The automorphism groups of some token graphs
Sofía Ibarra
Luis Manuel Rivera
Abstract
The token graphs of graphs have been studied at least from the 80’s with different names and by different authors. The Johnson graph J(n,k) is isomorphic to the k-token graph of the complete graph Kn. To our knowledge, the unique results about the automorphism groups of token graphs are for the case of the Johnson graphs. In this paper we begin the study of the automorphism groups of token graphs of another graphs. In particular we obtain the automorphism group of the k-token graph of the path graph Pn, for n=2k. Also, we obtain the automorphism group of the 2-token graph of the following graphs: cycle, star, fan and wheel graphs.
Keywords: Token graphs; automorphism groups, Johnson graphs.
AMS Subject Classification Numbers: 05C76, 05C60.
1 Introduction
Let Γ be a simple graph of order n. Let 1≤k≤n−1 be an integer. The k-token graph Fk(Γ) of Γ is defined as the graph with vertex set all k-subsets of V(Γ), where two vertices are adjacent in Fk(Γ) whenever their symmetric difference is an edge of Γ. If k∈{1,n−1}, then Fk(Γ) is isomorphic to Γ and in this case we say that Fk(Γ) is a trivial token graph of Γ. In fact, if Γ is a graph of order n, then Fk(Γ)≃Fn−k(Γ).
The token graphs have been redefined several times and with different names. When k=2, this class of graphs are called double vertex graphs that were widely studied by Alavi et al. [1, 2, 3, 4] and are the same that the 2-subgraph graphs defined in a thesis of G. Johns [19]. In the work of Zhu et al. [31], the k-token graphs are named n-tuple vertex graphs.
Later, T. Rudolph [29] redefined the double vertex graphs with the name of symmetric powers of graphs with the idea to study the graph isomorphism problem and some problems in quantum mechanics. There are several papers related with Rudolph’s work, see, e. g., [6, 7, 8, 15] and the references therein. Some of them motivated by the connection between token graphs and the Heisenberg Hamiltonian (see, e. g., [26] and the references therein), that is related with the Heisenberg model [17], a quantum theory of ferromagnetism.
Finally, R. Fabila-Monroy, et. al. [14], in an independent way, reintroduce this concept but now with the name of token graphs and began a systematic study of several combinatorial properties of this graphs: connectivity, diameter, cliques, chromatic number and Hamiltonian paths. In the last years, several groups of authors have continued with this line of research (see, e.g., [5, 11, 12, 16, 18, 22, 28]). For example, Carballosa et al. [10] studied the planarity and regularity of token graphs and Leaños and Trujillo-Negrete [22] proved a conjecture of Fabila-Monroy, et. al [14] about the connectivity of token graphs. Finally, Gómez Soto et al. [16] found the packing number of the 2-token graph of the path graph, that is equal to the size of largest binary code of length n and constant weight 2 that can correct a single adjacent transposition (sequence A085680 in [30]).
When Γ is the complete graph Kn, the k-token graph Fk(Γ) is isomorphic to the Johnson graph J(n,k) [20]. To the knowledge of the authors, the only results about the automorphism groups of token graphs are about Johnson graphs. It is known that if n=2k, then Aut(Fk(Kn))≃Sn and if n=2k, then Aut(Fk(Kn))≃S2×Sn, where Sn denotes the symmetric group on n symbols (see., e.g. [21, 25, 27]).
In this work, we study the automorphism group of other token graphs. Our main results can be stated as two theorems:
Theorem 1.1**.**
Let n=4 be an integer greater than 2. If Γ∈{Cn,K1,n−1,A1,n−1,W1,n−1},
then
[TABLE]
where Cn,K1,n−1,A1,n−1 and W1,n−1, denotes the cycle, star, fan and wheel graphs, respectively.
Theorem 1.2**.**
Let Pn be the path graph of order n>2, with n=2k. Then
[TABLE]
Theorem 1.1 is not true in general. For example, the automorphism group of the grid graph G2,3 is of order 4 but ∣Aut(F2(G2,3))∣=8.
In the proofs of our results, we use elementary group theory, as in [23, 24, 25], and properties of token graphs. For the case of the token graphs of path graphs we obtain a formula for the distance between pair of vertices in Fk(Pn) that generalizes the one given by Beaula et al. [9].
The outline of this paper is as follows. In Section 2 we present some definitions, notation and some preliminary results. We show that Aut(Γ) is a subgroup of Aut(Fk(Γ)), for every graph Γ. Also, we show that if n=2k, then ∣Aut(Fk(Γ))∣≥2∣Aut(Γ)∣. The proof of Theorem 1.1 is worked for each case separately. In Section 3 we show that Aut(F2(Cn))=Aut(Cn)), for n=4. In Section 4 we prove that if Γ∈{K1,n,A1,n,W1,n}, then Aut(F2(Γ))=Aut(Γ), for n=4. In Section 5 we prove Theorem 1.2.
