ∀w ∈ V (Γ)}. Γ,Σ Γ,Σ Γ,Σ Γ,Σ We will abbreviate f ((u, i), (v, j)), X (u, i) and Y (u, i) to f((u, i), (v, j)), Γ,Σ Γ,Σ Γ,Σ X(u, i) and Y (u, i), respectively, when there is no danger of confusion. The following lemma follows immediately from the definitions of Γ × Σ and f((u, i), (v, j)). Lemma 4.1. Let Γ and Σ be regular graphs, and let (u, i), (v, j) ∈ V (Γ × Σ). Then |N (u) ∩ N (v)| |N (i) ∩ N (j)| Γ Γ Σ Σ f((u, i), (v, j)) = · . (7) val(Γ) val(Σ) Now assume that Σ is vertex-transitive. Then the vertex-transitivity of Σ implies that there exist integers n > n > · · · > n ≥ 1 such that for any i ∈ V (Σ) and 1 2 t x ∈ N (i) we have B(Σ) |N (x) ∩ N (i)| ∈ {n , n , . . . , n } and |N (i)| > n . Σ Σ 1 2 t B(Σ) 1 Set D (i) = {i}, D (i) = {x ∈ N (i) | |N (x) ∩ N (i)| = n , 1 6 n < |N (i)|} for 1 ≤ k ≤ t, k B(Σ) Σ Σ k k B(Σ) D (i) = {x ∈ V (Σ) | N (i) ∩ N (x) = ∅}, t+1 Σ Σ and D (i) = ∅ for s > t + 2. Then t+1 t V (Σ) = ∪ D (i), val(Σ) = |D (i)| and N (i) = ∪ D (i). k k B(Σ) k k=1 k=1 k=1 For each (u, i) ∈ V (Γ × Σ), define X (u, i) = Y (u, i) = {(u, i)} 0 0 and X (u, i) = X(u, i) ∩ (V (Γ) × D (i)) for k ≥ 1. k k k−1 Define Y (u, i) to be the set of elements (v, j) of X(u, i) \ (∪ X (u, i)) such that k m m=0 k−1 f((u, i), (v, j)) > f((u, i), (w, y)) for all (w, y) ∈ X(u, i) \ (∪ X (u, i)). m=0 STABILITY OF GRAPH PAIRS 11 4.2. X(u, i) and X (u, i). Lemma 4.2. Let Γ and Σ be graphs. Then for any (u, i) ∈ V (Γ × Σ) and σ ∈ Aut(Γ × Σ), we have σ σ (X(u, i)) = X((u, i) ). Proof. For any (u, i), (v, j) ∈ V (Γ × Σ) and σ ∈ Aut(Γ × Σ), we have σ σ (u, i) 6= (v, j) ⇔ (u, i) 6= (v, j) and σ σ val(Σ) · f((u, i), (v, j)) ∈ N ⇔ val(Σ) · f((u, i) , (v, j) ) ∈ N. It follows that σ σ σ σ (v, j) ∈ (X(u, i)) ⇔ (v, j) ∈ X(u, i) ⇔ (v, j) ∈ X((u, i) ). σ σ Thus (X(u, i)) = X ((u, i) ) as required. Lemma 4.3. Let Γ and Σ be regular graphs with coprime valencies. Suppose that Γ is R-thin. Then for any (u, i) ∈ V (Γ × Σ) and (v, j) ∈ X(u, i), we have i 6= j. Proof. Suppose that i = j for some (u, i) ∈ V (Γ × Σ) and (v, j) ∈ X(u, i). Then |N (i) ∩ N (j)| = val(Σ). Since (v, j) ∈ X(u, i), it follows from Lemma 4.1 that Σ Σ |N (u) ∩ N (v)| Γ Γ · val(Σ) = val(Σ) · f((u, i), (v, j)) ∈ N. val(Γ) This together with gcd(val(Γ), val(Σ)) = 1 implies that |N (u) ∩ N (v)| = val(Γ). Γ Γ Since Γ is R-thin, we deduce that u = v, and so (u, i) = (v, j), a contradiction. Lemma 4.4. Let Γ and Σ be regular graphs with coprime valencies. Suppose that Σ is R-thin. Then for any u ∈ V (Γ), any σ ∈ Aut(Γ × Σ), and any edge {i, j} of B(Σ), we have σπ σπ Σ Σ (u, i) 6= (u, j) . Proof. Since {i, j} is an edge of B(Σ), we have i 6= j, and hence (u, i) 6= (u, j). σ σ Since σ ∈ Aut(Γ × Σ), we then have (u, i) 6= (u, j) . Suppose to the contrary that σπ σπ σ σ Σ Σ (u, i) = (u, j) = z for some z ∈ V (Σ). Then (u, i) = (g, z) and (u, j) = (h, z) for some g, h ∈ V (Γ) with g 6= h. Thus σ σ |N ((u, i) ) ∩ N ((u, j) )| = |N ((g, z)) ∩ N ((h, z))| Γ×Σ Γ×Σ Γ×Σ Γ×Σ = |N (g) ∩ N (h)| · |N (z)| Γ Γ Σ = |N (g) ∩ N (h)| · val(Σ). Γ Γ On the other hand, since σ ∈ Aut(Γ × Σ), we have σ σ |N ((u, i) ) ∩ N ((u, j) )| = |N ((u, i)) ∩ N ((u, j))| Γ×Σ Γ×Σ Γ×Σ Γ×Σ = |N (u)| · |N (i) ∩ N (j)| Γ Σ Σ = val(Γ) · |N (i) ∩ N (j)|. Σ Σ 12 QIN, XIA, ZHOU, AND ZHOU Hence |N (g) ∩ N (h)| · val(Σ) = val(Γ) · |N (i) ∩ N (j)|. Γ Γ Σ Σ Since gcd(val(Γ), val(Σ)) = 1, it follows that val(Σ) divides |N (i) ∩ N (j)|. Hence Σ Σ |N (i) ∩ N (j)| = val(Σ) or |N (i) ∩ N (j)| = 0. However, this is impossible as Σ Σ Σ Σ Σ is R-thin and {i, j} is an edge of B(Σ). This completes the proof. Lemma 4.5. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Then for any edge {u, v} of B(Γ), any edge {i, j} of B(Σ), and any element σ ∈ Aut(Γ × Σ), we have σπ σπ Σ Σ |N ((u, i) ) ∩ N ((v, j) )| > |N (i) ∩ N (j)|. (8) Σ Σ Σ Σ Proof. Since {u, v} is an edge of B(Γ), we have N (u) ∩ N (v) 6= ∅. Consider Γ Γ w ∈ N (u) ∩ N (v). We have Γ Γ {w} × (N (i) ∩ N (j)) ⊆ N ((u, i)) ∩ N ((v, j)). Σ Σ Γ×Σ Γ×Σ Since σ ∈ Aut(Γ × Σ), it follows that σ σ ({w} × (N (i) ∩ N (j))) ⊆ (N ((u, i)) ∩ N ((v, j))) Σ Σ Γ×Σ Γ×Σ σ σ = N ((u, i) ) ∩ N ((v, j) ). Γ×Σ Γ×Σ This together with Lemma 4.4 implies (8). Lemma 4.6. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Suppose further that Σ is vertex-transitive. Then for any (u, i) ∈ V (Γ × Σ), any σ ∈ Aut(Γ × Σ), and any integer k ≥ 0, we have σ σ (X (u, i)) = X ((u, i) ). k k Proof. Consider an arbitrary vertex (v, j) of X (u, i). By Lemma 4.2, σ σ σ σ (v, j) ∈ (X (u, i)) ⊆ (X(u, i)) = X((u, i) ). (9) Since X (u, i) = X(u, i) ∩ (V (Γ) × D (i)), we have j ∈ D (i) and (v, j) ∈ X(u, i), k k k whence {u, v} is an edge of B(Γ) and {i, j} is an edge of B(Σ). It then follows from Lemma 4.5 that σπ σπ Σ Σ |N ((u, i) ) ∩ N ((v, j) )| > |N (i) ∩ N (j)| = n . (10) Σ Σ Σ Σ k σ σ σ First assume that k = 0. Then (X (u, i)) = {(u, i) } = X ((u, i) ). 0 0 σ σ Next assume that k = 1. By (9), we have (v, j) ∈ X((u, i) ). By Lemma 4.3, we then have σπ σπ Σ Σ (u, i) 6= (v, j) , and so σπ σπ Σ Σ |N ((u, i) ) ∩ N ((v, j) )| 6 n . Σ Σ 1 This together with (10) implies that σπ σπ Σ Σ |N ((u, i) ) ∩ N ((v, j) )| = n . Σ Σ 1 STABILITY OF GRAPH PAIRS 13 σ σπ Thus (v, j) ∈ V (Γ) × D ((u, i) ). Combining this with (9), we then have σ σ σπ σ (v, j) ∈ X((u, i) ) ∩ (V (Γ) × D ((u, i) )) = (X (u, i) ). 1 1 Therefore, σ σ (X (u, i)) ⊆ X ((u, i) ). (11) 1 1 It then follows that −1 −1 σ σ σσ (X ((u, i) )) ⊆ X ((u, i) ) = X (u, i). 1 1 1 Thus σ σ |X ((u, i) )| 6 |(X (u, i)) |. 1 1 σ σ This together with (11) implies (X (u, i)) = X ((u, i) ) as required. 1 1 Now assume that k > 2. Suppose by induction that for all integers ℓ with 0 6 ℓ < k we have σ σ (X (u, i)) = X ((u, i) ). (12) ℓ ℓ k−1 σ σ Since (v, j) ∈ X (u, i), we have (v, j) ∈/ (∪ X (u, i)) . It then follows from (12) k m m=0 that σ k−1 σ k−1 σ (v, j) ∈/ ∪ (X (u, i)) = ∪ X ((u, i) ). m m m=0 m=0 That is, σ σ (v, j) ∈/ X ((u, i) ) for m = 0, 1, . . . , k − 1. By the definition of X ((u, i) and (9), we then obtain that σπ σπ Σ Σ (v, j) ∈/ D ((u, i) ) for m = 0, 1, . . . , k − 1. σπ σπ Σ Σ This together with (10) implies that (v, j) ∈ D ((u, i) ), and hence we drive from (9) that σ σ σπ σ (v, j) ∈ X((u, i) ) ∩ (V (Γ) × D ((u, i) )) = X ((u, i) ). k k Therefore, σ σ (X (u, i)) ⊆ X ((u, i) ). k k Note that this holds for any (u, i) ∈ V (Γ × Σ) and σ ∈ Aut(Γ × Σ). Replacing (u, i) σ −1 by (u, i) and σ by σ in this inclusion, we obtain −1 −1 σ σ σ σ (X ((u, i) )) ⊆ X (((u, i) ) ) = X (u, i), k k k or equivalently, σ σ X ((u, i) ) ⊆ (X (u, i)) . k k σ σ Therefore, (X (u, i)) = X ((u, i) ) and the proof is complete by induction. k k 14 QIN, XIA, ZHOU, AND ZHOU 4.3. Y (u, i) and Y (u, i). Lemma 4.7. Let Γ and Σ be regular graphs with coprime valencies. Suppose that Σ is Γ-thin. Then for any (u, i) ∈ V (Γ × Σ), we have Y (u, i) = {u} × N (i). B(Σ) Proof. Let j ∈ N (i). Then i 6= j and N (i) ∩ N (j) 6= ∅. Since Γ and Σ are B(Σ) Σ Σ regular, we drive from Lemma 4.1 that val(Σ) · f((u, i), (u, j)) = |N (i) ∩ N (j)| ∈ N. (13) Σ Σ For any w ∈ V (Γ), by Lemma 4.1 and (13) we have |N (u) ∩ N (w)| Γ Γ val(Σ) · f((u, i), (w, j)) = · |N (i) ∩ N (j)| Σ Σ val(Γ) 6 |N (i) ∩ N (j)| Σ Σ = val(Σ) · f((u, i), (u, j)). Hence f((u, i), (u, j)) > f((u, i), (w, j)) for all w ∈ V (Γ). (14) This together with (13) implies that (u, j) ∈ Y (u, i). Thus {u} × N (i) ⊆ Y (u, i). (15) B(Σ) Conversely, let (v, j) ∈ Y (u, i). Then (v, j) 6= (u, i), val(Σ) · f((u, i), (v, j)) ∈ N, and f((u, i), (v, j)) > f((u, i), (w, j)) for all w ∈ V (Γ). (16) Moreover, by Lemma 4.1, we have |N (u) ∩ N (v)| Γ Γ val(Σ) · f((u, i), (v, j)) = · |N (i) ∩ N (j)| ∈ N. (17) Σ Σ val(Γ) It follows that |N (i) ∩ N (j)| 6= 0, and so N (i) ∩ N (j) 6= ∅. Since (v, j) ∈ Σ Σ Σ Σ Y (u, i) ⊆ X(u, i), we derive from Lemma 4.3 that j 6= i, and hence j ∈ N (i). B(Σ) Thus, using the same argument as in the first paragraph of this proof, we can derive that (u, i), (u, j) and (w, j) satisfy (14). Combining this and (16), we then obtain v = u and therefore (v, j) ∈ {u} × N (i). So we have proved that B(Σ) Y (u, i) ⊆ {u} × N (i), B(Σ) which together with (15) yields Y (u, i) = {u} × N (i), as required. B(Σ) Lemma 4.8. