This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Commuting graph of a group on a transversal

Julio C. M. Pezzotta111The author acknowledges a scholarship from CAPES (Brazil) , Irene N. Nakaoka b
a,bUniversidade Estadual de Maringá, Brazil
a e-mail: juliopezzott@gmail.com
b e-mail: innakaoka@uem.br
Abstract

Given a finite group GG and a subset XX of GG, the commuting graph of GG on XX, denoted by 𝒞(G,X){\cal C}(G,X), is the graph that has XX as its vertex set with x,yXx,y\in X joined by an edge whenever xyx\neq y and xy=yxxy=yx. Let TT be a transversal of the center Z(G)Z(G) of GG. When GG is a finite non-abelian group and X=TZ(G)X=T\setminus Z(G), we denote the graph 𝒞(G,X){\cal C}(G,X) by 𝒯(G){\cal T}(G). In this paper, we show that 𝒯(G){\cal T}(G) is a connected strongly regular graph if and only if GG is isoclinic to an extraspecial 22-group of order at least 3232. We also characterize the finite non-abelian groups GG for which the graph 𝒯(G){\cal T}(G) is disconnected strongly regular.

Keywords: Commuting graph, isoclinism, extraspecial pp-group, strongly regular graph.

Mathematics Subject Classification 2010: 20D15, 05C25, 05E30

1 Introduction

Let XX be a subset of a finite group GG. The commuting graph of GG on the set XX, denoted by 𝒞(G,X){\cal C}(G,X), is the graph that has XX as its vertex set with x,yXx,y\in X joined by an edge whenever xyx\neq y and xy=yxxy=yx. Commuting graphs were first studied by Brauer and Fowler in [4] with X=G{1}X=G\setminus\{1\}. Many papers have investigated 𝒞(G,X){\cal C}(G,X) for different choices of XX. For example, the works [2, 7, 16] investigated the graph 𝒞(G,X){\cal C}(G,X) when XX consists of involutions and the papers [1, 8, 9, 14, 17] considered X=GZ(G)X=G\setminus Z(G), where Z(G)Z(G) denotes the center of GG. Let us denote the graph 𝒞(G,GZ(G)){\cal C}(G,G\setminus Z(G)) by Γ(G)\Gamma(G) and refer to it simply as commuting graph of GG. We observe that Γ(G)\Gamma(G) highlights the relations of commutativity between non-central elements of the group GG. However, given xGx\in G, it is clear that any two elements of the coset xZ(G)xZ(G) commute with each other. Furthermore, as observed in [17], vertices xx and yy are adjacent in Γ(G)\Gamma(G) if and only if gg and hh are adjacent for all gxZ(G)g\in xZ(G) and hyZ(G)h\in yZ(G). In this way, we may investigate the relations of commutativity in GG by observing only the relations of commutativity that occur between non-central elements of a transversal of Z(G)Z(G) in GG. Thus, in this work we consider the graph 𝒞(G,X){\cal C}(G,X) when GG is a finite non-abelian group and X=TZ(G)X=T\setminus Z(G), where TT is a transversal of Z(G)Z(G) in GG. Note that it is a subgraph of Γ(G)\Gamma(G). It is easy to see that if TT^{\prime} is another transversal of Z(G)Z(G) in GG, then the graphs 𝒞(G,TZ(G)){\cal C}(G,T\setminus Z(G)) and 𝒞(G,TZ(G)){\cal C}(G,T^{\prime}\setminus Z(G)) are isomorphic. Hence, we will denote the graph 𝒞(G,TZ(G)){\cal C}(G,T\setminus Z(G)) simply by 𝒯(G){\cal T}(G) without mentioning the choice of the transversal. It is worth mentioning that this type of graph has already been considered in [8, 9, 17]. Vahidi and Talebi [17] showed that the graphs Γ(G)\Gamma(G) and 𝒯(G){\cal T}(G) have the same independence number and diameter. Moreover, if ω(𝒢)\omega({\cal G}) denotes the clique number of the graph 𝒢{\cal G}, then ω(Γ(G))=ω(𝒯(G))|Z(G)|\omega(\Gamma(G))=\omega({\cal T}(G))|Z(G)|. In [8, 9] the authors examined the graph 𝒯(G){\cal T}(G) in their study of the diameter of a commuting graph.

We prove that 𝒯(G){\cal T}(G) presents the following relevant property: if GG and HH are isoclinic groups, then the graphs 𝒯(G){\cal T}(G) and 𝒯(H){\cal T}(H) are isomorphic (Proposition 2.3). This helps us to classify groups GG such that 𝒯(G){\cal T}(G) has certain particular properties.

We recall a graph is regular if each vertex has the same number of neighbors and it is kk-regular if each vertex has exactly kk neighbours. A strongly regular graph with parameters (v,k,λ,μ)(v,k,\lambda,\mu), where 0<k<v10<k<v-1, is a graph kk-regular on vv vertices such that each pair of adjacent vertices has precisely λ\lambda common neighbours and any two non-adjacent vertices have exactly μ\mu common neighbours.

Akbari and Moghaddamfar [1, Corollary 2] showed that if Γ(G)\Gamma(G) is strongly regular, then it is a disjoint union of at least two complete graphs and, therefore, it is disconnected. In this work, we prove that there exist groups GG for which the associated graph 𝒯(G){\cal T}(G) is connected strongly regular. Moreover, we present a description of these groups; more precisely, we get:

Theorem A.

Let GG be a finite non-abelian group. Then 𝒯(G){\cal T}(G) is a connected strongly regular graph if and only if GG admits a decomposition G=A×PG=A\times P, where AA is an abelian subgroup of GG and PP is a Sylow 22-subgroup of GG isoclinic to an extraspecial 22-group of order 22n+12^{2n+1}, for some integer n2n\geq 2. In this case, the parameters of 𝒯(G){\cal T}(G) are (22n1,22n12,22n23,22n21)(2^{2n}-1,2^{2n-1}-2,2^{2n-2}-3,2^{2n-2}-1).

We also characterize the finite non-abelian groups GG for which the graph 𝒯(G){\cal T}(G) is disconnected strongly regular (see Corollary 3.10).