2 Preliminaries and first results
In this paper, all our graphs are simple and finite, that is, a graph Γ is a pair (V(Γ),E(Γ)) where V(Γ) is a finite set and E(Γ) is a subset of the set of all 2-subsets of V(Γ). An edge of a graph Γ will be denoted by {u,v} or uv, for u,v∈V(Γ). We use u∼v to indicate that u and v are adjacent vertices, that is uv∈E(Γ). The neighborhood of a vertex v is defined as N(v)={u∈V(Γ):uv∈E(Γ)} and the degree d(v) of v is defined as ∣N(v)∣. The neighborhood of a set of vertices X is defined as N(X)=⋃x∈XN(x)∖X. Let U be a subset of V(Γ), we will use Γ⟨U⟩ to denote the subgraph of Γ induced by U. The graph difference Γ−U is defined as the graph Γ⟨V(Γ)∖U⟩.
Let Γ1 and Γ2 be two simple graphs. An isomorphism of Γ1 onto Γ2 is a bijection ϕ:V(Γ1)→V(Γ2) such that uv∈E(Γ1) if and only if ϕ(u)ϕ(v)∈E(Γ2). An automorphism of a graph Γ is an isomorphism of Γ onto itself. The set of all automorphism of a graph Γ is a subgroup of Sym(V(Γ)), the group of all permutations of V(Γ), an is denoted by Aut(Γ). To obtain the automorphism group of graphs in general is a difficult problem. But it is possible to obtain this group for particular cases.
It is well-known that Aut(Γ) acts on V(Γ). Let v∈V(Γ), the orbit of v is defined as O(v)={f(v):f∈Aut(Γ)} and the stabilizer of v is Stab(x)={f∈Aut(Γ):f(x)=x}. The orbit-stabilizer theorem says that ∣Aut(Γ)∣=∣O(v)∣∣Stab(v)∣, for every v∈V(Γ). Let f∈Aut(Γ), we say that a vertex x∈V(Γ) is a fixed point of f if f(x)=x. We use Fix(f) to denote the set of fixed points of f. As usual, sometimes we write Aut(Γ)=G instead of Aut(Γ)≃G.
We use Sn to denote the symmetric group over {1,…,n}.
The following observation (see [14]) will be used, sometimes without reference, when we compute the degree of vertices in Fk(Γ).
Observation 2.1**.**
The degree of a vertex A in Fk(Γ) is equal to the number of edges between A and V(Γ)∖A.
In particular if {x,y}∈V(F2(Γ)), then d({x,y})=d(x)+d(y) if x∼y and d({x,y})=d(x)+d(y)−2 if x∼y.
The following result, that appears in [10] and [14], will be useful in some of the proofs.
Proposition 2.2**.**
Let X be a subset of V(Γ) and Γ′=Γ−X. Then Fk(Γ′) is isomorphic to the graph obtained from Fk(Γ) by deleting the vertices A in Fk(Γ) such that A has al least one element of X.
2.1 First results
Our first result shows an important relation between Aut(Γ) and Aut(Fk(Γ)). The proof of the following theorem is straightforward.
Theorem 2.3**.**
Let Γ be a graph. Then Aut(Γ) is isomorphic to a subgroup of Aut(Fk(Γ)). In fact, if θ∈Aut(Γ), then the function fθ:V(Fk(Γ))→V(Fk(Γ)) defined as
[TABLE]
is an automorphism of Fk(Γ).
The automorphism fθ of Fk(Γ), for θ∈Aut(Γ), defined in previous theorem is called the *automorphism induced * by θ. When the context is clear, we use Aut(Γ) as the set of automorphism of Γ or as the subgroup of Aut(Fk(Γ)) induced by the automorphisms of Γ. We write Aut(Fk(Γ))=Aut(Γ) to mean that every automorphism of Fk(Γ) is induced by some automorphism of Γ.
Now, the following proposition shows that for n=2k, Aut(Fk(Γ)) has always more elements than Aut(Γ).
Theorem 2.4**.**
Let Γ be a graph of order n, with n even. The function fc:V(Fn/2(Γ))→V(Fn/2(Γ)) defined as fc(A)=Ac is an automorphism, where Ac=V(Γ)∖A. Even more fc is not an induced automorphism of any ϕ∈Aut(Γ).
Proof.
The proof is exactly the same that the given in the proof of Theorem 3.5 in [25] for the case when Γ is the complete graph Kn (the graph Fk(Kn) is isomorphic to the Johnson graph J(n,k)).
∎
Clearly, the function fc in previous theorem is a fixed point free involution.
Corollary 2.5**.**
Let n≥4 be an even integer. If Γ is a graph of order n, then
[TABLE]
3 Automorphism group of the 2-token graph of cycle graphs
In this section we prove that Aut(F2(Cn))=Aut(Cn), for n=4. In Figure 1 we show F2(C7). Let D2n denote the dihedral group of 2n elements. It is well-known that Aut(Cn)=D2n. Using computer software, we obtain that Aut(F2(C4))=S2×D8. First we present some observations and results that will be useful. In this section, V(Cn)={1,2…,n} and E(Cn)={{i,i+1}:1≤i≤n−1}∪{{1,n}}.
Observation 3.1**.**
Let n≥4 be an integer.
-
If v∈F2(Cn), then d(v)∈{2,4}.
2. 2.
∣N(u)∩N(v)∣≤2, for every pair of vertices u,v∈F2(Cn).
We use i⊕j and i⊖j to denote the sum (i+j)modn and (i−j)modn, respectively, with the convention that n≡n(modn).
Let r=⌊n/2⌋. We define the following subsets of V(F2(Cn)).
[TABLE]
where 1≤q≤r.