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Then the following hold for any (u, i) ∈ V (Γ × Σ) and positive integer k: (a) Y (u, i) ⊆ X (u, i); k k (b) Y (u, i) = Y (u, i) ∩ (V (Γ) × D (i)). k k STABILITY OF GRAPH PAIRS 15 Proof. If Y (u, i) = ∅, then statement (a) is obvious. Now assume that Y (u, i) 6= ∅ k k and let (v, j) ∈ Y (u, i). Then k−1 (v, j) ∈ X(u, i) \ (∪ X (u, i)) (18) m=0 and k−1 f((u, i), (v, j)) > f((u, i), (w, k)) for all (w, k) ∈ X(u, i) \ (∪ X (u, i)). (19) m=0 Suppose by way of contradiction that (v, j) ∈/ X (u, i). Then by (18) we have (v, j) ∈ X(u, i) but (v, j) ∈/ X (u, i) for m = 0, 1, . . . , k − 1, k. Since X (u, i) = X(u, i) ∩ (V (Γ) × D (i)), we have m m (v, j) ∈/ V (Γ) × D (i) for m = 0, 1, . . . , k − 1, k. So j ∈ D (i) for some t > k. Thus D (i) 6= ∅ and so D (i) 6= ∅. Take y ∈ D (i). t t k k Then |N (i) ∩ N (j)| < |N (i) ∩ N (y)| (20) Σ Σ Σ Σ and k−1 (u, y) ∈ X(u, i) \ (∪ X (u, i)). m=0 Using (20), we can easily obtain f((u, i), (v, j)) < f((u, i), (u, y)), which contradicts (19). This contradiction shows that (v, j) ∈ X (u, i), and therefore Y (u, i) ⊆ k k X (u, i). This completes the proof of statement (a). Now we prove statement (b). Let (v, j) ∈ Y (u, i). By statement (a), we have (v, j) ∈ X (u, i) = X(u, i) ∩ (V (Γ) × D (i)) ⊆ X(u, i). (21) k k It then follows from Lemmas 4.1 and 4.3 that |N (u) ∩ N (v)| Γ Γ val(Σ) · f((u, i), (v, j)) = · |N (i) ∩ N (j)| ∈ N Σ Σ val(Γ) and i 6= j. Hence N (i) ∩ N (j) 6= ∅. Using Lemma 4.1, it is straightforward to Σ Σ verify that val(Σ) · f((u, i), (u, j)) ∈ N (22) and val(Σ) · f((u, i), (u, j)) > val(Σ) · f((u, i), (v, j)). (23) k−1 Moreover, by (21), we have j ∈ D (i), and so j ∈/ ∪ D (i). Since i 6= j, we derive k m m=0 from (22) that k−1 (u, j) ∈ X(u, i) \ (∪ X (u, i)). m=0 This combined with (19) and (23) implies that v = u. Since N (i) ∩ N (j) 6= ∅, Σ Σ we have j ∈ N (i), and thus we obtain from Lemma 4.7 and (21) that (v, j) ∈ B(Σ) Y (u, i) ∩ V (Γ) × D (i). Therefore, Y (u, i) ⊆ Y (u, i) ∩ (V (Γ) × D (i)). (24) k k Conversely, let (v, j) ∈ Y (u, i) ∩ (V (Γ) × D (i)). Since Y (u, i) ⊆ X(u, i), we have k−1 (v, j) ∈ X(u, i) ∩ (V (Γ) × D (i)) ⊆ X(u, i) \ (∪ X (u, i)). (25) k m m=0 16 QIN, XIA, ZHOU, AND ZHOU k−1 Consider an arbitrary element (w, t) of X(u, i) \ (∪ X (u, i)). Since X (u, i) = m m m=0 X(u, i) ∩ (V (Γ) × D (i)), we have (w, t) ∈/ X(u, i) ∩ (V (Γ) × D (i)) for m = 0, 1, . . . , k − 1. Since (w, t) ∈ X(u, i), we deduce that t ∈/ D (i) for m = 0, 1, . . . , k − 1, k−1 and hence j ∈/ ∪ D (i). Note that (v, j) ∈ D (i). It follows that m k m=0 |N (i) ∩ N (j)| > |N (i) ∩ N (t)|. (26) Σ Σ Σ Σ Since (v, j) ∈ Y (u, i), by Lemma 4.7 we obtain that u = v. Thus |N (u) ∩ N (v)| = val(Γ) > |N (u) ∩ N (w)|. Γ Γ Γ Γ This together with (26) and Lemma 4.1 implies that val(Σ) · f((u, i), (v, j)) = |N (i) ∩ N (j)| Σ Σ |N (u) ∩ N (w)| Γ Γ > · |N (i) ∩ N (t)| Σ Σ val(Γ) = val(Σ) · f((u, i), (w, t)). Combining this with (25), we obtain (v, j) ∈ Y (u, i). Thus Y (u, i) ∩ (V (Γ) × D (i)) ⊆ Y (u, i). k k This together with (24) completes the proof of statement (b). Lemma 4.9. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Suppose further that Σ is vertex-transitive. Then for any (u, i) ∈ V (Γ × Σ) and σ ∈ Aut(Γ × Σ), we have σ σ (Y (u, i)) = Y ((u, i) ). t+1 Proof. Since Σ is finite, there exists a positive integer t such that V (Σ) = ∪ D (i), m=0 where as before D (i) = {i} and D (i) = {x ∈ V (Σ) | N (i) ∩ N (x) = ∅}. Since 0 t+1 Σ Σ Y (u, i) ⊆ X(u, i), by Lemma 4.3, we have Y (u, i) ∩ (V (Γ) × D (i)) = ∅. (27) For any (v, j) ∈ V (Γ) × D (i), we have j ∈/ N (i) as N (i) ∩ N (j) = ∅. By t+1 B(Σ) Σ Σ Lemma 4.7, we then have (v, j) ∈/ Y (u, i) and therefore Y (u, i) ∩ (V (Γ) × D (i)) = ∅. (28) t+1 Note that V (Γ × Σ) = V (Γ) × V (Σ) t+1 = V (Γ) × ∪ D (i) m=0 t+1 = ∪ (V (Γ) × D (i)) m=0 = ∪ (V (Γ) × D (i)) ∪ (V (Γ) × D (i)) ∪ (V (Γ) × D (i)) . m 0 t+1 m=1 STABILITY OF GRAPH PAIRS 17 Thus, by (27), (28) and Lemma 4.8(b), we have Y (u, i) = Y (u, i) ∩ V (Γ × Σ) = Y (u, i) ∩ ∪ (V (Γ) × D (i)) m=1 = ∪ (Y (u, i) ∩ (V (Γ) × D (i)) m=1 = ∪ Y (u, i). m=1 Since σ ∈ Aut(Γ × Σ), by Lemmas 4.2 and 4.6, we obtain that for each positive integer k, k−1 σ k−1 σ X(u, i) \ ∪ X (u, i) = (X(u, i)) \ ∪ (X (u, i)) m m m=0 m=0 σ k−1 σ = X((u, i) ) \ ∪ X ((u, i) ) . m=0 σ σ Thus (Y (u, i)) = Y ((u, i) ) by the definition of Y (u, i). Therefore, k k k σ t t σ t σ σ (Y (u, i)) = ∪ Y (u, i) = ∪ (Y (u, i)) = ∪ Y ((u, i) ) = Y ((u, i) ), m m m m=1 m=1 m=1 completing the proof. 5. Proof of Theorem 1.8 5.1. Lemmas. We need the following lemmas in our proof of Theorem 1.8. Lemma 5.1. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Suppose further that Σ is vertex-transitive. Let u ∈ V (Γ) and i, j ∈ V (Σ). If N (i) ∩ N (j) 6= ∅, then for any σ ∈ Aut(Γ × Σ), we have Σ Σ σπ σπ Γ Γ (u, i) = (u, j) . σ σ Proof. If i = j, then (u, i) = (u, j) and so the statement is true. Now assume that i 6= j and N (i) ∩ N (j) 6= ∅. Then j ∈ N (i). By Lemma 4.7, we then Σ Σ B(Σ) σ σ σ have (u, j) ∈ Y (u, i). Hence (u, j) ∈ (Y (u, i)) = Y ((u, i) ) by Lemma 4.9. This σπ σπ Γ Γ together with Lemma 4.7 implies that (u, i) = (u, j) , as required. Lemma 5.2. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Suppose further that Σ is vertex-transitive. Let u, v ∈ V (Γ) and i ∈ V (Σ). If there is a walk from u to v of even length, then for any σ ∈ Aut(Γ × Σ), we have σπ σπ Σ Σ (u, i) = (v, i) . (29) Proof. It suffices to prove (29) in the case when N (u) ∩ N (v) 6= ∅. The result in Γ Γ the general case follows by applying this equation to pairs of every other vertices on the walk. So let us assume that N (u) ∩ N (v) 6= ∅. Take a vertex w ∈ N (u) ∩ N (v). Set Γ Γ Γ Γ W = {w} × N (i). Then W ⊆ N ((u, i)) ∩ N ((v, i)). Since σ ∈ Aut(Γ × Σ), Σ Γ×Σ Γ×Σ it follows that σ σ σ W ⊆ N ((u, i) ) ∩ N ((v, i) ). Γ×Σ Γ×Σ 18 QIN, XIA, ZHOU, AND ZHOU Consequently, σπ σπ σπ Σ Σ Σ W ⊆ N ((u, i) ) ∩ N ((v, i) ). (30) Σ Σ Note that σ σπ σπ Σ Σ |W | = |W| = |N (i)| = |N ((u, i) )| = |N ((v, i) )| = val(Σ). (31) Σ Σ Σ Consider any two distinct vertices (w, j), (w, k) in W . We have j 6= k ∈ N (i) and i ∈ N (j) ∩ N (k), whence N (j) ∩ N (k) 6= ∅. By Lemma 5.1, we have Σ Σ Σ Σ σπ σπ Γ Γ (w, j) = (w, k) . Since (w, j) 6= (w, k) and σ ∈ Aut(Γ × Σ), we then obtain σπ σπ σπ σ Σ Σ Σ that (w, j) 6= (w, k) . Hence |W | = |W |. Combining this with (30) and (31), we obtain that σπ σπ σπ Σ Σ Σ N ((u, i) ) = N ((v, i) ) = W . Σ Σ σπ σπ Σ Σ Since Σ is R-thin, we then conclude that (u, i) = (v, i) . For a graph Δ and a partition B of V (Δ), the quotient graph Δ of Δ with respect to B is defined to have vertex set B such that two blocks B, C ∈ B are adjacent if and only if there exists at least one edge of Δ with one end-vertex in B and the other end-vertex in C. Lemma 5.3. Let Γ and Σ be connected regular graphs with coprime valencies. Sup- pose that both Γ and Σ are R-thin. Suppose further that Γ is bipartite and Σ is vertex-transitive and non-bipartite. Then Aut(Γ × Σ) = P (Γ, Σ). Proof. Let B and B be the biparts of Γ, and let V (Σ) = {1, 2, . . ., n}. Let u, v ∈ 0 1 V (Γ) and i ∈ V (Σ), and let σ be an arbitrary element of Aut(Γ × Σ). Let Δ = Γ × Σ. Set B = {V (Γ) × {j} | j ∈ V (Σ)} and D = {B × {j}, B × {j} | j ∈ V (Σ)}. 0 1 Then B and D are partitions of V (Δ), and hence Δ and Δ are well defined. B D First assume that u and v are in the same biparts of Γ. Since Γ is connected and bipartite, there is a walk in Γ from u to v of even length. It then follows from σπ σπ Σ Σ Lemma 5.2 that (u, i) = (v, i) . Hence σ preserves D, and so σ ∈ Aut(Δ ). Next assume that u and v are in different biparts of Γ. Without loss generality we may assume that u ∈ B and v ∈ B . Since Σ is non-bipartite and vertex-transitive, 0 1 every vertex of Σ is contained in an odd cycle, and so there exists a cycle of odd length containing i, say, C : i , i , . . . , i , i , where i = i and ℓ is odd. It follows 1 2 ℓ 1 1 that C : V (Γ) × {i }, V (Γ) × {i }, . . . , V (Γ) × {i }, V (Γ) × {i } B 1 2 ℓ 1 STABILITY OF GRAPH PAIRS 19 and C : B ×{i }, B ×{i }, . . . , B ×{i }, B ×{i }, B ×{i }, . . . , B ×{i }, B ×{i } D 0 1 1 2 0 ℓ 1 1 0 2 1 ℓ 0 1 are cycles in Δ and Δ , respectively. Note that C is of length 2ℓ. Since σ ∈ B D D Aut(Δ ) as shown above, σ maps C to the cycle (C ) of Δ , and also maps the D D D D pair of antipodals B × {i}, B × {i} to some pair of antipodals in (C ) . Note also 0 1 D π π Σ Σ that for each pair of antipodals X, Y of some cycle in Δ , we have X = Y and π π Σ Σ |X | = |Y | = 1. In particular, we have σπ σπ Σ Σ (B × {i}) = (B × {i}) 0 1 and σπ σπ Σ Σ |(B × {i}) | = |(B × {i}) | = 1. 