In this text, we use the following notations and conventions: CnC_{n} is the cyclic group of order nn and D2mD_{2m} (with m3m\geq 3) is the dihedral group of order 2m2m. Let GG be a group. For x,yGx,y\in G, the commutator of xx and yy is given by [x,y]=xyx1y1[x,y]=xyx^{-1}y^{-1} and the derived subgroup of GG is denoted by GG^{\prime}. We write CG(x)C_{G}(x) for the centralizer of xx in GG. The symbol cs(G)cs(G) represents the set formed by the sizes of the conjugation classes of the elements of GG. For 1<m1<<mn1<m_{1}<\ldots<m_{n}, we say that GG is of conjugate type {1,m1,,mn}\{1,m_{1},...,m_{n}\} if cs(G)={1,m1,,mn}cs(G)=\{1,m_{1},...,m_{n}\}. Thus, given a positive integer mm, the group GG is of conjugate type {1,m}\{1,m\} if [G:CG(x)]=m[G:C_{G}(x)]=m, for any xGZ(G)x\in G\setminus Z(G).

Given a graph 𝒢\mathcal{G} its vertex set is represented by V(𝒢)V(\mathcal{G}) and its edge set by E(𝒢)E(\mathcal{G}). The neighborhood of a vertex xx of 𝒢\mathcal{G} is denoted by N(x)N(x) (={yV(𝒢):{x,y}E(𝒢)})(=\{y\in V(\mathcal{G})\,:\,\{x,y\}\in E(\mathcal{G})\}) with degree degx=|N(x)|\,x=|N(x)|. When the graphs 𝒢1\mathcal{G}_{1} and 𝒢2\mathcal{G}_{2} are isomorphic, we write 𝒢1𝒢2\mathcal{G}_{1}\cong\mathcal{G}_{2} and the disjoint union of the graphs 𝒢1\mathcal{G}_{1} and 𝒢2\mathcal{G}_{2} is denoted by 𝒢1𝒢2\mathcal{G}_{1}\cup\mathcal{G}_{2}. As usual, the complete graph on nn vertices is denoted by KnK_{n} and the disjoint union of rr complete graphs KnK_{n} is indicated by rKnrK_{n}. The graph on mm vertices and without edges is represented by ImI_{m}. Other concepts and basic results on graphs can be seen in [6].

2 Preliminary results

In this section we present some concepts and results which will be needed to the proof of Theorem A.

First, we observe the commuting graph Γ(G)\Gamma(G) of GG can be obtained from 𝒯(G){\cal T}(G) in the following manner: for each vertex xx of 𝒯(G){\cal T}(G), put Vx=xZ(G)V_{x}=xZ(G). We can see that Γ(G)\Gamma(G) is the graph that has V:=xV(𝒯(G))VxV:=\bigcup_{x\in V({\cal T}(G))}V_{x} as vertex set and such that a subset {a,b}\{a,b\} of VV is an edge of Γ(G)\Gamma(G) if and only if one of the following conditions is satisfied: (1) a,bVxa,b\in V_{x}, for some vertex xx of 𝒯(G){\cal T}(G); or (2) aVxa\in V_{x} and bVyb\in V_{y}, where xx e yy are adjacent vertices in 𝒯(G){\cal T}(G).

Example 2.1.

Consider the group M(3)=a,b,c:a3,b3,c3,[a,b]c1,[a,c],[b,c]M(3)=\langle a,b,c:a^{3},b^{3},c^{3},[a,b]c^{-1},[a,c],[b,c]\rangle, which is an extraspecial 33-group of order 2727. Since Z(M(3))={e,c,c2}Z(M(3))=\{e,c,c^{2}\}, there are eight non-trivial cosets of Z(M(3))Z(M(3)) in M(3)M(3) and T={1,a,a2,b,b2,ab,ab2,a2b,a2b2}T=\{1,a,a^{2},b,b^{2},ab,ab^{2},a^{2}b,a^{2}b^{2}\} is a transversal of Z(M(3))Z(M(3)) in M(3)M(3). The commuting graph of M(3)M(3) is isomorphic to 4K64K_{6} and it can be seen in Figure 1. Each connected component of Γ(M(3))\Gamma(M(3)) contains exactly two distinct cosets of Z(M(3))Z(M(3)). In the figure below, in each connected component of Γ(M(3))\Gamma(M(3)), vertices that are in the same coset of Z(M(3))Z(M(3)) are represented with the same color. In Figure 2, we see that 𝒯(M(3))4K2{\cal T}(M(3))\cong 4K_{2}.

Refer to caption
Figure 1: Γ(M(3))4K6\Gamma(M(3))\cong 4K_{6}
Refer to caption
Figure 2: 𝒯(M(3))4K2{\cal T}(M(3))\cong 4K_{2}

The concept below was introduced by Hall [11]. We observe that for all group GG, the commutator map αG:G/Z(G)×G/Z(G)G\alpha_{G}:G/Z(G)\times G/Z(G)\rightarrow G^{\prime} given by αG(xZ(G),yZ(G))=[x,y]\alpha_{G}(xZ(G),yZ(G))=[x,y] is well defined.

Definition 2.2.

Let GG and HH be two groups. A pair (φ,ψ)(\varphi,\psi) is an isoclinism from GG to HH if the following conditions are satisfied:

(i)

φ\varphi is an isomorphism from G/Z(G)G/Z(G) to H/Z(H)H/Z(H);

(ii)

ψ\psi is an isomorphism from GG^{\prime} to HH^{\prime};

(iii)

ψ(αG(xZ(G),yZ(G)))=αH(φ(xZ(G)),φ(yZ(G)))\psi(\alpha_{G}(xZ(G),yZ(G)))=\alpha_{H}(\varphi(xZ(G)),\varphi(yZ(G))), for all x,yGx,y\in G.

When there is an isoclinism from GG to HH, we say that the groups GG and HH are isoclinic.

The relation of isoclinism is a equivalence relation on groups and the equivalence class of a group GG is called of isoclinism family of GG.

In general, isoclinic groups do not have isomophic commuting graphs, however for the graphs 𝒯(G){\cal T}(G) we have the following:

Proposition 2.3.

Let GG and HH be finite non-abelian groups. If GG and HH are isoclinic, then the graphs 𝒯(G){\cal T}(G) and 𝒯(H){\cal T}(H) are isomorphic.

Proof.