The proof of the following proposition is an easy exercise.
Proposition 3.2**.**
Let n≥3 be an integer and r=⌊n/2⌋.
-
If n is even, then ∣Ln/2∣=n/2 and ∣Lq∣=n, for 1≤q<r.
2. 2.
If n is odd, then ∣Lq∣=n, for 1≤q≤r.
3. 3.
The set L={L1,…,Lr} is a partition of V(F2(Cn)).
4. 4.
Let n≥6 and 3≤q≤r. If {i,i⊕q}∈Lq, with 1≤i≤n, then two neighbors, say B,C, of {i,i⊕q} belongs to Lq−1 and the vertex in N(B)∩N(C)∖{{i,i⊕q}} belongs to Lq−2.
5. 5.
Let v∈V(F2(Cn)), then d(v)=2 if and only if v∈L1.
In Figure 1 we show the subgraph of F2(C7) induced by L1∪L2.
Proposition 3.3**.**
Let n≥6. The subgraph of F2(Cn) induced by L1∪L2 is isomorphic to C2n.
Proof.
First note that if {i,i⊕1} in L1, then
[TABLE]
and if {i,i⊕2} in L2, then
[TABLE]
Since n≥6, then N({i,i⊕1})⊂L2, N({i,i⊕2})∩L1={{i,i⊕1},{i⊕1,i⊕2}} and N({i,i⊕2})∩L3={{i,i⊕3},{i⊖1,i⊕2}}. Thus, it is easy to check that the function ϕ:V(L1∪L2)→V(C2n) given by ϕ({i,i⊕1})=2i−1, for every {i,i⊕1}∈L1, and ϕ({i,i⊕2})=2i, for every {i,i⊕2}∈L2, is a graph isomorphism.
∎
We also need the following well-known observation.
Proposition 3.4**.**
If f∈Aut(Cn) fixes two adjacent vertices on Cn, then f=id.
Now we present our main result of this section.
Theorem 3.5**.**
Let n≥3 be an integer. If n=4, then Aut(F2(Cn))=Aut(Cn).
Proof.
If n=3, then F2(C3)≃C3 and the result follows for this case. Suppose now that n≥5. Let Γ=F2(Cn). In the view of Theorem 2.3, it is enough to show that ∣Aut(Γ)∣≤2n. Let x∈V(Γ) be the vertex {1,n} and let Γ1=Γ⟨N(x)⟩. Since N(x)={{1,n−1},{2,n}}, then ∣Aut(Γ1)∣=2. Let φ:Stab(x)→Aut(Γ1) be the function defined by φ(f)=f∣Γ1. As φ is a group homomorphism we have that ∣Stab(x)∣≤∣Ker φ∣∣Aut(Γ1)∣≤2∣Ker φ∣. We will prove that Ker φ={id}.
Let f∈Ker φ. By Proposition 3.2(3) it is enough to show that f(Lq)⊂Fix(f), for every q∈{1,…,⌊n/2⌋}.
First we prove that f(L1∪L2)⊂Fix(f). Let Γ2=⟨L1∪L2⟩. By Proposition 3.3 it follows that Γ2≃C2n. Note that f(L1∪L2)=L1∪L2. Indeed, L1 is equal to the set of vertices of degree 2 in Γ and the vertices in L2 are the unique vertices in Γ that have two of its neighbors in L1 (see Figure 1 for an example). Then f∣L1∪L2∈Aut(Γ2). Since f∈Ker φ and f∈Stab(x), we have that f({1,n})={1,n}, f({1,n−1})={1,n−1} and f({2,n})={2,n}. But {1,n} and {2,n} are adjacent vertices in Γ and then all the vertices in L1∪L2 are fixed by f (by Proposition 3.4). If n=5 we are done. For n≥6 we will prove that f(Lq)⊂Fix(f), for 3≤q≤⌊n/2⌋. Suppose by induction that f(Lj)⊂Fix(f), for 2≤j<q≤⌊n/2⌋. Let u∈Lq, that is u={i,i⊕q}, for some i∈{1,…,n}. By Proposition 3.2(4), we have that two neighbors of u, say v and w, belongs to Lq−1 and the vertex z in N(v)∩N(w)∖{u} belongs to Lq−2 (see Figure 2). By hypothesis, {v,w,z}⊂Fix(f) which implies that f(u)∈N(v)∩N(w) and hence f(u)=u (we are using Observation 3.1 that shows that ∣N(v)∩N(w)∣≤2). Then f(Lq)⊂Fix(f) as desired.
Therefore Ker φ={id} and hence ∣Stab(x)∣≤2. Now, since x∈L1, by Proposition 3.2(5) it follows that ∣O(x)∣≤n. Finally, by the orbit-stabilizer theorem we obtain ∣Aut(Γ)∣=∣O(x)∣∣Stab(x)∣≤2n which concludes the proof of the theorem.
∎
The following conjecture is based on experimental results obtained by computer.
Conjecture 3.6**.**
Let n be an integer and 3≤k≤n/2. If k=n/2 then Aut(Fk(Cn))=Aut(Cn) and if k=n/2, then Aut(Fk(Cn))=S2×Aut(Cn).
