betaFeedback

WorksMathematics

Generalized Power Graph of a Group

Abstract

In this paper, we introduce a new class of graphs, called generalized power graphs, and investigate some of their key properties, such as completeness, connectivity, and planarity.

Introduction

The concepts of directed and undirected power graphs of a group were introduced by Kelarev and Quinn [16] and Chakrabarty et al. [8], respectively, and have since attracted considerable interest (see for instance [1, 2, 9, 10, 17, 19]). For the reader’s convenience, we shortly define these two types of graphs. The directed power graph P⃗(G)\vec{\mathcal{P}}(G) of a group GG is the directed graph with vertex set GG, in which there is a directed edge from xx to yy if x≠yx \ne y and y=xny=x^{n} for some positive integer nn. The undirected power graph P(G)\mathcal{P}(G) of GG is the undirected graph obtained from P⃗(G)\vec{\mathcal{P}}(G) by ignoring edge directions. The proper power graph P∗(G)\mathcal{P}^{*}(G) of GG is the undirected power graph without identity element of a group GG.

An element of a group GG is called a generalized torsion element if a finite product of its conjugates equals the identity. We say that GG is a generalized torsion group if all of its elements have this property. It is immediate that any torsion element is a generalized torsion element. Although the converse holds for many important classes of groups (cf. [3, 4, 5, 15]), there exist infinite groups containing non-torsion elements that are nevertheless generalized torsion. For instance, consider the non-torsion elements of the infinite dihedral group.

Investigating generalized torsion elements provides important insights into combinatorial group theory and topology because they serve as a primary obstruction to bi-orderability. Indeed, if a group admits a bi-invariant order, then any finite product of conjugates of a positive element remains strictly positive. Consequently, every bi-orderable group is necessarily free of generalized torsion. This connection has motivated the study of generalized torsion and has established it as an active area of research (cf. [3, 4, 5, 12, 13, 14, 15]).

Given that graphs constructed from algebraic objects often reflect the structural properties of the underlying system, following [8, 16], we introduce the following classes of graphs. The generalized directed power graph P⃗∙(G)\vec{\mathcal{P}}_{\bullet}(G) of a group GG is the directed graph with vertex set GG, where for any x,y∈Gx,y\in G, there is a directed edge from xx to yy (denoted x→yx\to y or y←xy\leftarrow x) if x≠yx\ne y and yy can be expressed as a finite product of conjugates of xx. That is, x→yx\to y if there exist g1,…,gk∈Gg_{1},\ldots,g_{k}\in G such that

y=xg1xg2⋯xgk,y=x^{g_{1}}x^{g_{2}}\cdots x^{g_{k}},

where xg=g−1xgx^{g}=g^{-1}xg denotes the conjugate of xx by gg. The generalized undirected power graph (or generalized power graph, for short) P∙(G)\mathcal{P}_{\bullet}(G) of a group GG is the undirected graph obtained from P⃗∙(G)\vec{\mathcal{P}}_{\bullet}(G) by ignoring edge directions. The proper generalized directed (resp., undirected) power graph of GG, denoted by P⃗∙∗(G)\vec{\mathcal{P}}_{\bullet}^{*}(G) (resp., P∙∗(G)\mathcal{P}_{\bullet}^{*}(G)), is the induced subgraph obtained from P⃗∙(G)\vec{\mathcal{P}}_{\bullet}(G) (resp., P∙(G)\mathcal{P}_{\bullet}(G)) by deleting the vertex corresponding to the identity element.

2020 Mathematics Subject Classification. 05C25; 05C40; 05C10; 05C12; 20D15; 20F50; 20E45.

Key words and phrases. Power graph; Generalized power graph; generalized torsion; complete graph; connected graph; planar graph.

Let S3S_3 denote the symmetric group on three letters. The image below illustrates the generalized directed power graph, the proper generalized directed power graph, and the proper generalized undirected power graph of S3S_3, respectively.

Generalized directed power graph, proper generalized directed power graph, and proper generalized undirected power graph of $S_3$

Figure 1.

Given that every non-trivial element is adjacent to the identity in the generalized power graph of a generalized torsion group, our results are mainly devoted to the proper generalized power graph.

The paper is organized as follows. In Section 2, we present general results regarding generalized power graphs. Sections 3 and 4 are devoted to investigating the completeness and connectedness of the generalized power graph of some families of finite and infinite groups. Finally, in Section 5, we give necessary and sufficient conditions for the planarity of the proper generalized power graph of a finite group.

General Properties

Let xx and yy be elements in a group GG. We say that there exists a path connecting xx and yy if there exist elements x=z1,z2,…,zn=y∈Gx = z_1, z_2, \ldots, z_n = y \in G such that ziz_i and zi+1z_{i+1} are adjacent for all i=1,…,n−1i = 1, \ldots, n - 1.

For non-empty subsets X,Y⊆GX,Y \subseteq G, we write X→YX \to Y if x→yx \to y for all x∈Xx \in X and y∈Yy \in Y. We may also write X↔YX \leftrightarrow Y to mean that X→YX \to Y and Y→XY \to X.

We denote the conjugacy class of xx in GG by xG={xg∣g∈G}x^G = \{x^g \mid g \in G\}, ∣x∣|x| the order of xx and its centralizer by CG(x)C_G(x). Furthermore, CnC_n, D2nD_{2n}, Q8Q_8, KnK_n, and K3,3K_{3,3} denote the cyclic group of order nn, the dihedral group of order 2n2n, the quaternion group of order 8, the complete graph on nn vertices, and the complete bipartite graph, respectively.