By hypothesis, there are isomorphisms φ:G/Z(G)H/Z(H)\varphi:G/Z(G)\rightarrow H/Z(H) and ψ:GH\psi:G^{\prime}\rightarrow H^{\prime} such that ψ(αG(xZ(G),yZ(G)))=αH(φ(xZ(G)),φ(yZ(G)))\psi(\alpha_{G}(xZ(G),yZ(G)))=\alpha_{H}(\varphi(xZ(G)),\varphi(yZ(G))), for all x,yGx,y\in G. Let T(G)={1,t1,,tn}T(G)=\{1,t_{1},\ldots,t_{n}\} be a transversal of Z(G)Z(G) in GG, and, for each i{1,,n}i\in\{1,\ldots,n\}, let yiy_{i} be a representative of the coset φ(tiZ(G))\varphi(t_{i}Z(G)). Therefore T(H)={1,y1,,yn}T(H)=\{1,y_{1},\ldots,y_{n}\} is a transversal of Z(H)Z(H) in HH. If we put T(G){1}T(G)\setminus\{1\} as the vertex set of 𝒯(G){\cal T}(G) and T(H){1}T(H)\setminus\{1\} as the vertex set of 𝒯(H){\cal T}(H), the mapping tiyit_{i}\longmapsto y_{i} defines a bijection between V(𝒯(G))V({\cal T}(G)) and V(𝒯(H))V({\cal T}(H)). In addition, if tit_{i} e tjt_{j} are adjacent in 𝒯(G){\cal T}(G), then [yi,yj]=ψ([ti,tj])=1[y_{i},y_{j}]=\psi([t_{i},t_{j}])=1, that is, yiy_{i} and yjy_{j} are adjacent in 𝒯(H){\cal T}(H). On the other hand, if [yi,yj]=1[y_{i},y_{j}]=1, we have ψ([ti,tj])=1\psi([t_{i},t_{j}])=1. Since ψ\psi is an isomorphism, we obtain [ti,tj]=1[t_{i},t_{j}]=1. Therefore 𝒯(G){\cal T}(G) and 𝒯(H){\cal T}(H) are isomorphic. ∎

The next example shows that the converse of Proposition 2.3 does not hold in general.

Example 2.4.

Consider the following groups :

J:=a,b,ca4=b4=c4=1,a2=b2,ab=ba,cac1=a1,bcb1=c1J:=\langle a,b,c\mid a^{4}=b^{4}=c^{4}=1,a^{2}=b^{2},ab=ba,cac^{-1}=a^{-1},bcb^{-1}=c^{-1}\rangle

D16:=x,yx8=y2=1,yxy1=x1D_{16}:=\langle x,y\mid x^{8}=y^{2}=1,\,yxy^{-1}=x^{-1}\rangle.

We have Z(J)={1,a2,c2,a2c2}Z(J)=\{1,a^{2},c^{2},a^{2}c^{2}\} and Z(D16)={1,x4}Z(D_{16})=\{1,x^{4}\}; further T1={1,a,b,c,ab,bc,ac,abc}T_{1}=\{1,a,b,c,ab,bc,ac,abc\} is a transversal of Z(J)Z(J) in JJ and T2={1,x,x2,x3,y,xy,x2y,x3y}T_{2}=\{1,x,x^{2},x^{3},y,xy,x^{2}y,x^{3}y\} is a transversal of Z(D16)Z(D_{16}) in D16D_{16}. It is not difficult verify that 𝒯(J)𝒯(D16)K3I4{\cal T}(J)\cong{\cal T}(D_{16})\cong K_{3}\cup I_{4} (Figures 4 and 4); however, JJ e D16D_{16} are not isoclinic, because the groups J/Z(J)J/Z(J), D16/Z(D16)D_{16}/Z(D_{16}) are isomorphic to C23C_{2}^{3}, D8D_{8}, respectively.

Refer to caption
Figure 3: 𝒯(J){\cal T}(J)
Refer to caption
Figure 4: 𝒯(D16){\cal T}(D_{16})

The next result gives the size of the common neighborhood of two vertices of 𝒯(G){\cal T}(G).

Proposition 2.5.

Given a finite non-abelian group GG, let xx and yy be distinct vertices of 𝒯(G){\cal T}(G). The following statements are true:

(i)

deg(x)=[CG(x):Z(G)]2(x)=[C_{G}(x):Z(G)]-2;

(ii)

if xx and yy are adjacent, then |N(x)N(y)|=[CG(x)CG(y):Z(G)]3|N(x)\cap N(y)|=[C_{G}(x)\cap C_{G}(y):Z(G)]-3;

(iii)

if xx and yy are not adjacent, then |N(x)N(y)|=[CG(x)CG(y):Z(G)]1|N(x)\cap N(y)|=[C_{G}(x)\cap C_{G}(y):Z(G)]-1.

Proof.

Let xx be a vertex of 𝒯(G){\cal T}(G). If N(x)={x1,,xn}N(x)=\{x_{1},\ldots,x_{n}\} is the neighborhood of xx in 𝒯(G){\cal T}(G) and if 𝒜:=i=1nxiZ(G)\mathcal{A}:=\bigcup_{i=1}^{n}x_{i}Z(G), it is easy to verify that CG(x)=Z(G)xZ(G)𝒜C_{G}(x)=Z(G)\cup xZ(G)\cup\mathcal{A}, so that deg(x)=n=[CG(x):Z(G)]2(x)=n=[C_{G}(x):Z(G)]-2. This proves (i).

Let xx and yy be distinct vertices of 𝒯(G){\cal T}(G) and put N=N(x)N(y)N=N(x)\cap N(y) and :=aNaZ(G)\mathcal{B}:=\bigcup_{a\in N}aZ(G). We note that {x,y}N=\{x,y\}\cap N=\emptyset. If xx and yy are adjacent in 𝒯(G){\cal T}(G), then {x,y}CG(x)CG(y)\{x,y\}\subseteq C_{G}(x)\cap C_{G}(y) and, in this case, we obtain CG(x)CG(y)=Z(G)xZ(G)yZ(G)C_{G}(x)\cap C_{G}(y)=Z(G)\cup xZ(G)\cup yZ(G)\cup\mathcal{B}. It follows that |N|=[CG(x)CG(y):Z(G)]3|N|=[C_{G}(x)\cap C_{G}(y):Z(G)]-3, which proves (ii). If xx and yy are not adjacent in 𝒯(G){\cal T}(G), then xCG(y)x\notin C_{G}(y) and yCG(x)y\notin C_{G}(x) and it is easy to see that CG(x)CG(y)=Z(G)C_{G}(x)\cap C_{G}(y)=Z(G)\cup\mathcal{B}. Therefore |N|=[CG(x)CG(y):Z(G)]1|N|=[C_{G}(x)\cap C_{G}(y):Z(G)]-1 and part (iii) is proved. ∎

From Proposition 2.5 it follows that 𝒯(G){\cal T}(G) is regular if and only if |CG(x)|=|CG(y)||C_{G}(x)|=|C_{G}(y)|, for all x,yGZ(G)x,y\in G\setminus Z(G), that is, |cs(G)|=2|cs(G)|=2. Ito [13] gave a description of the finite groups of conjugate type {1,m}\{1,m\}.