4 Automorphism groups of the 2-token graphs of star, fan and wheel graphs
In this section we obtain the automorphism groups of the 2-token graphs of star, fan and wheel graphs. In all such cases we have that if ∣G∣≥5, then Aut(F2(G))=Aut(G).
Star graphs
First we consider the case of the star graph K1,n−1. In Figure 3 we show F2(K1,6). For small star graphs we have that F2(K1,2)≃P3 and F2(K1,3)≃C6 and hence Aut(F2(K1,2))≃S2 and Aut(F2(K1,3))≃D6. In this section V(K1,n−1)={1,2,…,n}, where 1 is the vertex of degree n−1.
Theorem 4.1**.**
Let n≥5 be an integer. Then
Aut(F2(K1,n−1))=Aut(K1,n−1).
Proof.
Let Γ=F2(K1,n−1). Since graph K1,n−1 is bipartite, it follows that Γ is also bipartite ([14, Proposition 1]). In fact, a bipartition of V(Γ) is {B,R}, where
[TABLE]
and R=V(Γ))∖B. Note that if x∈B, then d(x)=n−2 and if x∈R, then d(x)=2 (see Figure 3 for an example). Since n≥5, then Γ has exactly n−1 vertices of degree n−2. Let B={{1,2}} and O={{1,i}:3≤i≤n}. Clearly {B,O} is a partition of B. Let R={{2,i}:3≤i≤n} and G={{i,j}:3≤i<j≤n}. Clearly R=N({1,2}) and {R,G} is a partition of R. In Figure 3 we show the partition {B,O,R,G} of V(Γ).
Now, by Theorem 2.3, Aut(K1,n−1)≤Aut(Γ). It is well-known that Aut(K1,n−1)=Sn−1 and hence it is enough to show that ∣Aut(Γ)∣≤(n−1)!
Let x be the vertex {1,2} in Γ and let Γ1=Γ⟨N(x)⟩. In this case N(x)=R and Γ1=Kn−2. Therefore Aut(Γ1)=Sn−2. We have that ∣O(x)∣≤n−1 because x has degree n−2 and there are exactly n−1 vertices in Γ of degree n−2. If an automorphism f of Γ belongs to Stab(x), then f(N(x))=N(x) and hence f\raisebox{-2.15277pt}{|}_{N(x)}\in{\rm Aut}(\Gamma_{1}). Let φ:Stab(x)→Aut(Γ1) be the function defined by \varphi(f)=f\raisebox{-2.15277pt}{|}_{N(x)}. We will prove that Ker φ={id}.
Let f∈Ker φ. We will prove that f(Y)⊂Fix(f), for every Y∈{B,O,R,G} in the following order: B,R,O,G. Since f∈Stab(x) and f∈Ker φ, then f({1,2})={1,2} and f(y)=y, for every y∈N(x), respectively. That is, f(B)∪f(R)⊂Fix(f).
Now we prove that f(O)⊂Fix(f). Let w∈O, that is w={1,i}, for some i∈{3,…,n}. Note that N({2,i})={{1,2},{1,i}} and hence {1,i}∈N({2,i}). Since f({2,i})={2,i}, then f(N({2,i}))=N({2,i}). But f({1,2})={1,2} and then f({1,i})={1,i}, for every i∈{3,…,n}. Then f(O)⊂Fix(f). Finally, let w∈G, that is w={r,s}, with r,s∈{3,…,n}, r=s. We have that N({r,s})={{1,r},{1,s}}, that is a subset of O. Then N({r,s})⊂Fix(f). Since ∣N(y)∩N(z)∣<2, for every y,z∈R, with y=z, then f({r,s})={r,s}, for r,s∈{3,…,n}, r=s. Therefore f(G)⊂Fix(f). We conclude that f=id because {B,O,R,G} is a partition of V(Γ).
In this way Ker φ={id} and then ∣Stab(x)∣≤(n−2)!. Now we use that ∣Aut(Γ)∣=∣O(x)∣∣Stab(x)∣ to obtain that ∣Aut(Γ)∣≤(n−1)(n−2)!=(n−1)! as desired.
∎
The following conjecture is based on experimental results.
Conjecture 4.2**.**
Let n be an integer and 3≤k≤n/2. If k=n/2 then Aut(Fk(K1,n))=Aut(K1,n), and if k=n/2, then Aut(Fk(K1,n))=S2×Aut(K1,n).
Fan graphs
The fan graph A1,n is the joint graph K1+Pn. The vertices of A1,n are the disjoint union V(K1)∪V(Pn), where V(K1)={v} and V(Pn)={u1,…,un}, where ui∼ui+1 in Pn, for every 1≤i≤n−1. In Figure 4 we show F2(A1,7).
It is well-known that Aut(A1,n)≃S2, for n=3. For n≥5, it can be shown (by Observation 2.1) that the degrees of vertices in F2(A1,n−1) are as follows:
d({u1,u2})=3,d({u1,ui})=5, for 3≤i≤n−1, d({u1,v})=n, d({u1,un})=4;
d({ui,uj})∈{4,6}, for i,j∈{2,…,n−1},i=j;
d({ui,un})=5, for 2≤i≤n−2, d({un−1,un})=3,d({un,v})=n;
d(ui,v)=n+1, for every 2≤i≤n−1.