0 1 σπ σπ Σ Σ Since u ∈ B and v ∈ B , we then obtain that (u, i) = (v, i) . Since this 0 1 holds for any u, v ∈ V (Γ) and i ∈ V (Σ), we conclude that σ ∈ P (Γ, Σ). Since this holds for any σ ∈ Aut(Γ × Σ), we obtain that Aut(Γ × Σ) 6 P (Γ, Σ), which yields Aut(Γ × Σ) = P (Γ, Σ). Lemma 5.4. Let Γ and Σ be connected regular graphs with coprime valencies. Sup- pose that both Γ and Σ are R-thin. Suppose further that Γ is non-bipartite and Σ is vertex-transitive and bipartite. Then Aut(Γ × Σ) = P (Γ, Σ). Proof. Let Δ = Γ × Σ and B = {V (Γ) × {j} | j ∈ V (Σ)}. Then the quotient graph Δ of Δ with respect to B is isomorphic to Σ. Since Σ is bipartite, it follows that there is no odd cycle in Δ . This implies that, for any u, v ∈ V (Γ) and i ∈ V (Σ), there is no path of odd length from (u, i) to (v, i) in Δ. Since Γ is non-bipartite, we obtain from Lemma 3.1(a) that Δ is connected. It follows that there is a path of even length from (u, i) to (v, i) in Δ, and therefore there is a walk of even length from u to v in Γ. Thus, by Lemma 5.2, for any σπ σπ Σ Σ σ ∈ Aut(Γ × Σ), we have (u, i) = (v, i) , which implies σ ∈ P (Γ, Σ). Since this holds for any σ ∈ Aut(Γ × Σ), it follows that Aut(Γ × Σ) 6 P (Γ, Σ), yielding Aut(Γ × Σ) = P (Γ, Σ) as required. 5.2. Proof of Theorem 1.8. We are now ready to prove Theorem 1.8. Proof. Let Γ be a regular graph and Σ a vertex-transitive graph such that val(Γ) and val(Σ) are coprime. Suppose that (Γ, Σ) is nontrivially unstable. Then by the definition of a nontriv- ially unstable graph pair, Γ and Σ are coprime connected R-thin graphs and at least one of them is non-bipartite. Hence, by Lemma 3.1, Γ × Σ is connected and R-thin. Moreover, by Lemmas 5.3 and 5.4, we have Aut(Γ × Σ) = P (Γ, Σ). Since (Γ, Σ) is 20 QIN, XIA, ZHOU, AND ZHOU unstable, it follows from Lemma 2.6(b) that at least one Σ-automorphism of Γ is nondiagonal. Conversely, suppose that Γ × Σ is connected and R-thin and at least one Σ- automorphism of Γ is nondiagonal. Then, by Lemma 3.1, Γ and Σ are connected R-thin graphs and at least one of them is non-bipartite. If both Γ and Σ are non- bipartite, then by Lemma 3.7(b), (Γ, Σ) is stable, but this contradicts Lemma 2.6(a) as we assume that at least one Σ-automorphism of Γ is nondiagonal. Thus exactly one of Γ and Σ is non-bipartite. Since val(Γ) and val(Σ) are coprime, by Remark 1.2, Γ and Σ are coprime. Finally, by Lemmas 5.3 and 5.4, we have Aut(Γ×Σ) = P (Γ, Σ). Hence, by Lemma 2.6(b), (Γ, Σ) is unstable. Moreover, (Γ, Σ) is nontrivially unstable since Γ and Σ are coprime connected R-thin graphs. 6. Concluding remarks In the case when Σ = K , Theorem 1.8 gives rise to the following result: A con- nected regular graph is unstable if and only if it has a nontrivial two-fold automor- phism. As mentioned in the introduction, this is a special case of [5, Theorem 3.2], where the graph in the statement is not required to be regular. So Theorem 1.8 is a partial generalization of [5, Theorem 3.2]. It would be interesting to study whether the result in Theorem 1.8 is still true if Σ is not required to be regular with valency coprime to the valency of Γ. As seen in Theorem 1.8, there are close connections between stability of graph pairs (Γ, Σ) and Σ-automorphisms of Γ. However, the notion of Σ-automorphisms of Γ deserves further studies for its own sake. As far as we know, there are not many results on Σ-automorphisms of Γ in the literature, except in the case when ◦ ◦ Σ = K for which a K -automorphism of Γ is an automorphism of Γ in the usual 1 1 sense, and in the case when Σ = K for which a K -automorphism of Γ is exactly 2 2 a two-fold automorphism of Γ [5]. The concept of two-fold automorphisms was first introduced by Zelinka in [15, 16] for digraphs in his study of isotopies of digraphs and was extended to mixed graphs by Lauri et al. in [5]. It is readily seen that, if Σ is a spanning subgraph of Σ , then Aut (Γ) ≤ Aut (Γ). So among all graphs Σ 2 Σ Σ 2 1 of order n the complete graph K gives rise to the smallest possible group Aut (Γ). n Σ Therefore, it would be interesting to study the groups Aut (Γ) for n ≥ 2. Note TF that Aut (Γ) is exactly the two-fold automorphism group Aut (Γ) of Γ [5]. Finally, many questions about (nontrivially) unstable graph pairs may be asked. For example, by Lemma 2.6(a), if Γ admits a nondiagonal Σ-automorphism then (Γ, Σ) is unstable, and Theorem 1.8 determines a situation where this necessary condition is also sufficient. In general, one may ask the following question: For an unstable pair of graphs (Γ, Σ), under what conditions does Γ admit a nondiagonal Σ-automorphism? One may also study the stability of (Γ, Σ) for various special families of graphs Γ and/or various special families of graphs Σ. Possible candidates for Γ include circulant graphs [9, 14], arc-transitive graphs, generalized Petersen STABILITY OF GRAPH PAIRS 21 graphs GP(n, k) [10], etc. and potential choices for Σ include complete graphs K , complete bipartite graphs K , cycles C , etc. Obviously, this area of research is n,n n wide open and pleasantly inviting. Acknowledgements. We are grateful to the anonymous referees for their helpful comments and Dr Jiyong Chen for his valuable advices. Part of the work was done during a visit of the first author to The University of Melbourne. The first author would like to thank The University of Melbourne for its hospitality during her visit and Beijing Jiaotong University for its financial support. She also thanks the National Natural Science Foundation of China (11671030) for its financial support during her PhD. The first author was supported by the Fundamental Research Funds for Beijing Universities allocated to Capital University of Economics and Business(XRZ2020058). The third author was supported by the National Natural Science Foundation of China (11671030,12071023). References [1] W. Bosma, J. Cannon and C. Playoust, The Magma Algebra System I: The User Language, J. Symbolic Comput. 24 (1997), 235–265. [2] J. Chen, Embeddings of direct product graphs, in preparation, personal communication. [3] W. D¨orfler, Primfaktorzerlegung und Automorphismen des Kardinalproduktes von Graphen, Glasnik Mat. Ser. III 9(29) (1974), 15–27. [4] R. Hammack, W. Imrich and S. Klavˇzar, Handbook of Product Graphs, 2nd ed., CRC Press [5] J. Lauri, R. Mizzi and R. Scapellato, Unstable graphs: a fresh outlook via TF-automorphisms, Ars Math. Contemp. 8 (2015), 115–131. [6] D. Maruˇsiˇc, R. Scapellato and N. Zagaglia Salvi, A characterization of particular symmetric (0, 1) matrices, Linear Algebra Appl. 119 (1989), 153–162. [7] D. Maruˇsiˇc, R. Scapellato, and N. Zagaglia Salvi, Generalized Cayley graphs, Discrete Math. 102 (1992), 279–285. [8] R. Nedela and M. Skoviera, Regular embeddings of canonical double coverings of graphs, J. Combin. Theory Ser. B 67 (1996), 249–277. [9] Y-L. Qin, B. Xia and S. Zhou, Stability of circulant graphs, J. Combin. Theory Ser. B, 136 (2019), 154–169. [10] Y-L. Qin, B. Xia and S. Zhou, Stability of generalized Petersen graphs, J. Graph Theory, accepted, DOI: 10.1002/jgt.22642. [11] D. Surowski, Stability of arc-transitive graphs, J. Graph Theory 38 (2001), 95–110. [12] D. Surowski, Automorphism groups of certain unstable graphs, Math. Slovaca 53 (2003), 215–232. [13] P. M. Weichsel, The Kronecker product of graphs, Proc. Amer. Math. Soc. 13 (1962), 47–52. [14] S. Wilson, Unexpected symmetries in unstable graphs, J. Combin. Theory Ser. B 98 (2008), 359–383. [15] B. Zelinka, The group of autotopies of a digraph, Czech Math. J. 21 (1971), 619–624. [16] B. Zelinka, Isotopy of digraphs, Czech Math. J. 22 (1972), 353–360. 22 QIN, XIA, ZHOU, AND ZHOU School of Statistics, Capital University of Economics and Business, Beijing, 100070, P. R. China Email address: ylqin@cueb.edu.cn School of Mathematics and Statistics, The University of Melbourne, Parkville, VIC 3010, Australia Email address: binzhoux@unimelb.edu.au Department of Mathematics, Beijing Jiaotong University, Beijing, 100044, P. R. China Email address: jxzhou@bjtu.edu.cn School of Mathematics and Statistics, The University of Melbourne, Parkville, VIC 3010, Australia Email address: sanming@unimelb.edu.au http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Mathematics arXiv (Cornell University)

Stability of pair graphs

Mathematics , Volume 2020 (2010) – Oct 30, 2020

Loading next page...