Theorem 2.6.

([13, Theorem 1]) If GG is a finite group of conjugate type {1,m}\{1,m\}, then mm is a power of a prime pp and GG admits a decomposition G=A×PG=A\times P, where AA is an abelian subgroup of GG and PP is a Sylow pp-subgroup of the same conjugate type as GG.

Let pp be a prime. Recall that a finite non-abelian pp-group is extraspecial if G=Z(G)=Φ(G)G^{\prime}=Z(G)=\Phi(G) and |Z(G)|=p|Z(G)|=p, where Φ(G)\Phi(G) denotes the Frattini subgroup of GG. It is well known that any non-abelian group of order p3p^{3} is extraspecial. Further, an extraspecial pp-group has order p2n+1p^{2n+1}, for some positive integer nn.

In this paper we will use the following Ishikawa’s classification:

Theorem 2.7.

([12, Proposition 3.1]) Let GG be a finite pp-group, pp prime. Then GG is of conjugate type {1,p}\{1,p\} if and only if GG is isoclinic to an extraspecial pp-group.

3 Main results

In this section we will present the proof of Theorem A, which will be done in steps. First let us show that the graph 𝒯(G){\cal T}(G) associated to an extraspecial 22-group of order 22n+12^{2n+1}, n2n\geq 2, is connected strongly regular. For it, we will need the following two lemmas:

Lemma 3.1.

Let GG be an extraspecial 22-group. If x,yGx,y\in G and xyZ(G)x\notin yZ(G), then CG(x)CG(y)C_{G}(x)\neq C_{G}(y).

Proof.

If G=D8G=D_{8} or GG is the quaternion group of order 88, the result is clearly true. Consider an extraspecial 22-group GG of order mm and suppose that the result is true for all extraspecial 22-group of order less than mm. By [15, 5.3.8], GG contains subgroups AA and BB such that AA and BB are extraspecial 22-groups of order less than mm and G=ABG=AB is a central product. Hence, the elements of AA commute with the elements of BB and AB=Z(G)=Z(A)=Z(B)A\cap B=Z(G)=Z(A)=Z(B). By induction hypothesis, the following two conditions are satisfied:

(i) CA(a1)CA(a2)C_{A}(a_{1})\neq C_{A}(a_{2}), for all a1,a2Aa_{1},a_{2}\in A such that a1a2Z(A)a_{1}\notin a_{2}Z(A).

(ii) CB(b1)CB(b2)C_{B}(b_{1})\neq C_{B}(b_{2}), for all b1,b2Bb_{1},b_{2}\in B such that b1b2Z(B)b_{1}\notin b_{2}Z(B).

Let x=a1b1x=a_{1}b_{1} and y=a2b2y=a_{2}b_{2} be non-central elements of GG such that xyZ(G)x\notin yZ(G), where a1,a2Aa_{1},a_{2}\in A and b1,b2Bb_{1},b_{2}\in B. Thus, in virtue of (i) and (ii), we must have CA(a1)CA(a2)C_{A}(a_{1})\neq C_{A}(a_{2}) or CB(b1)CB(b2)C_{B}(b_{1})\neq C_{B}(b_{2}). In the first situation, there is wCA(a1)CA(a2)w\in C_{A}(a_{1})\setminus C_{A}(a_{2}) (or there is uCA(a2)CA(a1)u\in C_{A}(a_{2})\setminus C_{A}(a_{1})) and it follows that wCG(x)CG(y)w\in C_{G}(x)\setminus C_{G}(y) (or uCG(y)CG(x)u\in C_{G}(y)\setminus C_{G}(x)), which gives us CG(x)CG(y)C_{G}(x)\neq C_{G}(y). Analogously, CB(b1)CB(b2)C_{B}(b_{1})\neq C_{B}(b_{2}) implies CG(x)CG(y)C_{G}(x)\neq C_{G}(y), as required. ∎

The following result is taken from [10].

Lemma 3.2.

([10, Lemma 10.1.1] )If 𝒢{\cal G} is a strongly regular graph with parameters (v,k,λ,μ)(v,k,\lambda,\mu), then the following statements are equivalent:

(i)

𝒢{\cal G} is disconnected;

(ii)

μ=0\mu=0;

(iii)

λ=k1\lambda=k-1;

(iv)

𝒢{\cal G} is isomorphic to mKk+1mK_{k+1} for some integer m>1m>1.

Proposition 3.3.

If GG is an extraspecial 22-group of order 22n+12^{2n+1}, with n2n\geq 2, then 𝒯(G){\cal T}(G) is a connected strongly regular graph with parameters (22n1,22n12,22n23,22n21)(2^{2n}-1,2^{2n-1}-2,2^{2n-2}-3,2^{2n-2}-1).

Proof.

Let x,yGZ(G)x,y\in G\setminus Z(G) be such that xyZ(G)x\notin yZ(G). By Lemma 3.1, we have CG(x)CG(y)C_{G}(x)\neq C_{G}(y) and, by Theorem 2.7, GG is of conjugate type {1,2}\{1,2\}. This shows that [CG(x):CG(x)CG(y)]2[C_{G}(x):C_{G}(x)\cap C_{G}(y)]\geq 2. Write G={1,d}G^{\prime}=\{1,d\}. Given g,hCG(x)CG(y)g,h\in C_{G}(x)\setminus C_{G}(y), we have gy=ygdgy=ygd, h1y=yh1dh^{-1}y=yh^{-1}d and so (gh1)y=y(gh1)(gh^{-1})y=y(gh^{-1}), that is, gh1CG(x)CG(y)gh^{-1}\in C_{G}(x)\cap C_{G}(y). Hence [CG(x):CG(x)CG(y)]=2[C_{G}(x):C_{G}(x)\cap C_{G}(y)]=2. Applying Proposition 2.5 to calculate the degree of each vertex of 𝒯(G){\cal T}(G) and the size of the common neighborhood of two vertices, we conclude that 𝒯(G){\cal T}(G) is strongly regular with parameters (22n1,22n12,22n23,22n21)(2^{2n}-1,2^{2n-1}-2,2^{2n-2}-3,2^{2n-2}-1). Since 22n2102^{2n-2}-1\neq 0, Lemma 3.2 give us 𝒯(G){\cal T}(G) is connected. ∎

The next result establishes a necessary condition for a graph 𝒯(G){\cal T}(G) to be connected strongly regular.