Therefore, if w∈V(F2(A1,n)), then d(w)∈{3,4,5,6,n,n+1}. Note that for n≥8, there exists exactly two vertices of degree 3 and exactly two vertices of degree n in F2(A1,n). We obtained that Aut(A1,3)≃S2×S2 by using Mathematica software.
Theorem 4.3**.**
Let n≥4 be an integer. Then Aut(F2(A1,n))=Aut(A1,n).
Proof.
The proof is by induction on n. The cases for n∈{4,…,8} were obtained by computer. In the rest of the proof n≥9. Assume as induction hypothesis that the result is true for every 4≤m<n.
Let Γ=F2(A1,n). By Theorem 2.3, Aut(A1,n)≤Aut(Γ) and hence it is enough to prove that ∣Aut(Γ)∣≤2. Let
[TABLE]
and ΓR=Γ⟨R⟩. We have that
[TABLE]
Note that ΓR is isomorphic to A1,n−1, where {v,un} is the vertex of degree n−1 in ΓR. Let ΓR=Γ−R. Note that ΓR=F2(A1,n−1), where the vertices of A1,n−1 are given by V(K1)={v} and V(Pn−1)={u1,…,un−1}.
Let x∈V(Γ) be the vertex {u1,u2}. We have that ∣Orb(x)∣≤2 because d(x)=3 and there are exactly two vertices of degree 3 in Γ.
Let Γ1=Γ⟨N(x)⟩, where N(x)={{u1,u3},{v,u1},{v,u2}}. Since the graph Γ1 is isomorphic to P3, then Aut(Γ1)=S2. Let φ:Stab(x)→Aut(Γ1) be the homomorphism given by f↦f∣N(x). Since the vertices {u1,u3} and {v,u2} have different degree in Γ, then f∣N(x)=id, for every f∈Stab(x). Therefore Im(φ)={id} and hence ∣Stab(x)∣=∣Ker φ∣. We will prove that Ker φ={id}.
Let f∈Ker φ. As f∈Stab(x) and f∣N(x)=id, we have that
[TABLE]
The unique vertices in Γ of degree n are {v,u1} and {v,un}. Therefore f({v,un})={v,un} and this implies that f(N({v,un}))=N({v,un}). Note that {v,un−1} is the unique vertex of degree n+1 in N({v,un}) and hence f({v,un−1})={v,un−1}. Then f(R)=R and hence f∣R∈Aut(ΓR) (remember that R=N({v,un})∪{{v,un}}∖{{v,un−1}}). This implies that f∣V(ΓR)∈Aut(ΓR).
Since Aut(ΓR) is isomorphic to Aut(A1,n−1), the image of {u1,un} under f∣R has only two possibilities: {u1,un} or {un−1,un}. But f({u1,un})={un−1,un} is imposible because {u1,un} and {un−1,un} have different degrees in Γ. Thus f∣R=id. Now Aut(ΓR)=Aut(F2(A1,n−1)) and by induction we have that either f∣V(ΓR)=id or f∣V(ΓR)=g, where g is the automorphism in Aut(F2(A1,n−1)) that moves the vertex {u1,u2}. But f({u1,u2})={u1,u2} and hence f∣V(ΓR)=id. Therefore f=id. In this way ∣Stab(x)∣=1 and hence ∣Aut(Γ)∣≤2 as desired.
∎
Conjecture 4.4**.**
Let n be an integer and 3≤k≤n/2. If k=n/2 then Aut(Fk(A1,n))=Aut(A1,n) and if k=n/2, then Aut(Fk(A1,n))=S2×Aut(A1,n).
Wheel graphs
The wheel graph W1,n, n≥3, is defined as the join graph K1+Cn. In Figure 5 we show F2(W1,7). It is well-known that Aut(W1,n)≃D2n. We obtained that Aut(W1,3)≃S2×D6 by using Mathematica software.
Theorem 4.5**.**
Let n≥4 be integer. Then Aut(F2(W1,n))=Aut(W1,n).
Proof.
In this proof, the vertex set of W1,n is {v,u1,…,un}, where V(K1)={v} and V(Cn)={u1,…,un}, with u1∼un and ui∼ui+1, 1≤i≤n in Cn. The cases n∈{4,5} were verified by computer and hence we suppose that n≥6. Let Γ=F2(W1,n). Let T={{ui,uj}∈V(Γ):1≤i<j≤n} and C={{ui,v}∈V(Γ):1≤i≤n}. Let ΓT=Γ⟨T⟩ and ΓC=Γ⟨C⟩. We have that ΓT=F2(Cn) and ΓC≃Cn (see Figure 5 for an example), where an isomorphism between Cn and ΓC is given by ui↦(ui,v). In this proof we use that every automorphism in Aut(ΓT) (that is equal to Aut(F2(Cn))) is induced by some automorphism in Aut(Cn). For the case of Aut(ΓC) we have that every automorphism θ∈Aut(Cn) induces an automorphism g in Aut(ΓC) given by g({ui,v})=({θ(ui),v}). We know that Aut(W1,n)≤Aut(F2(W1,n)) and we will prove that Aut(F2(W1,n))=Aut(W1,n) by showing that every automorphism f in Aut(F2(W1,n)) is induced by some automorphism θ in Aut(W1,n), i.e., f({a,b})={θ(a),θ(b)}, for every {a,b}∈V(F2(W1,n)).