 
/lp/arxiv-cornell-university/stability-of-pair-graphs-b9KcvIx02y

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

ISSN
0095-8956
eISSN
ARCH-3343
DOI
10.1016/j.jctb.2020.10.002
Publisher site
See Article on Publisher Site

Abstract

STABILITY OF GRAPH PAIRS YAN-LI QIN, BINZHOU XIA, JIN-XIN ZHOU, AND SANMING ZHOU Abstract. We start up the study of the stability of general graph pairs. This notion is a generalization of the concept of the stability of graphs. We say that a pair of graphs (Γ, Σ) is stable if Aut(Γ × Σ) Aut(Γ) × Aut(Σ) and unstable otherwise, where Γ × Σ is the direct product of Γ and Σ. An unstable graph pair (Γ, Σ) is said to be a nontrivially unstable graph pair if Γ and Σ are connected coprime graphs, at least one of them is non-bipartite, and each of them has the property that different vertices have distinct neighbourhoods. We obtain necessary conditions for a pair of graphs to be stable. We also give a characterization of a pair of graphs (Γ, Σ) to be nontrivially unstable in the case when both graphs are connected and regular with coprime valencies and Σ is vertex-transitive. This characterization is given in terms of the Σ-automorphisms of Γ, which are a new concept introduced in this paper as a generalization of both automorphisms and two-fold automorphisms of a graph. Key words: stable graph; stable graph pair; direct product of graphs 1. Introduction All graphs considered in this paper are finite, undirected and simple, unless stated otherwise. As usual, for a graph Γ we use V (Γ), E(Γ) and Aut(Γ) to denote its vertex set, edge set and full automorphism group, respectively, and we use val(Γ) to denote the valency of Γ if Γ is regular. For a vertex u of Γ, the neighborhood of u in Γ, denoted by N (u), is the set of vertices adjacent to u in Γ. For two adjacent vertices u, v in a graph, the edge between them is denoted by the unordered pair {u, v}. Let Γ and Σ be graphs. The direct product of Γ and Σ, denoted by Γ × Σ, is the graph with vertex set V (Γ)×V (Σ) such that two vertices (u, x), (v, y) ∈ V (Γ)×V (Σ) are adjacent if and only if u and v are adjacent in Γ and x and y are adjacent in Σ. Clearly, Aut(Γ) × Aut(Σ) . Aut(Γ × Σ). (1) Herein and in the sequel we use X . Y to indicate that X is isomorphic to a subgroup of Y , and × on the left-hand side denotes the direct product of groups. In the literature much attention has been paid to the automorphism group Aut(Γ×Σ) of Γ × Σ. In particular, the question of when the equality in (1) holds has attracted considerable interest (see, for example, [3, 4]). In line with this we introduce the following definition. Definition 1.1. A graph pair (Γ, Σ) is called stable if Aut(Γ×Σ) Aut(Γ)×Aut(Σ) and unstable otherwise. arXiv:2010.16137v1 [math.CO] 30 Oct 2020 2 QIN, XIA, ZHOU, AND ZHOU Note that (Γ, Σ) is stable if and only if (Σ, Γ) is stable. The notion of the stability of graph pairs generalizes the concept of the stability of graphs [6], in the sense that a graph Γ is stable if and only if the graph pair (Γ, K ) is stable, where K is the 2 2 complete graph with two vertices. Introduced by Maruˇsiˇc et al. [6] in the language of symmetric (0, 1) matrices, the stability of graphs has been studied extensively (see, for example, [9, 11, 12, 14]) owing to its close connections with regular embeddings of canonical double covers [8], two-fold automorphisms of graphs [5], and general- ized Cayley graphs [7]. For example, the stability of circulant graphs was studied by Wilson in [14], and an open question in [14] about the stability of arc-transitive circulant graphs was answered and an infinite family of counterexamples to a con- jecture of Maruˇsiˇc et al. [6] was constructed by Qin et al. in [9]. A conjecture of Wilson [14] about the stability of generalized Petersen graphs was recently proved by Qin et al. in [10]. A graph Γ is said to be R-thick [4] if there exist distinct vertices u, v of Γ such that N (u) = N (v). Graphs that are not R-thick are said to be R-thin [4] or Γ Γ vertex-determining [9, 14]. A graph is said to be prime (with respect to the direct product) if it has order greater than 1 and cannot be represented as a direct product of two graphs of smaller orders, where the order of a graph is defined as its number of vertices. Since the direct product of graphs is an associative and commutative operation, the direct product of more than two graphs is well defined up to isomor- phism. An expression Γ = Γ × Γ × · · · × Γ with each Γ prime is called a prime 1 2 k i factorization of Γ (with respect to the direct product). It is well known [4, Theorem 8.17] that up to permutation of factors any non-bipartite graph with order greater than 1 has a unique prime factorization. Two graphs are called coprime (with re- spect to the direct product) if they do not have any common factor of order greater than 1. In particular, any two graphs of coprime orders must be coprime. Remark 1.2. If two graphs have a common factor with respect to the direct product, then their valencies must have a common divisor greater than 1. Thus regular graphs with coprime valencies must be coprime. Our first main result in this paper gives necessary conditions for a graph pair to be stable. Theorem 1.3. Let (Γ, Σ) be a stable pair of graphs. Then Γ and Σ are coprime R- thin graphs. Moreover, if in addition both Aut(Γ) and Aut(Σ) are nontrivial groups, then both Γ and Σ are connected and at least one of them is non-bipartite. In this paper we are only interested in graphs with nontrivial automorphism groups. Under this assumption Theorem 1.3 implies that in studying the stability of graph pairs we can focus on those pairs (Γ, Σ) such that Γ and Σ are connected coprime R-thin graphs and at least one of them is non-bipartite. This fact motivates the following definition. STABILITY OF GRAPH PAIRS 3 Definition 1.4. An unstable graph pair (Γ, Σ) is said to be nontrivially unstable if Γ and Σ are connected coprime R-thin graphs and at least one of them is non-bipartite. Since a graph not coprime to K is necessarily bipartite, we see that (Γ, K ) is 2 2 nontrivially unstable if and only if Γ is a non-bipartite connected R-thin unstable graph. Such a graph Γ is called nontrivially unstable by Wilson in [14] in a study of the stability of graphs. So the above definition of nontrivially unstable graph pairs generalizes the concept of nontrivially unstable graphs. Remark 1.5. It can be easily shown that Γ × Σ is R-thin if and only if both Γ and Σ are R-thin ([9, Lemma 2.3], stated as Lemma 3.1(b) in the present paper), and Γ × Σ is connected if and only if both Γ and Σ are connected and at least one of them is non-bipartite ([4, Theorem 5.9], stated as Lemma 3.1(a) in the present paper). Therefore, our definition of graph pairs (Γ, Σ) being nontrivially unstable is equivalent to requiring that Γ and Σ are coprime graphs with Γ × Σ connected and R-thin. Needless to say, the following problem is of central importance to the study of the stability of graph pairs. Problem 1.6. Characterize nontrivially unstable pairs of graphs (Γ, Σ) with both Aut(Σ) and Aut(Γ) nontrivial. We will study this problem in the case when Γ and Σ are regular graphs of coprime valencies. Our study is motivated by orientably regular embeddings of the canonical double cover Γ×K of a given graph Γ. More precisely, it was shown by Nedela and Skoviera [8] that for any stable graph Γ (that is, for any stable graph pair (Γ, K )), all orientably regular embeddings of Γ × K can be described in terms of orientably regular embeddings of Γ. As a natural extension, one would expect that for a stable pair (Γ, Σ) of graphs we may be able to describe all orientably regular embeddings of Γ × Σ in terms of orientably regular embeddings of Γ and Σ. In this regard it has been proved by Chen [2] that, if (Γ, Σ) is a stable pair of regular graphs such that Γ × Σ has an orientably regular embedding, then the valencies of Γ and Σ must be coprime. It is thus natural to impose the extra condition gcd(val(Γ), val(Σ)) = 1 when studying Problem 1.6, and we will do so in this paper. Note that this condition is satisfied by (Γ, K ) for any regular graph Γ. Let Γ be a graph. A pair of permutations (α, β) of V (Γ) is called a two-fold auto- α β morphism of Γ if for all u, v ∈ V (Γ), {u, v} ∈ E(Γ) if and only if {u , v } ∈ E(Γ). A two-fold automorphism (α, β) is said to be nontrivial if in addition α 6= β. It is proved in [5, Theorem 3.2] that a graph is unstable if and only if it has a nontrivial two-fold automorphism. The second main result in our paper, Theorem 1.8 below, generalizes this result to the setting of nontrvially unstable graph pairs when Γ is regular. To present our result we need the following definition. 