Lemma 3.4.

Let GG be a finite non-abelian group such that 𝒯(G){\cal T}(G) is connected strongly regular. Then, for all x,yGZ(G)x,y\in G\setminus Z(G) such that xyZ(G)x\notin yZ(G), we have CG(x)CG(y)C_{G}(x)\neq C_{G}(y).

Proof.

Suppose there is a non-abelian group GG such that 𝒯(G){\cal T}(G) is a connected strongly regular graph with parameters (v,k,λ,μ)(v,k,\lambda,\mu) and that there are x,yGZ(G)x,y\in G\setminus Z(G) such that xyZ(G)x\notin yZ(G) and CG(x)=CG(y)C_{G}(x)=C_{G}(y). For all gxZ(G)g\in xZ(G) and hyZ(G)h\in yZ(G) we have [g,h]=[x,y]=1,[g,h]=[x,y]=1, CG(g)=CG(x)C_{G}(g)=C_{G}(x) and CG(h)=CG(y)C_{G}(h)=C_{G}(y). Hence, we may assume that x,yx,y are vertices of 𝒯(G){\cal T}(G). From Proposition 2.5, it follows that if n=[G:Z(G)]n=[G:Z(G)] and m=[CG(x):Z(G)]m=[C_{G}(x):Z(G)], then v=n1v=n-1, k=m2k=m-2 and λ=m3=k1\lambda=m-3=k-1. By Lemma 3.2 the graph 𝒯(G){\cal T}(G) is disconnected, a contradiction. ∎

Our intention is to characterize the groups GG for which 𝒯(G){\cal T}(G) is connected strongly regular . As we deal with a case where the graph 𝒯(G){\cal T}(G) is regular (and so GG is of conjugate type {1,m}\{1,m\}), in virtue of Theorem 2.6, we can restrict our study to pp-groups.

Lemma 3.5.

Given an odd prime pp let GG be a finite non-abelian pp-group. If 𝒯(G){\cal T}(G) is strongly regular, then 𝒯(G){\cal T}(G) is disconnected.

Proof.

Suppose that 𝒯(G){\cal T}(G) is strongly regular. Given xGZ(G)x\in G\setminus Z(G), we have x2xZ(G)x^{2}\notin xZ(G); further, as CG(x)CG(x2)C_{G}(x)\subseteq C_{G}(x^{2}) and deg(x)(x)=deg(x2)(x^{2}) we get CG(x)=CG(x2)C_{G}(x)=C_{G}(x^{2}). Hence, by Lemma 3.4, the graph 𝒯(G){\cal T}(G) is disconnected, as desired. ∎

Lemma 3.5 say us that if GG is a pp-group such that 𝒯(G){\cal T}(G) is a connected strongly regular graph, then p=2p=2.

For the proof of the next two results we will need the following relation, that holds for any strongly regular graph with parameters (v,k,λ,μ)(v,k,\lambda,\mu) (see [5, Proposition 2.6]):

μ(vk1)=k(kλ1).\mu(v-k-1)=k(k-\lambda-1). (1)
Lemma 3.6.

Let GG be a finite non-abelian 22-group such that 𝒯(G){\cal T}(G) is a connected strongly regular graph with parameters (v,k,λ,μ)(v,k,\lambda,\mu). Suppose that |CG(x)CG(y)|=|CG(t)CG(w)||C_{G}(x)\cap C_{G}(y)|=|C_{G}(t)\cap C_{G}(w)|, for any vertices x,y,t,wx,y,t,w of 𝒯(G){\cal T}(G), with xyx\neq y and twt\neq w. Then GG is isoclinic to an extraspecial 22-group of order at least 3232.

Proof.

From the hypotheses it follows that there exist non-negative integers m,n,rm,n,r such that [G:Z(G)]=2m+n+r[G:Z(G)]=2^{m+n+r}, [CG(x):Z(G)]=2m+n[C_{G}(x):Z(G)]=2^{m+n} and [CG(x)CG(y):Z(G)]=2m[C_{G}(x)\cap C_{G}(y):Z(G)]=2^{m}, for any distinct vertices xx and yy of 𝒯(G){\cal T}(G). Thus, Proposition 2.5 provides us

v=2m+n+r1,k=2m+n2,λ=2m3,μ=2m1.v=2^{m+n+r}-1,\qquad k=2^{m+n}-2,\qquad\lambda=2^{m}-3,\qquad\mu=2^{m}-1.

Because λ0\lambda\geq 0, we must have m2m\geq 2. From (1) we obtain (2n+r2n)(2m1)=(2n1)(2m+n2)(2^{n+r}-2^{n})(2^{m}-1)=(2^{n}-1)(2^{m+n}-2) and, consequently, 2n(2m+r2r+12m+n+2)=2.2^{n}(2^{m+r}-2^{r}+1-2^{m+n}+2)=2. Because μk\mu\leq k we get n1n\geq 1 and from the last equality above we conclude n=1n=1. Thus 2m+r2r+12m+1+2=12^{m+r}-2^{r}+1-2^{m+1}+2=1 and so 2m(2r2)=2r22^{m}(2^{r}-2)=2^{r}-2, which implies r=1r=1. Hence, [G:CG(x)]=2[G:C_{G}(x)]=2, for any xGZ(G)x\in G\setminus Z(G), that is, GG is of conjugate type {1,2}\{1,2\}. Now, from Theorem 2.7 it follows that GG is isoclinic to an extraspecial 22-group EE. As m2m\geq 2 and n=r=1n=r=1 we obtain [E:Z(E)]=[G:Z(G)]16[E:Z(E)]=[G:Z(G)]\geq 16 and so EE has order at least 3232. ∎

Theorem 3.7.

If GG is a finite non-abelian 22-group such that 𝒯(G){\cal T}(G) is connected strongly regular, then GG is isoclinic to an extraspecial 22-group of order at least 3232.

Proof.