Let y∈V(F2(W1,n)). Note that if y∈T, then d(y)∈{4,6}, and if y∈C, then d(y)=n+1. As n≥6, then d(y)∈{4,6} when y∈C. Therefore, we have that f∣T∈Aut(ΓT) and f∣C∈Aut(ΓC), for every f∈Aut(F2(W1,n)). As f∣T∈Aut(F2(Cn)), then there exists α∈Aut(Cn) such that f∣T({a,b})={α(a),α(b)}, for any {a,b}∈V(ΓT). For the case of f∣C we have that f∣C({a,v})={β(a),v}, for some β∈Aut(Cn).
We will prove by contradiction that α=β. Suppose that α=β. Without loss of generality, we suppose that α(u1)=β(u1).
Claim 4.6**.**
If α(u1)=β(u1), then α(u1)=β(u2),α(u2)=β(u1) and α(uj)=β(uj), for every j∈{3,…,n}.
Proof.
As {u1,u2}∼{u1,v}, then f({u1,u2})∼f({u1,v}), that is {α(u1),α(u2)}∼{β(u1),v}. By the definition of 2-token graph ∣{α(u1),α(u2)}∩{β(u1),v}∣=1. But v∈{α(u1),α(u2)} and α(u1)=β(u1), and then we have that α(u2)=β(u1). On the other hand {u1,u2}∼{u2,v} and then f({u1,u2})∼f({u2,v}). That is {α(u1),α(u2)}∼{β(u2),v}. But we have proved that α(u2)=β(u1) and hence {α(u1),β(u1)}∼{β(u2),v}. Similarly as in previous case the equality ∣{α(u1),β(u1)}∩{β(u2),v}∣=1 implies that α(u1)=β(u2).
Now, for j∈{3,…,n}, we have that {u1,uj}∼{uj,v}. Then
f({u1,uj})∼f({uj,v}), that is {α(u1),α(uj)}∼{β(uj),v}. But α(u1)=β(u2) and hence {β(u2),α(uj)}∼{β(uj),v}.
The unique option is that α(uj)=β(uj) and the proof of the claim is completed.
∎
Let a=β(u1) and b=β(u2). Using previous claim, it is easy to check that αβ−1(a)=b, αβ−1(b)=a, and αβ−1(c)=c, for every c∈V(Cn)∖{a,b}. That is, αβ−1 is equal to the transposition (a,b) (written in cyclic notation).
But αβ−1∈Aut(Cn)=D2n and the dihedral group D2n has not transpositions when n>3.
This contradiction shows that α=β. Therefore, f is the automorphism induced by θ∈Aut(W1,n), where θ(v)=v and θ(ui)=α(ui), for every i∈{1,…,n}.
∎
Conjecture 4.7**.**
Let n be an integer and 3≤k≤n/2. If k=n/2 then Aut(Fk(W1,n))=Aut(W1,n) and if k=n/2, then Aut(Fk(W1,n))=S2×Aut(W1,n).
5 Proof of Theorem 1.2
In this section we use Pn to denote the path graph with V(Pn)={1,…,n} and E(Pn)={{i,i+1}:1≤i≤n−1}. It is well-known that Aut(P1)≃S1 and Aut(Pn)≃S2, for n≥2. Explicitly, if n≥2, then Aut(Pn)={id,θ}, where
[TABLE]
and
[TABLE]
(We are writing permutation θ in its cycle notation).
Our main result in this section is Theorem 1.2 about the automorphism group of Fk(Pn), for n=2k. In Figure 6 and 7 we show F2(P6) and F3(P7), respectively.
First, we present some auxiliary results. Without loss of generality, we write the elements of a vertex {a1,…,ak}∈V(Fk(Pn)) in ascending order, that is a1<⋯<ak. Using Observation 2.1 we obtain the following facts about the vertices in Fk(Pn).
Observation 5.1**.**
Let Pn be a path graph of order n≥3 and let 2≤k≤n−2 be an integer.
-
Let v={a1,a2,…,ak} be a vertex in Fk(Pn). The degree of v is even if and only if {a1,ak}={1,n} or {a1,ak}∩{1,n}=∅.
2. 2.
The vertices of degree 2 in Fk(Pn) are either of the form {a,a+1,…,a+(k−1)}, for 2≤a≤n−k, or {1,…,m,n−(k−m−1),n−(k−m),…,n}, for 1≤m≤k−1.
3. 3.
The graph Fk(Pn) has exactly two vertices of degree one, say {1,…,k} and {n−(k−1),…,n}.
The formula for the graph distance d(u,v) between vertices u and v in Fk(Pn) is given in the following result.
Lemma 5.2**.**
Let u={u1,…,uk} and v={v1,…,vk} be vertices in Fk(Pn), where u1<⋯<uk and v1<⋯<vk. Then
[TABLE]
Proof.
The L1-distance on Zk is defined as
[TABLE]
for every a={a1,…,ak},b={b1,…,bk}∈Zk. The grid graph Zk is constructed as follows: two points in Zk are adjacent if their L1-distance is equal to one. It is well-known (see, e.g., [13, p. 333]) that the L1-distance is equal to the path distance on the grid graph Zk. The result follows by observing that Fk(Pn) is a subgraph of the grid graph Zk, where vertex {ai1,…,aik} in Fk(Pn) correspond to the unique vertex (a1,…,ak) in Zk such that a1<⋯<ak and {ai1,…,aik}={a1,…,ak}.