4 QIN, XIA, ZHOU, AND ZHOU Definition 1.7. Let Γ and Σ be graphs with V (Σ) = {1, . . . , n}, and let α , . . . , α 1 n be permutations of V (Γ). We say that the n-tuple (α , . . . , α ) is a Σ-automorphism 1 n α α i j of Γ if for all u, v ∈ V (Γ), {u, v} ∈ E(Γ) if and only if {u , v } ∈ E(Γ) for all i, j ∈ V (Σ) with {i, j} ∈ E(Σ). Such a Σ-automorphism (α , . . . , α ) of Γ is said 1 n to be nondiagonal if there exists at least one pair of vertices i, j ∈ V (Σ) such that α 6= α . i j It is readily seen that a two-fold automorphism of a graph Γ is exactly a K - automorphism of Γ, and a nontrivial two-fold automorphism of Γ is precisely a nondiagonal K -automorphism of Γ. As a side note, we mention that the definition above applies when Γ and Σ are pseudographs. (A pseudograph is a graph in which both loops and multiple edges are permitted.) And we observe that, if Σ is the ◦ ◦ pseudograph K with only one vertex and one self-loop, then a K -automorphism 1 1 of Γ is an automorphism of Γ in the usual sense. Nevertheless, in this paper we only consider the case when both Γ and Σ are simple graphs with order at least two. The second main result in this paper, presented below, settles Problem 1.6 in the case when Γ and Σ are regular with coprime valencies and Σ is vertex-transitive. Theorem 1.8. Let Γ be a connected regular graph and Σ a connected vertex-transitive graph such that val(Γ) and val(Σ) are coprime. Suppose that both Γ and Σ are R- thin and at least one of them is non-bipartite. Then (Γ, Σ) is nontrvially unstable if and only if at least one Σ-automorphism of Γ is nondiagonal. The rest of this paper is organized as follows. In the next section we will introduce the Σ-automorphism group of Γ and two subgroups (see Definitions 2.1 and 2.3 in the next section) of Aut(Γ×Σ), and study connections between these groups and the stability of (Γ, Σ). The proof of Theorem 1.3 will be given in Section 3. In Section 4, we will prove a number of lemmas concerning pairs of graphs with coprime valencies. These lemmas will be used to prove Theorem 1.8 in Section 5. We will conclude the paper with some remarks and questions in Section 6. 2. Σ-automorphism group of Γ Throughout this section Γ and Σ are graphs with V (Σ) = {1, . . . , n} and n > 1. 2.1. Aut (Γ), P (Γ, Σ) and Q(Γ, Σ). It is not difficult to see that the set of all Σ-automorphisms of Γ with multiplication defined by (α , . . . , α )(β , . . . , β ) = (α β , . . . , α β ) 1 n 1 n 1 1 n n is a group. We call this group the Σ-automorphism group of Γ and denote it by Aut (Γ). Note that (α, α, . . . , α) ∈ Aut (Γ) if and only if α ∈ Aut(Γ). Hence Σ Σ Aut(Γ) . Aut (Γ). (2) Moreover, Aut(Γ) = Aut (Γ) ⇔ α = · · · = α for each (α , . . . , α ) ∈ Aut (Γ). (3) Σ 1 n 1 n Σ STABILITY OF GRAPH PAIRS 5 Definition 2.1. Define Q(Γ, Σ) = {σ ∈ Aut(Γ × Σ) : (V (Γ) × {i}) = V (Γ) × {i} for each i ∈ V (Σ)}. Note that Q(Γ, Σ) is a subgroup of Aut(Γ × Σ). Denote the projections from V (Γ × Σ) to V (Γ) and V (Σ) by π and π , respectively. In other words, Γ Σ π π Γ Σ (u, i) = u and (u, i) = i, for (u, i) ∈ V (Γ × Σ). Lemma 2.2. Let Γ and Σ be graphs. Then Q(Γ, Σ) Aut (Γ). Proof. Set V (Σ) = {1, . . . , n}. Define f : Q(Γ, Σ) → Sym(V (Γ)) × · · · × Sym(V (Γ)), σ 7→ (α , . . . , α ) 1 n σπ π α Γ Γ i such that (u, i) = (u, i) for each (u, i) ∈ V (Γ × Σ). For any σ, τ ∈ Q(Γ, Σ), let f(σ) = (α , . . . , α ) and f(τ) = (β , . . . , β ). It is straightforward to verify that 1 n 1 n στπ π (α β ) Γ Γ i i (u, i) = (u, i) for each (u, i) ∈ V (Γ × Σ). Hence f(σ)f(τ) = (α , . . . , α )(β , . . . , β ) 1 n 1 n = (α β , . . . , α β ) 1 1 n n = f(στ). Thus f is a group homomorphism. Since f is injective, we have Q(Γ, Σ) = f(Q(Γ, Σ)). σ α Let (α , . . . , α ) ∈ Aut (Γ). Define σ by (u, i) = (u , i) for all i ∈ V (Σ). Then 1 n Σ σ ∈ Q(Γ, Σ) and f(σ) = (α , . . . , α ). Thus Aut (Γ) ⊆ f(Q(Γ, Σ)). 1 n Σ Conversely, let σ ∈ Q(Γ, Σ) and f(σ) = (α , . . . , α ). Then for any i, j ∈ V (Σ) 1 n with {i, j} ∈ E(Σ), we have {u, v} ∈ E(Γ) ⇔ {(u, i), (v, j)} ∈ E(Γ × Σ) σ σ α α i j ⇔ {(u, i) , (v, j) } = {(u , i), (v , j)} ∈ E(Γ × Σ) α α i j ⇔ {u , v } ∈ E(Γ), α α i j whence {u, v} ∈ E(Γ) if and only if {u , v } ∈ E(Γ). Thus f(σ) = (α , . . . , α ) ∈ 1 n Aut (Γ), and so f(Q(Γ, Σ)) ⊆ Aut (Γ). Therefore, Q(Γ, Σ) f(Q(Γ, Σ)) = Σ Σ Aut (Γ), completing the proof. Obviously, {V (Γ) × {i} | i ∈ V (Σ)} is a partition of V (Γ × Σ). Definition 2.3. Define P (Γ, Σ) to be the set of elements of Aut(Γ × Σ) that leave the partition {V (Γ) × {i} | i ∈ V (Σ)} invariant. That is, P (Γ, Σ) is the setwise stabilizer of {V (Γ) × {i} | i ∈ V (Σ)} under Aut(Γ × Σ). Hence it is a subgroup of Aut(Γ × Σ). Of course P (Γ, Σ) induces an action on {V (Γ) × {i} | i ∈ V (Σ)}, and the kernel of this action is exactly Q(Γ, Σ). Hence P (Γ, Σ)/Q(Γ, Σ) = Aut(Σ). This together with Lemma 2.2 implies the following result. 6 QIN, XIA, ZHOU, AND ZHOU Lemma 2.4. Let Γ and Σ be graphs. Then P (Γ, Σ) Aut (Γ) ⋊ Aut(Σ). 2.2. P (Γ, Σ) and the stability of (Γ, Σ). Lemma 2.5. Let Γ and Σ be graphs. If (Γ, Σ) is stable, then Aut(Γ × Σ) = P (Γ, Σ). Proof. By (2) and Lemma 2.4, we have Aut(Γ) × Aut(Σ) . Aut (Γ) ⋊ Aut(Σ) P (Γ, Σ) 6 Aut(Γ × Σ). Thus, if (Γ, Σ) is stable, then Aut(Γ × Σ) = P (Γ, Σ). Lemma 2.6. Let Γ and Σ be graphs, where V (Σ) = {1, . . . , n}. (a) If at least one Σ-automorphism of Γ is nondiagonal, then (Γ, Σ) is unstable. (b) If Aut(Γ × Σ) = P (Γ, Σ), then (Γ, Σ) is unstable if and only if at least one Σ-automorphism of Γ is nondiagonal. Proof. (a) Suppose that there is at least one nondiagonal Σ-automorphism of Γ. Then we see from (2) and (3) that Aut(Γ) is isomorphic to a proper subgroup of Aut (Γ). Combining this with Lemma 2.4, we obtain |Aut(Γ) × Aut(Σ)| < |Aut (Γ) ⋊ Aut(Σ)| = |P (Γ, Σ)| 6 |Aut(Γ × Σ)|. Thus Aut(Γ × Σ) ≇ Aut(Γ) × Aut(Σ) and so (Γ, Σ) is unstable. (b) Suppose that Aut(Γ×Σ) = P (Γ, Σ). Then Aut(Γ×Σ) Aut (Γ)⋊Aut(Σ) by Lemma 2.4. Thus (Γ, Σ) is unstable if and only if Aut (Γ) ≇ Aut(Γ), which, by (2) and (3), is true if and only if at least one Σ-automorphism of Γ is nondiagonal. 3. Proof of Theorem 1.3 3.1. Preparation. We need a few lemmas before we can prove Theorem 1.3. First, the following known results will be used in our proofs of Theorems 1.3 and 1.8. Lemma 3.1. Let Γ and Σ be graphs. (a) If Γ and Σ are connected with order at least 2, then Γ×Σ is connected if at least one of Γ and Σ is non-bipartite, and Γ × Σ has exactly two components if both Γ and Σ are bipartite ([13]; see also [4, Theorem 5.9]). (b) Γ × Σ is R-thin if and only if both Γ and Σ are R-thin ([9, Lemma 2.3]). (c) Γ × Σ is non-bipartite if and only if both Γ and Σ are non-bipartite ([4, Exercise 8.13]). Lemma 3.2. Let Γ and Σ be graphs. If (Γ, Σ) is stable, then both Γ and Σ are R-thin. STABILITY OF GRAPH PAIRS 7 Proof. We prove the contrapositive of the statement. Suppose that at least one of Γ and Σ is R-thick. Without loss of generality we may assume that Γ is R-thick. Then there exist two distinct vertices u and v of Γ such that N (u) = N (v). Let α be Γ Γ the permutation of V (Γ) which swaps u and v and fixes each vertex in V (Γ)\{u, v}, and let τ = (α, 1, 1, . . ., 1). Then τ is a Σ-automorphism of Γ. Moreover, τ is nondiagonal as α 6= 1. Thus, by Lemma 2.6(a), (Γ, Σ) is unstable. Lemma 3.3. Let Γ and Σ be graphs. If one of them is disconnected and the other has a nontrivial automorphism group, then (Γ, Σ) is unstable. Proof. Without loss of generality we may assume that Γ is disconnected and Aut(Σ) 6= 1. Take a connected component Γ of Γ and an element α 6= 1 of Aut(Σ). Let σ be the permutation of V (Γ × Σ) which fixes each vertex in V (Γ × Σ) \ V (Γ × Σ) and permutes the vertices in V (Γ × Σ) in the following way: σ α (u, i) = (u, i ) for each (u, i) ∈ V (Γ × Σ). It is straightforward to verify that σ ∈ Aut(Γ × Σ) \ P (Γ, Σ). So (Γ, Σ) is unstable by Lemma 2.5. In regard to Lemma 3.3, there exist both stable pairs (Γ, Σ) and unstable pairs (Γ, Σ) with Γ disconnected and Aut(Σ) = 1, as illustrated by the following example. Example 3.4. Let Γ be the graph with V (Γ ) = {1, 2, 3, 4, 5} and 1 1 E(Γ ) = {{1, 2}, {3, 4}, {4, 5}, {3, 5}}. Let Γ be the graph with V (Γ ) = {1, 2, 3, 4, 5, 6, 7, 8} and 2 2 E(Γ ) = {{1, 2}, {2, 3}, {3, 4}, {4, 5}, {3, 6}, {4, 6}, {7, 8}}. Let Σ be the graph with V (Σ) = {1, 2, 3, 4, 5, 6} and E(Σ) = {{1, 2}, {2, 3}, {3, 4}, {4, 5}, {3, 6}, {4, 6}}. Then both Γ and Γ are disconnected and Aut(Σ) = 1. Computation in Magma [1] 1 2 shows that |Aut(Γ × Σ)| = |Aut(Γ )| = 12, |Aut(Γ )| = 2, and |Aut(Γ × Σ)| = 4. 1 1 2 2 Hence (Γ , Σ) is stable but (Γ , Σ) is unstable. 1 2 As usual, for a graph Γ and a subset U ⊆ V (Γ), we use hUi to denote the subgraph of Γ induced by U, namely the graph with vertex set U in which two vertices u, v ∈ U are adjacent if and only if they are adjacent in Γ. Lemma 3.5. Let Γ and Σ be connected bipartite graphs. If both Aut(Γ) and Aut(Σ) are nontrivial groups, then (Γ, Σ) is unstable. Proof. Let {B , B } be the bipartition of Γ and {C , C } the bipartition of Σ. Set 1 2 1 2 U = (B × C ) ∪ (B × C ) and U = (B × C ) ∪ (B × C ). 