By Lemma 3.6, it is sufficient to show that |CG(x)CG(y)|=|CG(t)CG(w)||C_{G}(x)\cap C_{G}(y)|=|C_{G}(t)\cap C_{G}(w)|, for any vertices x,y,t,wx,y,t,w of 𝒯(G){\cal T}(G), where xyx\neq y and twt\neq w. Let us assume the contrary, that is, there are vertices x,y,t,wx,y,t,w of 𝒯(G){\cal T}(G) such that

|CG(x)CG(y)|>|CG(t)CG(w)|.|C_{G}(x)\cap C_{G}(y)|>|C_{G}(t)\cap C_{G}(w)|. (2)

Taking into account Proposition 2.5 and the fact that 𝒯(G){\cal T}(G) is strongly regular, we have two cases to analyze:

Case 1: [x,y]1[x,y]\neq 1 and [t,w]=1[t,w]=1. Due to Proposition 2.5 and (2), we can write

v=2m+n+r+s1k=2m+n+r2λ=2m3,μ=2m+n1,v=2^{m+n+r+s}-1\qquad k=2^{m+n+r}-2\qquad\lambda=2^{m}-3,\qquad\mu=2^{m+n}-1,

where m,n,r,sm,n,r,s are non-negative integers, with n1n\geq 1. From μ(vk1)=k(kλ1)\mu(v-k-1)=k(k-\lambda-1) we obtain

(2m+n1)(2n+r+s2n+r)=(2m+n+r2)(2n+r1),(2^{m+n}-1)(2^{n+r+s}-2^{n+r})=(2^{m+n+r}-2)(2^{n+r}-1),

which is equivalent to

2n(2m+n+r+s2m+n+r2r+s+2r2m+n+2r+2m+r+2r+1)=2.2^{n}(2^{m+n+r+s}-2^{m+n+r}-2^{r+s}+2^{r}-2^{m+n+2r}+2^{m+r}+2^{r+1})=2.

Since n1n\geq 1, this forces n=1n=1 and

2m+n+r+s2m+n+r2r+s+2r2m+n+2r+2m+r+2r+1=1,2^{m+n+r+s}-2^{m+n+r}-2^{r+s}+2^{r}-2^{m+n+2r}+2^{m+r}+2^{r+1}=1,

that is, 2r(2m+s+12m+12s+12m+r+1+2m+2)=12^{r}(2^{m+s+1}-2^{m+1}-2^{s}+1-2^{m+r+1}+2^{m}+2)=1. For such equality to occur, we must have r=0r=0, which give us k<μk<\mu, a contradiction.

Case 2: [x,y]=1[x,y]=1 and [t,w]1[t,w]\neq 1. Considering (2), we may write

v=2m+n+r+s1,k=2m+n+r2,λ=2m+n3,μ=2m1,v=2^{m+n+r+s}-1,\qquad k=2^{m+n+r}-2,\qquad\lambda=2^{m+n}-3,\qquad\mu=2^{m}-1,

where m,n,r,sm,n,r,s are non-negative integers, with n1n\geq 1. Note that s,m0s,m\neq 0 because kv1k\neq v-1 and μ0\mu\neq 0. From the identity μ(vk1)=k(kλ1)\mu(v-k-1)=k(k-\lambda-1) we get (2m1)(2r+s2r)=(2m+n+r2)(2r1)(2^{m}-1)(2^{r+s}-2^{r})=(2^{m+n+r}-2)(2^{r}-1) and so 2r(2m+s2m2s+12m+n+r+2m+n+2)=22^{r}(2^{m+s}-2^{m}-2^{s}+1-2^{m+n+r}+2^{m+n}+2)=2, which implies r{0,1}r\in\{0,1\}. However, by Lemma 3.2, λk1\lambda\neq k-1; thus r=1r=1 and, consequently,

2m+s2m2s+12m+n+1+2m+n+2=1.2^{m+s}-2^{m}-2^{s}+1-2^{m+n+1}+2^{m+n}+2=1.

It follows from the last equality that 2m(2s12n+1+2n)=2(2s11)2^{m}(2^{s}-1-2^{n+1}+2^{n})=2(2^{s-1}-1).

If s=1s=1, we obtain 12n+1+2n=01-2^{n+1}+2^{n}=0, that is, n=0n=0, a contradiction.

Suppose now s2s\geq 2. Since m1m\geq 1, we conclude that the number 2m1(2s12n+1+2n)=2s112^{m-1}(2^{s}-1-2^{n+1}+2^{n})=2^{s-1}-1 is odd and this implies m=1m=1. From the penultimate equality we get 2s12n+1+2n=2s112^{s}-1-2^{n+1}+2^{n}=2^{s-1}-1, which gives us 2s1=2n2^{s-1}=2^{n}, that is, s1=ns-1=n. Hence, in this case, the parameters of 𝒯(G){\cal T}(G) are:

v=22s+11,k=2s+12,λ=2s3,μ=1.v=2^{2s+1}-1,\qquad k=2^{s+1}-2,\qquad\lambda=2^{s}-3,\qquad\mu=1. (3)

Putting γ=(μλ)2+4(kμ)\gamma=(\mu-\lambda)^{2}+4(k-\mu), it follows from ([5], Theorem 2.16) that the numbers m1m_{1} and m2m_{2} given below are non-negative integers:

m1=12(v1+2k+(v1)(μλ)γ),m2=12(v12k+(v1)(μλ)γ).\displaystyle{m_{1}=\frac{1}{2}\left(v-1+\frac{2k+(v-1)(\mu-\lambda)}{\sqrt{\gamma}}\right)},\qquad\displaystyle{m_{2}=\frac{1}{2}\left(v-1-\frac{2k+(v-1)(\mu-\lambda)}{\sqrt{\gamma}}\right)}.

Note that 2k+(v1)(μλ)=2s+1(2s+222s+3)122k+(v-1)(\mu-\lambda)=2^{s+1}(2^{s+2}-2^{2s}+3)-12 and this number is nonzero for s2s\geq 2.

From the parameters obtained in (3), we obtain γ=(42s)2+4(2s+13)=4(4s1+1)\gamma=(4-2^{s})^{2}+4(2^{s+1}-3)=4(4^{s-1}+1) and, thus, γ=24s1+1\sqrt{\gamma}=2\sqrt{4^{s-1}+1}. Since m1m_{1} and m2m_{2} are non-negative integers, there is aa\in\mathbb{Q} such that 4s1=a214^{s-1}=a^{2}-1. Then aa must be an odd integer, say a=2l+1a=2l+1, and we get 4s1=4l(l+1)4^{s-1}=4l(l+1), a contradiction.