∎
The proof of previous lemma also follows by using the formula of distance in [15]. The case for F2(Pn) was also proved in [9]. In the proof of the following theorem, we use several times the following fact: if H is a graph, then
[TABLE]
for every g∈Aut(H) and every a,b∈V(H).
Proof of Theorem 1.2.
If k=1, then F1(Pn)≃Pn and the result is trivially true. The cases n≤7, 1≤k≤n/2, were verified by computer and hence we suppose that n≥8. Let Γ=Fk(Pn).
By Theorem 2.3, it follows that Aut(Pn)≤Aut(Γ) and hence it is enough to prove that ∣Aut(Γ)∣≤2.
Let x∈Γ be the vertex {1,…,k}. Note that d(x)=1 and hence ∣O(x)∣≤2. In fact, ∣O(x)∣=2 because the non-identity automorphism, say ϕ, in Aut(Pn) induces a non-identity automorphism gϕ in Aut(Γ) such that gϕ({1,…,k})={n−(k−1),…,n}. Let Γ1=Γ⟨N(x)⟩. Let φ:Stab(x)→Aut(Γ1) be the function defined by \varphi(f)=f\raisebox{-2.15277pt}{|}_{N(x)}. As Aut(Γ1)={id}, we have that Stab(x)=Ker φ. We will prove that Ker φ={id}, which implies that ∣Aut(Γ)∣≤2.
First we prove the case k=2. Suppose by induction that F2(Pm)=Aut(Pm), for every 5≤m<n. Let L1={{a1,a2}∈V(Γ):a1=1 or a2=n} and L2=V(Γ)∖L1. Let ΓL1=Γ⟨L1⟩ and let ΓL2=Γ⟨L2⟩. Note that ΓL1≃P2n−3 and, by Proposition 2.2, we have that ΓL2≃F2(Pn−2) (see Figure 6 for an example). It is easy to see that, with the exception of {1,n}, all the vertices in L1 have either degree 1 or degree 3. In fact L1 has all the vertices of Γ that have degree 1 or 3. Note that {1,n} is the unique vertex in Γ of degree 2 in with its two neighbors of degree 3. Consequently, if f∈Aut(Γ), then f(L1)=L1 and hence f∣L1∈Aut(ΓL1). Now, let f∈Ker φ. Since Aut(L1)≃Aut(P2n−3) and f({1,2})={1,2}, we have that f∣L1=id which shows that f(L1)⊂Fix(f).
Using this, the proof that f(L2)⊂Fix(f) follows immediately by induction because ΓL2≃F2(Pn−2) and n−2≥6 (see Figure 6 for an example).
Now we prove the case 3≤k<n/2. The proof is by induction on n. As bases cases, we have proved the result for F2(Pn), for every n≥5, and for n=8, we have verified every k∈{2,3} by computer. We suppose that the result is true for any graph Pn′ with n′<n. That is, Aut(Fk′(Pn′))=Aut(Pn′), for 2≤k′<n′/2<n/2.
Let f∈Ker φ. Since f∈Stab(x), we have that f({1,…,k−1,k+1})=({1,…,k−1,k+1} and f({n−(k−1),…,n−1,n})=({n−(k−1),…,n−1,n}) (the grey vertices in the example in Figure 7).
We define the following subsets of V(Γ)
[TABLE]
Note that C=V(Γ)∖(A∪B). Let ΓA=Γ⟨A⟩, ΓB=Γ⟨B⟩ and ΓC=Γ⟨C⟩. We will prove that if f in Ker φ, then f is the identity permutation by showing that X⊂Fix(f), for every X∈{A,B,C}. We do this in four steps: first we prove that if f∈Ker φ, then f({1,…,k−1,n})=({1,…,k−1,n} (Claim 5.4), second we work the case of A, third the case of B and finally the case of C. In Figure 7, we use colors to illustrate the four steps.
We need the following:
Claim 5.3**.**
Let y be the vertex {1,…,k−1,n}.
-
If 3≤a≤n−k−1, then {a,a+1,…,a+(k−1)}∈O(y),
2. 2.
If 2≤m≤k−2, then {1,…,m,n−(k−m−1),n−(k−m),…,n}∈O(y).
Proof.
The elements in O(y) should be vertices of degree 2. We will use the fact that if g∈Aut(Γ), then N(g(v))=g(N(v)), for every v∈V(Γ). Note that N(y)={y1,y2}, where y1={1,…,k−2,k,n} and y2={1,…,k−1,n−1}. The condition k≥3 implies that d(y1)=4 and d(y2)=3.
Proof of (1). Let w={a,a+1,…,a+(k−1)}, with 3≤a≤n−k−1, then N(w)={w1,w2}, where w1={a−1,a+1,…,a+(k−1)}, w2={a,a+1,…,a+(k−2),a+k}, d(w1)=d(w2)=4. Therefore f(y)=w.
Proof of (2). Let z={1,…,m,n−(k−m−1),n−(k−m),…,n}, with 2≤m≤k−2. Then N(z)={z1,z2}, where z1={{1,…,m−1,m+1,n−(k−m−1),n−(k−m),…,n}, z2={1,…,m,n−(k−m−1)−1,n−(k−m),…,n} and d(z1)=d(z2)=4 (here we are using that k<n/2). Therefore f(y)=z.