1 1 1 2 2 2 1 2 2 1 8 QIN, XIA, ZHOU, AND ZHOU Then hU i and hU i are two connected components of Γ×Σ = hU ∪U i. Since both 1 2 1 2 Aut(Γ) and Aut(Σ) are nontrivial, we have |Aut(Γ × Σ)| > |Aut(Γ)| · |Aut(Σ)| > 2 · 2 = 4. (4) If both Aut(hU i) and Aut(hU i) are trivial, then 1 2 1, if hU i ≇ hU i 1 2 Aut(Γ × Σ) = Aut(hU ∪ U i) = 1 2 Z , if hU i hU i, 2 1 2 which contradicts (4). Hence at least one of Aut(hU i) and Aut(hU i) is nontriv- 1 2 ial. Without loss generality we may assume that Aut(hU i) is nontrivial. Take an element 1 6= α ∈ Aut(hU i). Let σ be the permutation of V (Γ × Σ) defined by (u, i), if (u, i) ∈ U (u, i) = (5) (u, i) , if (u, i) ∈ U for (u, i) ∈ V (Γ × Σ). Then σ ∈ Aut(Γ × Σ). Suppose to the contrary that (Γ, Σ) is stable. Then by Lemma 2.5, Aut(Γ × Σ) = P (Γ, Σ), and so σ ∈ P (Γ, Σ). Set V (Σ) = {1, . . . , n} and let α , . . . , α be 1 n α σπ i Γ permutations of V (Γ) such that u = (u, i) for u ∈ V (Γ) and i ∈ V (Σ). Since σ ∈ P (Γ, Σ), by (5) we have for (u, i) ∈ V (Γ × Σ), (u, i), if (u, i) ∈ U (u, i) = (6) (u , i), if (u, i) ∈ U . Hence σ σ {u, v} ∈ E(Γ) ⇔ {(u, i) , (v, j) } ∈ E(Γ × Σ) for any i, j ∈ V (Σ) with {i, j} ∈ E(Σ) α α i j ⇔ {(u , i), (v , j)} ∈ E(Γ × Σ) for any i, j ∈ V (Σ) with {i, j} ∈ E(Σ) α α i j ⇔ {u , v } ∈ E(Γ) for any i, j ∈ V (Σ) with {i, j} ∈ E(Σ). In other words, (α , . . . , α ) is a Σ-automorphism of Γ. Since (Γ, Σ) is stable, by 1 n Lemma 2.6(a) we deduce that α = · · · = α . This together with (6) implies that σ 1 n fixes each vertex of Γ × Σ, whence σ = 1. It then follows from (5) that α = 1. This contradiction shows that (Γ, Σ) is unstable and the proof is complete. The following example shows that the condition that both Aut(Γ) and Aut(Σ) are nontrivial cannot be removed from Lemma 3.5 for otherwise the result may not be true. Example 3.6. Let Γ be the graph with V (Γ) = {1, 2, 3, 4, 5, 6, 7} and E(Γ) = {{1, 2}, {2, 3}, {3, 4}, {4, 5}, {2, 6}, {3, 7}, {6, 7}}. Let Σ = K with V (Σ) = {a, b}. Then both Γ and Σ are bipartite and Aut(Γ) = 1. Let σ be the permutation of V (Γ × Σ) which interchanges (i, a) and (i, b) for each i ∈ V (Γ). Then Aut(Γ × Σ) = ∼ ∼ hσi Z . Hence Aut(Γ × Σ) Aut(Γ) × Aut(Σ) and (Γ, Σ) is stable. = = 2 STABILITY OF GRAPH PAIRS 9 Let Σ and Γ be graphs each with a nontrivial automorphism group. If one of them is R-thick or disconnected, then by Lemma 3.2 or 3.3 we know that (Γ, Σ) is unstable. Part (b) of the following lemma determines the stability of (Γ, Σ) when both Σ and Γ are connected, R-thin and non-bipartite. Lemma 3.7. Let Γ and Σ be graphs. (a) If (Γ, Σ) is stable, then Γ and Σ are coprime. (b) If both Γ and Σ are connected, R-thin and non-bipartite, then (Γ, Σ) is stable if and only if Γ and Σ are coprime. Proof. (a) We prove the contrapositive of this statement. Suppose that Γ and Σ are not coprime. Then there exist graphs Γ , Σ and Δ such that Γ = Γ × Δ, 1 1 1 Σ = Σ × Δ and |V (Δ)| > 1. So Γ × Σ = Γ × Δ × Σ × Δ. Let σ be the 1 1 1 permutation of V (Γ× Σ) = V (Γ × Δ× Σ × Δ) defined by (x, u, y, v) = (x, v, y, u) 1 1 for (x, u, y, v) ∈ V (Γ × Δ × Σ × Δ). Since |V (Δ)| > 1, it is straightforward to 1 1 verify that σ ∈ Aut(Γ × Σ) \ P (Γ, Σ). This together with Lemma 2.5 implies that (Γ, Σ) is unstable. (b) The “only if” part follows from (a), so it remains to prove the “if” part. Since both Γ and Σ are connected, R-thin and non-bipartite, by Lemma 3.1, so is Γ × Σ. If Γ and Σ are coprime, then by [4, Theorem 8.18], we have Aut(Γ × Σ) = Aut(Γ) × Aut(Σ) and hence (Γ, Σ) is stable. 3.2. Proof of Theorem 1.3. We are now ready to prove Theorem 1.3. Proof. Suppose that (Γ, Σ) is stable. Then by Lemmas 3.2 and 3.7(a) we know that Γ and Σ are coprime R-thin graphs. Moreover, if both Aut(Γ) and Aut(Σ) are nontrivial, then by Lemmas 3.3 and 3.5 both Γ and Σ are connected and at least one of them is non-bipartite. 4. Pairs of regular graphs with coprime valencies As a preparation for our proof of Theorem 1.8, we study pairs of regular graphs with coprime valencies in this section. As before, throughout this section Γ and Σ are graphs with V (Σ) = {1, . . . , n} and n > 1. As usual, the set of positive integers is denoted by N. A key concept used in this section is the Boolean square [4] of a graph Δ, denoted by B(Δ), which is the graph with vertex set V (Δ) and edge set {{u, v} | u, v ∈ V (Δ), u 6= v, N (u) ∩ N (v) 6= ∅}. Δ Δ 4.1. Notation. For (u, i), (v, j) ∈ V (Γ × Σ), define |N ((u, i)) ∩ N ((v, j))| Γ×Σ Γ×Σ f ((u, i), (v, j)) = . Γ,Σ |N ((u, i))| Γ×Σ Define X (u, i) = {(v, j) ∈ V (Γ × Σ) \ {(u, i)} | val(Σ) · f ((u, i), (v, j)) ∈ N} Γ,Σ Γ,Σ 10 QIN, XIA, ZHOU, AND ZHOU and Y (u, i) = {(v, j) ∈ X (u, i) | f ((u, i), (v, j)) > f ((u, i), (w, j)) ∀w ∈ V (Γ)}. Γ,Σ Γ,Σ Γ,Σ Γ,Σ We will abbreviate f ((u, i), (v, j)), X (u, i) and Y (u, i) to f((u, i), (v, j)), Γ,Σ Γ,Σ Γ,Σ X(u, i) and Y (u, i), respectively, when there is no danger of confusion. The following lemma follows immediately from the definitions of Γ × Σ and f((u, i), (v, j)). Lemma 4.1. Let Γ and Σ be regular graphs, and let (u, i), (v, j) ∈ V (Γ × Σ). Then |N (u) ∩ N (v)| |N (i) ∩ N (j)| Γ Γ Σ Σ f((u, i), (v, j)) = · . (7) val(Γ) val(Σ) Now assume that Σ is vertex-transitive. Then the vertex-transitivity of Σ implies that there exist integers n > n > · · · > n ≥ 1 such that for any i ∈ V (Σ) and 1 2 t x ∈ N (i) we have B(Σ) |N (x) ∩ N (i)| ∈ {n , n , . . . , n } and |N (i)| > n . Σ Σ 1 2 t B(Σ) 1 Set D (i) = {i}, D (i) = {x ∈ N (i) | |N (x) ∩ N (i)| = n , 1 6 n < |N (i)|} for 1 ≤ k ≤ t, k B(Σ) Σ Σ k k B(Σ) D (i) = {x ∈ V (Σ) | N (i) ∩ N (x) = ∅}, t+1 Σ Σ and D (i) = ∅ for s > t + 2. Then t+1 t V (Σ) = ∪ D (i), val(Σ) = |D (i)| and N (i) = ∪ D (i). k k B(Σ) k k=1 k=1 k=1 For each (u, i) ∈ V (Γ × Σ), define X (u, i) = Y (u, i) = {(u, i)} 0 0 and X (u, i) = X(u, i) ∩ (V (Γ) × D (i)) for k ≥ 1. k k k−1 Define Y (u, i) to be the set of elements (v, j) of X(u, i) \ (∪ X (u, i)) such that k m m=0 k−1 f((u, i), (v, j)) > f((u, i), (w, y)) for all (w, y) ∈ X(u, i) \ (∪ X (u, i)). m=0 STABILITY OF GRAPH PAIRS 11 4.2. X(u, i) and X (u, i). Lemma 4.2. Let Γ and Σ be graphs. Then for any (u, i) ∈ V (Γ × Σ) and σ ∈ Aut(Γ × Σ), we have σ σ (X(u, i)) = X((u, i) ). Proof. For any (u, i), (v, j) ∈ V (Γ × Σ) and σ ∈ Aut(Γ × Σ), we have σ σ (u, i) 6= (v, j) ⇔ (u, i) 6= (v, j) and σ σ val(Σ) · f((u, i), (v, j)) ∈ N ⇔ val(Σ) · f((u, i) , (v, j) ) ∈ N. It follows that σ σ σ σ (v, j) ∈ (X(u, i)) ⇔ (v, j) ∈ X(u, i) ⇔ (v, j) ∈ X((u, i) ). σ σ Thus (X(u, i)) = X ((u, i) ) as required. Lemma 4.3. Let Γ and Σ be regular graphs with coprime valencies. Suppose that Γ is R-thin. Then for any (u, i) ∈ V (Γ × Σ) and (v, j) ∈ X(u, i), we have i 6= j. Proof. Suppose that i = j for some (u, i) ∈ V (Γ × Σ) and (v, j) ∈ X(u, i). Then |N (i) ∩ N (j)| = val(Σ). Since (v, j) ∈ X(u, i), it follows from Lemma 4.1 that Σ Σ |N (u) ∩ N (v)| Γ Γ · val(Σ) = val(Σ) · f((u, i), (v, j)) ∈ N. val(Γ) This together with gcd(val(Γ), val(Σ)) = 1 implies that |N (u) ∩ N (v)| = val(Γ). Γ Γ Since Γ is R-thin, we deduce that u = v, and so (u, i) = (v, j), a contradiction. Lemma 4.4. Let Γ and Σ be regular graphs with coprime valencies. Suppose that Σ is R-thin. Then for any u ∈ V (Γ), any σ ∈ Aut(Γ × Σ), and any edge {i, j} of B(Σ), we have σπ σπ Σ Σ (u, i) 6= (u, j) . Proof. Since {i, j} is an edge of B(Σ), we have i 6= j, and hence (u, i) 6= (u, j). σ σ Since σ ∈ Aut(Γ × Σ), we then have (u, i) 6= (u, j) . Suppose to the contrary that σπ σπ σ σ Σ Σ (u, i) = (u, j) = z for some z ∈ V (Σ). Then (u, i) = (g, z) and (u, j) = (h, z) for some g, h ∈ V (Γ) with g 6= h. Thus σ σ |N ((u, i) ) ∩ N ((u, j) )| = |N ((g, z)) ∩ N ((h, z))| Γ×Σ Γ×Σ Γ×Σ Γ×Σ = |N (g) ∩ N (h)| · |N (z)| Γ Γ Σ = |N (g) ∩ N (h)| · val(Σ). Γ Γ On the other hand, since σ ∈ Aut(Γ × Σ), we have σ σ |N ((u, i) ) ∩ N ((u, j) )| = |N ((u, i)) ∩ N ((u, j))| Γ×Σ Γ×Σ Γ×Σ Γ×Σ = |N (u)| · |N (i) ∩ N (j)| Γ Σ Σ = val(Γ) · |N (i) ∩ N (j)|. Σ Σ 12 QIN, XIA, ZHOU, AND ZHOU Hence |N (g) ∩ N (h)| · val(Σ) = val(Γ) · |N (i) ∩ N (j)|. Γ Γ Σ Σ Since gcd(val(Γ), val(Σ)) = 1, it follows that val(Σ) divides |N (i) ∩ N (j)|. Hence Σ Σ |N (i) ∩ N (j)| = val(Σ) or |N (i) ∩ N (j)| = 0. However, this is impossible as Σ Σ Σ Σ Σ is R-thin and {i, j} is an edge of B(Σ). This completes the proof. Lemma 4.5. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Then for any edge {u, v} of B(Γ), any edge {i, j} of B(Σ), and any element σ ∈ Aut(Γ × Σ), we have σπ σπ Σ Σ |N ((u, i) ) ∩ N ((v, j) )| > |N (i) ∩ N (j)|. (8) Σ Σ Σ Σ Proof. Since {u, v} is an edge of B(Γ), we have N (u) ∩ N (v) 6= ∅. Consider Γ Γ w ∈ N (u) ∩ N (v). We have Γ Γ {w} × (N (i) ∩ N (j)) ⊆ N ((u, i)) ∩ N ((v, j)). Σ Σ Γ×Σ Γ×Σ Since σ ∈ Aut(Γ × Σ), it follows that σ σ ({w} × (N (i) ∩ N (j))) ⊆ (N ((u, i)) ∩ N ((v, j))) Σ Σ Γ×Σ Γ×Σ σ σ = N ((u, i) ) ∩ N ((v, j) ). Γ×Σ Γ×Σ This together with Lemma 4.4 implies (8). Lemma 4.6. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Suppose further that Σ is vertex-transitive. Then for any (u, i) ∈ V (Γ × Σ), any σ ∈ Aut(Γ × Σ), and any integer k ≥ 0, we have σ σ (X (u, i)) = X ((u, i) ). k k Proof. Consider an arbitrary vertex (v, j) of X (u, i). By Lemma 4.2, σ σ σ σ (v, j) ∈ (X (u, i)) ⊆ (X(u, i)) = X((u, i) ). (9) Since X (u, i) = X(u, i) ∩ (V (Γ) × D (i)), we have j ∈ D (i) and (v, j) ∈ X(u, i), k k k whence {u, v} is an edge of B(Γ) and {i, j} is an edge of B(Σ). It then follows from Lemma 4.5 that σπ σπ Σ Σ |N ((u, i) ) ∩ N ((v, j) )| > |N (i) ∩ N (j)| = n . (10) Σ Σ Σ Σ k σ σ σ First assume that k = 0. Then (X (u, i)) = {(u, i) } = X ((u, i) ). 0 0 σ σ Next assume that k = 1. By (9), we have (v, j) ∈ X((u, i) ). By Lemma 4.3, we then have σπ σπ Σ Σ (u, i) 6= (v, j) , and so σπ σπ Σ Σ |N ((u, i) ) ∩ N ((v, j) )| 6 n . Σ Σ 1 This together with (10) implies that σπ σπ Σ Σ |N ((u, i) ) ∩ N ((v, j) )| = n . Σ Σ 1 STABILITY OF GRAPH PAIRS 13 σ σπ Thus (v, j) ∈ V (Γ) × D ((u, i) ). Combining this with (9), we then have σ σ σπ σ (v, j) ∈ X((u, i) ) ∩ (V (Γ) × D ((u, i) )) = (X (u, i) ). 1 1 Therefore, σ σ (X (u, i)) ⊆ X ((u, i) ). (11) 1 1 It then follows that −1 −1 σ σ σσ (X ((u, i) )) ⊆ X ((u, i) ) = X (u, i). 1 1 1 Thus σ σ |X ((u, i) )| 6 |(X (u, i)) |. 1 1 σ σ This together with (11) implies (X (u, i)) = X ((u, i) ) as required. 