Now, the result follows from Lemma 3.6. ∎

We are now ready to show Theorem A.

Proof of Theorem A.

Let GG be a finite non-abelian group such that 𝒯(G){\cal T}(G) is connected strongly regular. Theorem 2.6 tells us that GG admits a decomposition G=A×PG=A\times P, where AA is an abelian subgroup of GG and PP is a Sylow pp-subgroup of GG of conjugate type {1,pr}\{1,p^{r}\} for some positive integer rr. Now [3, page 286] ensures us that GG and PP are isoclinic and so 𝒯(G)𝒯(P){\cal T}(G)\cong{\cal T}(P). Lemma 3.5 forces p=2p=2 and Theorem 3.7 yields us that PP is isoclinic to an extraspecial 22-group EE of order at least 3232, that is, |E|=22n+1|E|=2^{2n+1}, for some n2n\geq 2. Now, the result follows from Propositions 3.3 and 2.3, taking into account that GG and EE are isoclinic groups. ∎

As an immediate consequence of the proof of Theorem A, we obtain

Corollary 3.8.

Let GG be a finite non-abelian group. The graph 𝒯(G){\cal T}(G) is connected strongly regular if and only if GG is isoclinic to an extraspecial 22-group of order at least 3232.

We also characterize the finite non-abelian groups GG for which the graph 𝒯(G){\cal T}(G) is a disjoint union of complete graphs of the same size.

Proposition 3.9.

Let GG be a finite non-abelian group and let m,nm,n be positive integers. The graph 𝒯(G){\cal T}(G) is a disjoint union of mm copies of KnK_{n} if and only if mn=[G:Z(G)]1mn=[G:Z(G)]-1 and for any xGZ(G)x\in G\setminus Z(G) we have CG(x)C_{G}(x) is abelian and [CG(x):Z(G)]=n+1[C_{G}(x):Z(G)]=n+1.

Proof.

Suppose that 𝒯(G){\cal T}(G) is a disjoint union of mm copies of KnK_{n}. It is clear that mn=[G:Z(G)]1mn=[G:Z(G)]-1. Let T={1,x1,,xmn}T=\{1,x_{1},\ldots,x_{mn}\} be the transversal of Z(G)Z(G) in GG such that V(𝒯(G))=T{1}V({\cal T}(G))=T\setminus\{1\}. Given xGZ(G)x\in G\setminus Z(G), there is i{1,,mn}i\in\{1,\ldots,mn\} such that xxiZ(G)x\in x_{i}Z(G). If N(xi)={y1,,yn1}N(x_{i})=\{y_{1},\ldots,y_{n-1}\} is the neighborhood of xix_{i} in 𝒯(G){\cal T}(G), then deg(xi)=n1(x_{i})=n-1. Since deg(xi)=[CG(xi):Z(G)]2(x_{i})=[C_{G}(x_{i}):Z(G)]-2 (by Proposition 2.5), we have [CG(xi):Z(G)]=n+1[C_{G}(x_{i}):Z(G)]=n+1. Moreover, CG(xi)=Z(G)xiZ(G)(yN(xi)yZ(G))C_{G}(x_{i})=Z(G)\cup x_{i}Z(G)\cup\left(\bigcup_{y\in N(x_{i})}yZ(G)\right) and CG(xi)C_{G}(x_{i}) is abelian, because the subgraph generated by {xi}N(xi)\{x_{i}\}\cup N(x_{i}) is complete. Since CG(xi)=CG(x)C_{G}(x_{i})=C_{G}(x), we conclude that [CG(x):Z(G)]=n+1[C_{G}(x):Z(G)]=n+1 and CG(x)C_{G}(x) is abelian.

Conversely, assume that [G:Z(G)]1=mn[G:Z(G)]-1=mn and for any xGZ(G)x\in G\setminus Z(G), CG(x)C_{G}(x) is abelian and [CG(x):Z(G)]=n+1[C_{G}(x):Z(G)]=n+1. Let xx be a vertex of 𝒯(G){\cal T}(G). If N(x)={y1,,yn1}N(x)=\{y_{1},\ldots,y_{n-1}\} is the neighborhood of xx in 𝒯(G){\cal T}(G), then the subgraph generated by {x}N(x)\{x\}\cup N(x) is isomorphic to KnK_{n}, because CG(x)=Z(G)xZ(G)(yN(x)yZ(G))C_{G}(x)=Z(G)\cup xZ(G)\cup\left(\bigcup_{y\in N(x)}yZ(G)\right) and CG(x)C_{G}(x) is abelian. In addiction, as the non-central elements of GG have centralizers of equal size, 𝒯(G){\cal T}(G) is a disjoint union of mm graphs isomorphic to KnK_{n}. ∎

It is worth mentioning that when GG is a non-abelian pp-group, the case n=1n=1 in Proposition 3.9 is also a consequence of [1, Corollary 1]. Further, if GG is a group such that 𝒯(G)mKn{\cal T}(G)\cong mK_{n}, the result above produces m3m\geq 3.

We have already seen ([10, Lemma 10.1.1]) that if a graph 𝒢{\cal G} is disconnected strongly regular, then it is isomorphic to mKnmK_{n} for some integers m,n2m,n\geq 2. Of course, the reciprocal also holds. Thus, as an immediate consequence from Proposition 3.9, we obtain

Corollary 3.10.

Let GG be a finite non-abelian group. The graph 𝒯(G){\cal T}(G) is disconnected strongly regular if and only if the centralizers of the non-central elements of GG are all abelian and there exists an integer r3r\geq 3 such that [CG(x):Z(G)]=r[C_{G}(x):Z(G)]=r, for all xGZ(G)x\in G\setminus Z(G).

Akbari and Moghaddamfar [1, Corollary 2(a)] proved that a commuting graph Γ(G)\Gamma(G) is strongly regular if and only if it is a disjoint union of complete graphs on |Z(G)||Z(G)| vertices. However, Example 2.1 shows that the connected components of a strongly regular graph Γ(G)\Gamma(G) do not necessarily have size equal to |Z(G)||Z(G)|. In fact, we have the following result, which corrects the error in [1, Corollary 2(a)]:

Proposition 3.11.

Let GG be a finite non-abelian group. The graph Γ(G)\Gamma(G) is strongly regular if and only if Γ(G)mKs\Gamma(G)\cong mK_{s}, where ms=|GZ(G)|ms=|G\setminus Z(G)| and s=([CG(x):Z(G)]1)|Z(G)|s=([C_{G}(x):Z(G)]-1)|Z(G)|, for all xGZ(G)x\in G\setminus Z(G).