∎
Step 1. We prove the following fact.
Claim 5.4**.**
Let y be the vertex {1,…,k−1,n}. If f∈Ker φ, then f(y)=y.
Proof.
By Observation 5.1(2) and Claim 5.3 we have that f(y) has only the following options:
-
{2,3,…,k+1},
2. 2.
{n−k,n−(k−1),…,n−1},
3. 3.
{1,n−(k−2),…,n}
4. 4.
{1,…,k−1,n},
Now we use several times the fact that d(a,b)=d(g(a),g(b)), for every g∈Aut(Γ) and every a,b∈V(Γ).
Case 1. Suppose that f({1,…,k−1,n})={2,3,…,k+1}. Then
[TABLE]
which implies that n=2k, a contradiction.
Case 2. Suppose that f({1,…,k−1,n})={n−k,n−(k−1),…,n−1}. Then
[TABLE]
From which we obtain k2−nk+n=0. The discriminant of equation x2−nx+n=0 is D=n2−4n. As n≥7, D is positive and not a square number. Therefore x2−nx+n=0 has not integer solutions, which is a contradiction.
Case 3. Suppose that f({1,…,k−1,n})={1,n−(k−2),…,n}}. Then
[TABLE]
As n=k, then k=2, a contradiction that k≥3.
Therefore, the unique option is that f(y)=y as desired.
∎
In Figure 7 we show an example where vertex {1,…,k−1,n} is colored black.
Step 2. We will prove that f(A)⊂Fix(f) (the red vertices in the example in Figure 7).
First we prove that f(A)=A. Let v={a1,…,ak−1,n} be a vertex in A. Suppose that f(v)={c1,…,ck}. Then
[TABLE]
from which follows that
[TABLE]
Now, by Claim 5.4, f({1,…,k−1,n})={1,…,k−1,n}. Then
[TABLE]
from which follows that
[TABLE]
Combining equations (1) and (2) we obtain that ck=n and hence f(v)∈A. Then f(A)⊆A. In fact, f(A)=A because f is a bijection. Therefore f∣A∈Aut(ΓA) .
Notice that ΓA is isomorphic to Fk−1(Pn−1) (see Figure 8 for the case of F3(P7)), with an isomorphism given by {a1,…,ak−1,n}↦{a1,…,ak−1}. The inequality 3≤k<n/2, implies that 2≤k−1<(n−1)/2. By the induction hypothesis, Aut(Fk−1(Pn−1))=S2. The unique option is that f∣A=id because {n−(k−1),…,n−1,n} is the unique vertex in A with degree 1 simultaneously in ΓA and Fk(Pn) (see Figure 8 for an example).
Step 3. We will prove that f(B)⊂Fix(f) (the blue vertices in the example in Figure 7).
First, we prove that f(B)=B. Let v={1,v2,…,vk}∈B. We have two cases: vk=n or vk=n. If vk=n, then v∈A and hence f(v)∈B, because we have proved that f(v)=v, for every v∈A. Suppose now that vk=n, that is v∈A. By Observation 5.1(1) it follows that d(v) is odd and hence d(f(v)) should be odd. Also by Observation 5.1(1), f(v) has only two options. The first one is f(v)={1,…,dk}, with dk=n, in which case f(v)∈B as desired. The second option is f(v)={a1,…,n}, with a1=1. In this case f(v)∈A and hence f(v)=v∈A, a contradiction.
Therefore, f(B)⊆B and hence f(B)=B because f is a bijection. As B≃Fk−1(Pn−1), with an isomorphism given by {1,b2,…,bk}↦{b2,…,bk}, then Aut(ΓB)≃S2 and hence f∣B=id as in the case of ΓA. Therefore f(A∪B)⊂Fix(f).
Step 4. We will prove that f(C)⊂Fix(f) (the green vertices in the example in Figure 7).
It is easy to see that Γ⟨C⟩≃Fk(Pn−2). However, we can not apply the induction hypothesis on C because it is possible that k≥(n−2)/2 (for example, if n=10 and k=4). We solve this inconvenient by using the graph distance in Γ.
Suppose that f({c1,…,ck})={d1,…,dk}.
[TABLE]
and this implies that c1≥d1. Now
[TABLE]
and this implies that d1≥c1. Therefore, d1=c1.
Now we will prove that dk=ck.
[TABLE]
and this implies that dk≥ck. Now
[TABLE]
and this implies that dk≤ck. Therefore, dk=ck.
Finally we prove that ci=di, for every 2≤i≤k−1.
We have proved that if c={c1,c2,…,ck−1,ck}, then f(c)={c1,d2,…,dk−1,ck}. Therefore
[TABLE]
From which we obtain that ∑i=2k−1∣di−ci∣=0. But as every ∣di−ci∣ is non negative, then ∣di−ci∣=0, for every i∈{2,…,k−1} and hence di=ci, for every i∈{2,…,k−1}.
So we have that f(c)=c, for every c∈C.
Therefore f=id and the proof of the theorem is completed.
∎
We finish this paper with the following:
Conjecture 5.5**.**
Let n be an even integer. Then Aut(Fn/2(Pn))≃S2×S2.
Acknowledgments
The authors would like to thank the reviewer for her/his useful corrections.