1 1 Now assume that k > 2. Suppose by induction that for all integers ℓ with 0 6 ℓ < k we have σ σ (X (u, i)) = X ((u, i) ). (12) ℓ ℓ k−1 σ σ Since (v, j) ∈ X (u, i), we have (v, j) ∈/ (∪ X (u, i)) . It then follows from (12) k m m=0 that σ k−1 σ k−1 σ (v, j) ∈/ ∪ (X (u, i)) = ∪ X ((u, i) ). m m m=0 m=0 That is, σ σ (v, j) ∈/ X ((u, i) ) for m = 0, 1, . . . , k − 1. By the definition of X ((u, i) and (9), we then obtain that σπ σπ Σ Σ (v, j) ∈/ D ((u, i) ) for m = 0, 1, . . . , k − 1. σπ σπ Σ Σ This together with (10) implies that (v, j) ∈ D ((u, i) ), and hence we drive from (9) that σ σ σπ σ (v, j) ∈ X((u, i) ) ∩ (V (Γ) × D ((u, i) )) = X ((u, i) ). k k Therefore, σ σ (X (u, i)) ⊆ X ((u, i) ). k k Note that this holds for any (u, i) ∈ V (Γ × Σ) and σ ∈ Aut(Γ × Σ). Replacing (u, i) σ −1 by (u, i) and σ by σ in this inclusion, we obtain −1 −1 σ σ σ σ (X ((u, i) )) ⊆ X (((u, i) ) ) = X (u, i), k k k or equivalently, σ σ X ((u, i) ) ⊆ (X (u, i)) . k k σ σ Therefore, (X (u, i)) = X ((u, i) ) and the proof is complete by induction. k k 14 QIN, XIA, ZHOU, AND ZHOU 4.3. Y (u, i) and Y (u, i). Lemma 4.7. Let Γ and Σ be regular graphs with coprime valencies. Suppose that Σ is Γ-thin. Then for any (u, i) ∈ V (Γ × Σ), we have Y (u, i) = {u} × N (i). B(Σ) Proof. Let j ∈ N (i). Then i 6= j and N (i) ∩ N (j) 6= ∅. Since Γ and Σ are B(Σ) Σ Σ regular, we drive from Lemma 4.1 that val(Σ) · f((u, i), (u, j)) = |N (i) ∩ N (j)| ∈ N. (13) Σ Σ For any w ∈ V (Γ), by Lemma 4.1 and (13) we have |N (u) ∩ N (w)| Γ Γ val(Σ) · f((u, i), (w, j)) = · |N (i) ∩ N (j)| Σ Σ val(Γ) 6 |N (i) ∩ N (j)| Σ Σ = val(Σ) · f((u, i), (u, j)). Hence f((u, i), (u, j)) > f((u, i), (w, j)) for all w ∈ V (Γ). (14) This together with (13) implies that (u, j) ∈ Y (u, i). Thus {u} × N (i) ⊆ Y (u, i). (15) B(Σ) Conversely, let (v, j) ∈ Y (u, i). Then (v, j) 6= (u, i), val(Σ) · f((u, i), (v, j)) ∈ N, and f((u, i), (v, j)) > f((u, i), (w, j)) for all w ∈ V (Γ). (16) Moreover, by Lemma 4.1, we have |N (u) ∩ N (v)| Γ Γ val(Σ) · f((u, i), (v, j)) = · |N (i) ∩ N (j)| ∈ N. (17) Σ Σ val(Γ) It follows that |N (i) ∩ N (j)| 6= 0, and so N (i) ∩ N (j) 6= ∅. Since (v, j) ∈ Σ Σ Σ Σ Y (u, i) ⊆ X(u, i), we derive from Lemma 4.3 that j 6= i, and hence j ∈ N (i). B(Σ) Thus, using the same argument as in the first paragraph of this proof, we can derive that (u, i), (u, j) and (w, j) satisfy (14). Combining this and (16), we then obtain v = u and therefore (v, j) ∈ {u} × N (i). So we have proved that B(Σ) Y (u, i) ⊆ {u} × N (i), B(Σ) which together with (15) yields Y (u, i) = {u} × N (i), as required. B(Σ) Lemma 4.8. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Then the following hold for any (u, i) ∈ V (Γ × Σ) and positive integer k: (a) Y (u, i) ⊆ X (u, i); k k (b) Y (u, i) = Y (u, i) ∩ (V (Γ) × D (i)). k k STABILITY OF GRAPH PAIRS 15 Proof. If Y (u, i) = ∅, then statement (a) is obvious. Now assume that Y (u, i) 6= ∅ k k and let (v, j) ∈ Y (u, i). Then k−1 (v, j) ∈ X(u, i) \ (∪ X (u, i)) (18) m=0 and k−1 f((u, i), (v, j)) > f((u, i), (w, k)) for all (w, k) ∈ X(u, i) \ (∪ X (u, i)). (19) m=0 Suppose by way of contradiction that (v, j) ∈/ X (u, i). Then by (18) we have (v, j) ∈ X(u, i) but (v, j) ∈/ X (u, i) for m = 0, 1, . . . , k − 1, k. Since X (u, i) = X(u, i) ∩ (V (Γ) × D (i)), we have m m (v, j) ∈/ V (Γ) × D (i) for m = 0, 1, . . . , k − 1, k. So j ∈ D (i) for some t > k. Thus D (i) 6= ∅ and so D (i) 6= ∅. Take y ∈ D (i). t t k k Then |N (i) ∩ N (j)| < |N (i) ∩ N (y)| (20) Σ Σ Σ Σ and k−1 (u, y) ∈ X(u, i) \ (∪ X (u, i)). m=0 Using (20), we can easily obtain f((u, i), (v, j)) < f((u, i), (u, y)), which contradicts (19). This contradiction shows that (v, j) ∈ X (u, i), and therefore Y (u, i) ⊆ k k X (u, i). This completes the proof of statement (a). Now we prove statement (b). Let (v, j) ∈ Y (u, i). By statement (a), we have (v, j) ∈ X (u, i) = X(u, i) ∩ (V (Γ) × D (i)) ⊆ X(u, i). (21) k k It then follows from Lemmas 4.1 and 4.3 that |N (u) ∩ N (v)| Γ Γ val(Σ) · f((u, i), (v, j)) = · |N (i) ∩ N (j)| ∈ N Σ Σ val(Γ) and i 6= j. Hence N (i) ∩ N (j) 6= ∅. Using Lemma 4.1, it is straightforward to Σ Σ verify that val(Σ) · f((u, i), (u, j)) ∈ N (22) and val(Σ) · f((u, i), (u, j)) > val(Σ) · f((u, i), (v, j)). (23) k−1 Moreover, by (21), we have j ∈ D (i), and so j ∈/ ∪ D (i). Since i 6= j, we derive k m m=0 from (22) that k−1 (u, j) ∈ X(u, i) \ (∪ X (u, i)). m=0 This combined with (19) and (23) implies that v = u. Since N (i) ∩ N (j) 6= ∅, Σ Σ we have j ∈ N (i), and thus we obtain from Lemma 4.7 and (21) that (v, j) ∈ B(Σ) Y (u, i) ∩ V (Γ) × D (i). Therefore, Y (u, i) ⊆ Y (u, i) ∩ (V (Γ) × D (i)). (24) k k Conversely, let (v, j) ∈ Y (u, i) ∩ (V (Γ) × D (i)). Since Y (u, i) ⊆ X(u, i), we have k−1 (v, j) ∈ X(u, i) ∩ (V (Γ) × D (i)) ⊆ X(u, i) \ (∪ X (u, i)). (25) k m m=0 16 QIN, XIA, ZHOU, AND ZHOU k−1 Consider an arbitrary element (w, t) of X(u, i) \ (∪ X (u, i)). Since X (u, i) = m m m=0 X(u, i) ∩ (V (Γ) × D (i)), we have (w, t) ∈/ X(u, i) ∩ (V (Γ) × D (i)) for m = 0, 1, . . . , k − 1. Since (w, t) ∈ X(u, i), we deduce that t ∈/ D (i) for m = 0, 1, . . . , k − 1, k−1 and hence j ∈/ ∪ D (i). Note that (v, j) ∈ D (i). It follows that m k m=0 |N (i) ∩ N (j)| > |N (i) ∩ N (t)|. (26) Σ Σ Σ Σ Since (v, j) ∈ Y (u, i), by Lemma 4.7 we obtain that u = v. Thus |N (u) ∩ N (v)| = val(Γ) > |N (u) ∩ N (w)|. Γ Γ Γ Γ This together with (26) and Lemma 4.1 implies that val(Σ) · f((u, i), (v, j)) = |N (i) ∩ N (j)| Σ Σ |N (u) ∩ N (w)| Γ Γ > · |N (i) ∩ N (t)| Σ Σ val(Γ) = val(Σ) · f((u, i), (w, t)). Combining this with (25), we obtain (v, j) ∈ Y (u, i). Thus Y (u, i) ∩ (V (Γ) × D (i)) ⊆ Y (u, i). k k This together with (24) completes the proof of statement (b). Lemma 4.9. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Suppose further that Σ is vertex-transitive. Then for any (u, i) ∈ V (Γ × Σ) and σ ∈ Aut(Γ × Σ), we have σ σ (Y (u, i)) = Y ((u, i) ). t+1 Proof. Since Σ is finite, there exists a positive integer t such that V (Σ) = ∪ D (i), m=0 where as before D (i) = {i} and D (i) = {x ∈ V (Σ) | N (i) ∩ N (x) = ∅}. Since 0 t+1 Σ Σ Y (u, i) ⊆ X(u, i), by Lemma 4.3, we have Y (u, i) ∩ (V (Γ) × D (i)) = ∅. (27) For any (v, j) ∈ V (Γ) × D (i), we have j ∈/ N (i) as N (i) ∩ N (j) = ∅. By t+1 B(Σ) Σ Σ Lemma 4.7, we then have (v, j) ∈/ Y (u, i) and therefore Y (u, i) ∩ (V (Γ) × D (i)) = ∅. (28) t+1 Note that V (Γ × Σ) = V (Γ) × V (Σ) t+1 = V (Γ) × ∪ D (i) m=0 t+1 = ∪ (V (Γ) × D (i)) m=0 = ∪ (V (Γ) × D (i)) ∪ (V (Γ) × D (i)) ∪ (V (Γ) × D (i)) . m 0 t+1 m=1 STABILITY OF GRAPH PAIRS 17 Thus, by (27), (28) and Lemma 4.8(b), we have Y (u, i) = Y (u, i) ∩ V (Γ × Σ) = Y (u, i) ∩ ∪ (V (Γ) × D (i)) m=1 = ∪ (Y (u, i) ∩ (V (Γ) × D (i)) m=1 = ∪ Y (u, i). m=1 Since σ ∈ Aut(Γ × Σ), by Lemmas 4.2 and 4.6, we obtain that for each positive integer k, k−1 σ k−1 σ X(u, i) \ ∪ X (u, i) = (X(u, i)) \ ∪ (X (u, i)) m m m=0 m=0 σ k−1 σ = X((u, i) ) \ ∪ X ((u, i) ) . m=0 σ σ Thus (Y (u, i)) = Y ((u, i) ) by the definition of Y (u, i). Therefore, k k k σ t t σ t σ σ (Y (u, i)) = ∪ Y (u, i) = ∪ (Y (u, i)) = ∪ Y ((u, i) ) = Y ((u, i) ), m m m m=1 m=1 m=1 completing the proof. 5. Proof of Theorem 1.8 5.1. Lemmas. We need the following lemmas in our proof of Theorem 1.8. Lemma 5.1. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Suppose further that Σ is vertex-transitive. Let u ∈ V (Γ) and i, j ∈ V (Σ). If N (i) ∩ N (j) 6= ∅, then for any σ ∈ Aut(Γ × Σ), we have Σ Σ σπ σπ Γ Γ (u, i) = (u, j) . σ σ Proof. If i = j, then (u, i) = (u, j) and so the statement is true. Now assume that i 6= j and N (i) ∩ N (j) 6= ∅. Then j ∈ N (i). By Lemma 4.7, we then Σ Σ B(Σ) σ σ σ have (u, j) ∈ Y (u, i). Hence (u, j) ∈ (Y (u, i)) = Y ((u, i) ) by Lemma 4.9. This σπ σπ Γ Γ together with Lemma 4.7 implies that (u, i) = (u, j) , as required. Lemma 5.2. Let Γ and Σ be regular graphs with coprime valencies. Suppose that both Γ and Σ are R-thin. Suppose further that Σ is vertex-transitive. Let u, v ∈ V (Γ) and i ∈ V (Σ). If there is a walk from u to v of even length, then for any σ ∈ Aut(Γ × Σ), we have σπ σπ Σ Σ (u, i) = (v, i) . (29) Proof. It suffices to prove (29) in the case when N (u) ∩ N (v) 6= ∅. The result in Γ Γ the general case follows by applying this equation to pairs of every other vertices on the walk. So let us assume that N (u) ∩ N (v) 6= ∅. Take a vertex w ∈ N (u) ∩ N (v). Set Γ Γ Γ Γ W = {w} × N (i). Then W ⊆ N ((u, i)) ∩ N ((v, i)). Since σ ∈ Aut(Γ × Σ), Σ Γ×Σ Γ×Σ it follows that σ σ σ W ⊆ N ((u, i) ) ∩ N ((v, i) ). Γ×Σ Γ×Σ 18 QIN, XIA, ZHOU, AND ZHOU Consequently, σπ σπ σπ Σ Σ Σ W ⊆ N ((u, i) ) ∩ N ((v, i) ). (30) Σ Σ Note that σ σπ σπ Σ Σ |W | = |W| = |N (i)| = |N ((u, i) )| = |N ((v, i) )| = val(Σ). (31) Σ Σ Σ Consider any two distinct vertices (w, j), (w, k) in W . We have j 6= k ∈ N (i) and i ∈ N (j) ∩ N (k), whence N (j) ∩ N (k) 6= ∅. By Lemma 5.1, we have Σ Σ Σ Σ σπ σπ Γ Γ (w, j) = (w, k) . Since (w, j) 6= (w, k) and σ ∈ Aut(Γ × Σ), we then obtain σπ σπ σπ σ Σ Σ Σ that (w, j) 6= (w, k) . Hence |W | = |W |. Combining this with (30) and (31), we obtain that σπ σπ σπ Σ Σ Σ N ((u, i) ) = N ((v, i) ) = W . Σ Σ σπ σπ Σ Σ Since Σ is R-thin, we then conclude that (u, i) = (v, i) . For a graph Δ and a partition B of V (Δ), the quotient graph Δ of Δ with respect to B is defined to have vertex set B such that two blocks B, C ∈ B are adjacent if and only if there exists at least one edge of Δ with one end-vertex in B and the other end-vertex in C. Lemma 5.3. Let Γ and Σ be connected regular graphs with coprime valencies. Sup- pose that both Γ and Σ are R-thin. Suppose further that Γ is bipartite and Σ is vertex-transitive and non-bipartite. Then Aut(Γ × Σ) = P (Γ, Σ). Proof. Let B and B be the biparts of Γ, and let V (Σ) = {1, 2, . . ., n}. Let u, v ∈ 0 1 V (Γ) and i ∈ V (Σ), and let σ be an arbitrary element of Aut(Γ × Σ). Let Δ = Γ × Σ. Set B = {V (Γ) × {j} | j ∈ V (Σ)} and D = {B × {j}, B × {j} | j ∈ V (Σ)}. 