Proof.

Suppose that Γ(G)\Gamma(G) is a strongly regular graph with parameters (v,k,λ,μ)(v,k,\lambda,\mu). As Γ(G)\Gamma(G) is regular, from Theorem 2.6 it follows |Z(G)|>1|Z(G)|>1. For all xGZ(G)x\in G\setminus Z(G) and z1,z2Z(G)z_{1},z_{2}\in Z(G), with z1z2z_{1}\neq z_{2}, it is clear that CG(xz1)=CG(xz2)C_{G}(xz_{1})=C_{G}(xz_{2}) and [xz1,xz2]=1[xz_{1},xz_{2}]=1. Thus, we must have λ=k1\lambda=k-1. As the graph Γ(G)\Gamma(G) is not complete, from the relation μ(vk1)=k(kλ1)\mu(v-k-1)=k(k-\lambda-1), we obtain μ=0\mu=0. Hence, Lemma 3.2 ensures us Γ(G)mKs\Gamma(G)\cong mK_{s}, for some integers m,s2m,s\geq 2. Clearly, ms=|GZ(G)|ms=|G\setminus Z(G)| and by calculating the degree of each vertex we get s=|CG(x)||Z(G)|=([CG(x):Z(G)]1)|Z(G)|s=|C_{G}(x)|-|Z(G)|=([C_{G}(x):Z(G)]-1)|Z(G)|, for all xGZ(G)x\in G\setminus Z(G). ∎

Let GG be a finite non-abelian group. Since for x,yGx,y\in G we have [x,y]=1[x,y]=1 if and only if [g,h]=1[g,h]=1, for all gxZ(G)g\in xZ(G), hyZ(G)h\in yZ(G), as a consequence of Proposition 3.11 we conclude that if Γ(G)\Gamma(G) is a strongly regular graph, then each connected component of this graph is an union of nn non-trivial cosets of Z(G)Z(G), where n=[CG(x):Z(G)]1n=[C_{G}(x):Z(G)]-1.

Let pp be a prime and let PP be an extraspecial pp-group. If pp is odd and |P|>p3|P|>p^{3}, Theorem 2.7 and Lemma 3.5 ensure us that 𝒯(P){\cal T}(P) is regular but not strongly regular. The next result tells us that if GG is a group isoclinic to an extraspecial pp-group of order p3p^{3}, then 𝒯(G){\cal T}(G) is disconnected strongly regular.

Proposition 3.12.

Let GG be a finite non-abelian group and let pp be a prime number. We have 𝒯(G)(p+1)Kp1{\cal T}(G)\cong(p+1)K_{p-1} if and only if GG is isoclinic to an extraspecial pp-group of order p3p^{3}.

Proof.

Let GG be a group such that 𝒯(G)(p+1)Kp1{\cal T}(G)\cong(p+1)K_{p-1}. Hence [G:Z(G)]=p2[G:Z(G)]=p^{2} and as the graph 𝒯(G){\cal T}(G) is regular, by Lemma 2.6 and Proposition 2.3, we may assume that GG is a pp-group. Now Proposition 3.9 provides us [CG(x):Z(G)]=p[C_{G}(x):Z(G)]=p, for all xGZ(G)x\in G\setminus Z(G). Thus we conclude [G:CG(x)]=p[G:C_{G}(x)]=p, for all xGZ(G)x\in G\setminus Z(G) and, consequently, GG is of conjugate type {1,p}\{1,p\}. From Theorem 2.7 it follows that GG is isoclinic to an extraspecial pp-group, say PP. Since [P:Z(P)]=[G:Z(G)]=p2[P:Z(P)]=[G:Z(G)]=p^{2} and |Z(P)|=p|Z(P)|=p, we must have |P|=p3|P|=p^{3}. The converse follows from Proposition 3.9 and Proposition 2.3. ∎

References

  • [1] M. Akbari, A. R. Moghaddamfar, The Existence or Nonexistence of Non-Commuting Graphs with Particular Properties. J. Algebra Appl. 13 (2014) 1350064, 11 pp.
  • [2] C. Bates, D. Bundy, S. Hart and P. Rowley, Commuting involution graphs for sporadic simple groups, J. Algebra 316 (2007), 849-868.
  • [3] Y. Berkovich, Groups of Prime Power Order, Walter de Gruyter, vol. 1, Berlin/New York (2008).
  • [4] R. Brauer and K. A. Fowler, On groups of even order. Ann. of Math. (2) 62 (1955), 567-583.
  • [5] P. J. Cameron and J. H. van Lint, Designs, Graphs, Codes and Their Links. Cambridge University Press, 1991.
  • [6] G. Chartrand and P. Zhang, A First Course in Graph Theory, Dover Publications (2012).
  • [7] B. Fischer, Finite groups generated by 3-transpositions, J. Invent. Math. 13 (1971), 232-246.
  • [8] M. Giudici and C. Parker, There is no upper bound for the diameter of the commuting graph of a finite group, J. Combin. Theory Ser. A 120 (2013), 1600-1603.
  • [9] M. Giudici and A. Pope, On bounding the diameter of the commuting graph of a group, J. Group Theory 17 (2014), 131-149.
  • [10] C. Godsil, G. Royle, Algebraic Graph Theory, Graduate Texts in Math. 207 Springer-Verlag (2001).
  • [11] P. Hall, The classification of prime-power groups. J. Reine Angew. Math. 182 (1940), 130-141.
  • [12] K. Ishikawa, Finite p-groups up to isoclinism, which have only two conjugacy lengths. J. Algebra 220 (1999), 333-345.
  • [13] N. It , On finite groups with given conjugate types I. Nagoya Math. 6 (1953), 17-28.
  • [14] G. L. Morgan and C. W. Parker, The diameter of the commuting graph of a finite group with trivial center, J. Algebra 393(1) (2013), 41-59.
  • [15] D. J. S. Robinson, A course in the theory of groups, 2nd edition, Springer-Verlag, New York, 1996.
  • [16] S. Perkins, Commuting involution graphs for n, Arch. Math. 86 (2006), 16-25.
  • [17] J. Vahidi and A. A. Talebi, The commuting graphs on groups D2nD_{2n} and QnQ_{n}, J. Math. Comput. Sci. 1 (2010), 123-127.