0 1 Then B and D are partitions of V (Δ), and hence Δ and Δ are well defined. B D First assume that u and v are in the same biparts of Γ. Since Γ is connected and bipartite, there is a walk in Γ from u to v of even length. It then follows from σπ σπ Σ Σ Lemma 5.2 that (u, i) = (v, i) . Hence σ preserves D, and so σ ∈ Aut(Δ ). Next assume that u and v are in different biparts of Γ. Without loss generality we may assume that u ∈ B and v ∈ B . Since Σ is non-bipartite and vertex-transitive, 0 1 every vertex of Σ is contained in an odd cycle, and so there exists a cycle of odd length containing i, say, C : i , i , . . . , i , i , where i = i and ℓ is odd. It follows 1 2 ℓ 1 1 that C : V (Γ) × {i }, V (Γ) × {i }, . . . , V (Γ) × {i }, V (Γ) × {i } B 1 2 ℓ 1 STABILITY OF GRAPH PAIRS 19 and C : B ×{i }, B ×{i }, . . . , B ×{i }, B ×{i }, B ×{i }, . . . , B ×{i }, B ×{i } D 0 1 1 2 0 ℓ 1 1 0 2 1 ℓ 0 1 are cycles in Δ and Δ , respectively. Note that C is of length 2ℓ. Since σ ∈ B D D Aut(Δ ) as shown above, σ maps C to the cycle (C ) of Δ , and also maps the D D D D pair of antipodals B × {i}, B × {i} to some pair of antipodals in (C ) . Note also 0 1 D π π Σ Σ that for each pair of antipodals X, Y of some cycle in Δ , we have X = Y and π π Σ Σ |X | = |Y | = 1. In particular, we have σπ σπ Σ Σ (B × {i}) = (B × {i}) 0 1 and σπ σπ Σ Σ |(B × {i}) | = |(B × {i}) | = 1. 0 1 σπ σπ Σ Σ Since u ∈ B and v ∈ B , we then obtain that (u, i) = (v, i) . Since this 0 1 holds for any u, v ∈ V (Γ) and i ∈ V (Σ), we conclude that σ ∈ P (Γ, Σ). Since this holds for any σ ∈ Aut(Γ × Σ), we obtain that Aut(Γ × Σ) 6 P (Γ, Σ), which yields Aut(Γ × Σ) = P (Γ, Σ). Lemma 5.4. Let Γ and Σ be connected regular graphs with coprime valencies. Sup- pose that both Γ and Σ are R-thin. Suppose further that Γ is non-bipartite and Σ is vertex-transitive and bipartite. Then Aut(Γ × Σ) = P (Γ, Σ). Proof. Let Δ = Γ × Σ and B = {V (Γ) × {j} | j ∈ V (Σ)}. Then the quotient graph Δ of Δ with respect to B is isomorphic to Σ. Since Σ is bipartite, it follows that there is no odd cycle in Δ . This implies that, for any u, v ∈ V (Γ) and i ∈ V (Σ), there is no path of odd length from (u, i) to (v, i) in Δ. Since Γ is non-bipartite, we obtain from Lemma 3.1(a) that Δ is connected. It follows that there is a path of even length from (u, i) to (v, i) in Δ, and therefore there is a walk of even length from u to v in Γ. Thus, by Lemma 5.2, for any σπ σπ Σ Σ σ ∈ Aut(Γ × Σ), we have (u, i) = (v, i) , which implies σ ∈ P (Γ, Σ). Since this holds for any σ ∈ Aut(Γ × Σ), it follows that Aut(Γ × Σ) 6 P (Γ, Σ), yielding Aut(Γ × Σ) = P (Γ, Σ) as required. 5.2. Proof of Theorem 1.8. We are now ready to prove Theorem 1.8. Proof. Let Γ be a regular graph and Σ a vertex-transitive graph such that val(Γ) and val(Σ) are coprime. Suppose that (Γ, Σ) is nontrivially unstable. Then by the definition of a nontriv- ially unstable graph pair, Γ and Σ are coprime connected R-thin graphs and at least one of them is non-bipartite. Hence, by Lemma 3.1, Γ × Σ is connected and R-thin. Moreover, by Lemmas 5.3 and 5.4, we have Aut(Γ × Σ) = P (Γ, Σ). Since (Γ, Σ) is 20 QIN, XIA, ZHOU, AND ZHOU unstable, it follows from Lemma 2.6(b) that at least one Σ-automorphism of Γ is nondiagonal. Conversely, suppose that Γ × Σ is connected and R-thin and at least one Σ- automorphism of Γ is nondiagonal. Then, by Lemma 3.1, Γ and Σ are connected R-thin graphs and at least one of them is non-bipartite. If both Γ and Σ are non- bipartite, then by Lemma 3.7(b), (Γ, Σ) is stable, but this contradicts Lemma 2.6(a) as we assume that at least one Σ-automorphism of Γ is nondiagonal. Thus exactly one of Γ and Σ is non-bipartite. Since val(Γ) and val(Σ) are coprime, by Remark 1.2, Γ and Σ are coprime. Finally, by Lemmas 5.3 and 5.4, we have Aut(Γ×Σ) = P (Γ, Σ). Hence, by Lemma 2.6(b), (Γ, Σ) is unstable. Moreover, (Γ, Σ) is nontrivially unstable since Γ and Σ are coprime connected R-thin graphs. 6. Concluding remarks In the case when Σ = K , Theorem 1.8 gives rise to the following result: A con- nected regular graph is unstable if and only if it has a nontrivial two-fold automor- phism. As mentioned in the introduction, this is a special case of [5, Theorem 3.2], where the graph in the statement is not required to be regular. So Theorem 1.8 is a partial generalization of [5, Theorem 3.2]. It would be interesting to study whether the result in Theorem 1.8 is still true if Σ is not required to be regular with valency coprime to the valency of Γ. As seen in Theorem 1.8, there are close connections between stability of graph pairs (Γ, Σ) and Σ-automorphisms of Γ. However, the notion of Σ-automorphisms of Γ deserves further studies for its own sake. As far as we know, there are not many results on Σ-automorphisms of Γ in the literature, except in the case when ◦ ◦ Σ = K for which a K -automorphism of Γ is an automorphism of Γ in the usual 1 1 sense, and in the case when Σ = K for which a K -automorphism of Γ is exactly 2 2 a two-fold automorphism of Γ [5]. The concept of two-fold automorphisms was first introduced by Zelinka in [15, 16] for digraphs in his study of isotopies of digraphs and was extended to mixed graphs by Lauri et al. in [5]. It is readily seen that, if Σ is a spanning subgraph of Σ , then Aut (Γ) ≤ Aut (Γ). So among all graphs Σ 2 Σ Σ 2 1 of order n the complete graph K gives rise to the smallest possible group Aut (Γ). n Σ Therefore, it would be interesting to study the groups Aut (Γ) for n ≥ 2. Note TF that Aut (Γ) is exactly the two-fold automorphism group Aut (Γ) of Γ [5]. Finally, many questions about (nontrivially) unstable graph pairs may be asked. For example, by Lemma 2.6(a), if Γ admits a nondiagonal Σ-automorphism then (Γ, Σ) is unstable, and Theorem 1.8 determines a situation where this necessary condition is also sufficient. In general, one may ask the following question: For an unstable pair of graphs (Γ, Σ), under what conditions does Γ admit a nondiagonal Σ-automorphism? One may also study the stability of (Γ, Σ) for various special families of graphs Γ and/or various special families of graphs Σ. Possible candidates for Γ include circulant graphs [9, 14], arc-transitive graphs, generalized Petersen STABILITY OF GRAPH PAIRS 21 graphs GP(n, k) [10], etc. and potential choices for Σ include complete graphs K , complete bipartite graphs K , cycles C , etc. Obviously, this area of research is n,n n wide open and pleasantly inviting. Acknowledgements. We are grateful to the anonymous referees for their helpful comments and Dr Jiyong Chen for his valuable advices. Part of the work was done during a visit of the first author to The University of Melbourne. The first author would like to thank The University of Melbourne for its hospitality during her visit and Beijing Jiaotong University for its financial support. She also thanks the National Natural Science Foundation of China (11671030) for its financial support during her PhD. The first author was supported by the Fundamental Research Funds for Beijing Universities allocated to Capital University of Economics and Business(XRZ2020058). The third author was supported by the National Natural Science Foundation of China (11671030,12071023). References [1] W. Bosma, J. Cannon and C. Playoust, The Magma Algebra System I: The User Language, J. Symbolic Comput. 24 (1997), 235–265. [2] J. Chen, Embeddings of direct product graphs, in preparation, personal communication. [3] W. D¨orfler, Primfaktorzerlegung und Automorphismen des Kardinalproduktes von Graphen, Glasnik Mat. Ser. III 9(29) (1974), 15–27. [4] R. Hammack, W. Imrich and S. Klavˇzar, Handbook of Product Graphs, 2nd ed., CRC Press [5] J. Lauri, R. Mizzi and R. Scapellato, Unstable graphs: a fresh outlook via TF-automorphisms, Ars Math. Contemp. 8 (2015), 115–131. [6] D. Maruˇsiˇc, R. Scapellato and N. Zagaglia Salvi, A characterization of particular symmetric (0, 1) matrices, Linear Algebra Appl. 119 (1989), 153–162. [7] D. Maruˇsiˇc, R. Scapellato, and N. Zagaglia Salvi, Generalized Cayley graphs, Discrete Math. 102 (1992), 279–285. [8] R. Nedela and M. Skoviera, Regular embeddings of canonical double coverings of graphs, J. Combin. Theory Ser. B 67 (1996), 249–277. [9] Y-L. Qin, B. Xia and S. Zhou, Stability of circulant graphs, J. Combin. Theory Ser. B, 136 (2019), 154–169. [10] Y-L. Qin, B. Xia and S. Zhou, Stability of generalized Petersen graphs, J. Graph Theory, accepted, DOI: 10.1002/jgt.22642. [11] D. Surowski, Stability of arc-transitive graphs, J. Graph Theory 38 (2001), 95–110. [12] D. Surowski, Automorphism groups of certain unstable graphs, Math. Slovaca 53 (2003), 215–232. [13] P. M. Weichsel, The Kronecker product of graphs, Proc. Amer. Math. Soc. 13 (1962), 47–52. [14] S. Wilson, Unexpected symmetries in unstable graphs, J. Combin. Theory Ser. B 98 (2008), 359–383. [15] B. Zelinka, The group of autotopies of a digraph, Czech Math. J. 21 (1971), 619–624. [16] B. Zelinka, Isotopy of digraphs, Czech Math. J. 22 (1972), 353–360. 22 QIN, XIA, ZHOU, AND ZHOU School of Statistics, Capital University of Economics and Business, Beijing, 100070, P. R. China Email address: ylqin@cueb.edu.cn School of Mathematics and Statistics, The University of Melbourne, Parkville, VIC 3010, Australia Email address: binzhoux@unimelb.edu.au Department of Mathematics, Beijing Jiaotong University, Beijing, 100044, P. R. China Email address: jxzhou@bjtu.edu.cn School of Mathematics and Statistics, The University of Melbourne, Parkville, VIC 3010, Australia Email address: sanming@unimelb.edu.au

Journal

MathematicsarXiv (Cornell University)

Published: Oct 30, 2020

References