We define commutators of weight 1,2,…1,2,\ldots in the variables x1,x2,…x_1,x_2,\ldots recursively as formal bracket expressions. The variables x1,x2,…x_1,x_2,\ldots are commutators of weight 1, and commutators of weight 2 take the form [xi,xj]=xi−1xj−1xixj[x_i,x_j] = x_i^{-1}x_j^{-1}x_ix_j. If w1w_1 and w2w_2 are commutators of weight k1k_1 and k2k_2, respectively, then [w1,w2][w_1,w_2] is a commutator of weight k1+k2k_1+k_2. If ww is a commutator of weight kk, we view it primarily as a function of kk variables defined on a group GG. Any value of ww in GG is called a commutator in GG (of weight kk).

Throughout this paper, the following properties may be used without explicit reference.

Lemma 2.1. Let x,y,zx,y,z elements of a group GG.

(1) If x→y→zx \to y \to z, then x→zx \to z.

(2) If x→yx \to y and x→zx \to z, then x→yzx \to yz.

(3) xG↔xGx^G \leftrightarrow x^G.

(4) If xg→yhx^g \to y^h for some g,h∈Gg,h \in G, then xG→yGx^G \to y^G.

If additionally xx is a torsion element, we obtain the following results:

(5) If y≠1y \ne1 and y∈⟨xG⟩y \in\langle x^G\rangle, then xG→yGx^G \to y^G.

(6) If (∣x∣,i)=1(|x|,i)=1, then xi→xjx^i \to x^j for every jj.

(7) x↔x−1x \leftrightarrow x^{-1} and y→(xy)∣x∣y \to(xy)^{|x|}.

(8) If xx is any entry of some commutator ww, then x→wx \to w.

Proof. (1) There exist elements g1,…,gm,h1,…,hn∈Gg_1,\ldots,g_m,h_1,\ldots,h_n \in G such that y=xg1⋯xgmy=x^{g_1}\cdots x^{g_m} and z=yh1⋯yhnz=y^{h_1}\cdots y^{h_n}. Hence,

z=∏j=1n(xg1⋯xgm)hj=∏j=1n∏i=1mxgihj,z=\prod_{j=1}^{n}(x^{g_1}\cdots x^{g_m})^{h_j}=\prod_{j=1}^{n}\prod_{i=1}^{m}x^{g_i h_j},

showing that x→zx \to z.

(2) There exist elements g1,…,gm,h1,…,hn∈Gg_1,\ldots,g_m,h_1,\ldots,h_n \in G such that y=xg1⋯xgmy=x^{g_1}\cdots x^{g_m} and z=xh1⋯xhnz=x^{h_1}\cdots x^{h_n}. Hence, yz=xg1⋯xgmxh1⋯xhnyz=x^{g_1}\cdots x^{g_m}x^{h_1}\cdots x^{h_n}, showing that x→yzx \to yz.

(3) Let z1,z2∈xGz_1,z_2 \in x^G. There exist g,h∈Gg,h \in G such that z1=xgz_1=x^g and z2=xhz_2=x^h. Thus, z1=(xh)h−1g=z2h−1gz_1=(x^h)^{h^{-1}g}=z_2^{h^{-1}g}, which means z2→z1z_2 \to z_1. By a similar argument, we obtain z1→z2z_1 \to z_2.

(4) Note that xg→yhx^g \to y^h implies x→yx \to y. Applying Item (3), we obtain xG→{x}→{y}→yGx^G \to\{x\} \to\{y\} \to y^G, and the result follows from Item (1).

(5) Since xx is torsion and y∈⟨xG⟩y \in\langle x^G\rangle, we have x→yx \to y. Thus, the result follows by Item (4).

(6) By Bézout’s identity, there exist integers mm and nn such that mi+n∣x∣=1mi+n|x|=1. Thus, x=xmi+n∣x∣=xmix=x^{mi+n|x|}=x^{mi}, and so xj=(xmi)j=(xi)mjx^j=(x^{mi})^j=(x^i)^{mj}. This implies that xi→xjx^i \to x^j.

(7) If xx has finite order nn, then x−1x^{-1} also has order nn and, consequently, xn−1=x−1x^{n-1}=x^{-1} and (x−1)n−1=x(x^{-1})^{n-1}=x.

The second part follows from the fact that, for any integer n≥2n \ge2, we have (xy)n=xnyxn−1yxn−2⋯yxy(xy)^n=x^n y^{x^{n-1}}y^{x^{n-2}}\cdots y^x y. Setting ∣x∣=n|x|=n, we have (xy)n=yxn−1yxn−2⋯yxy(xy)^n=yx^{n-1}y^{x^{n-2}}\cdots y^x y, which means y→(xy)ny \to(xy)^n.

(8) We proceed by induction on the length of the commutator ww. Let kk be the order of xx. If ww has length 22, then w=[x,y]w=[x,y] or w=[y,x]w=[y,x]. In the first case, we have w=x−1y−1xy=xk−1xyw=x^{-1}y^{-1}xy=x^{k-1}xy, and thus x→wx \to w. The second case is similar. Now assume the result holds for any commutator of length less than tt, and let ww be a commutator of length tt. Then w=[w1,w2]w=[w_1,w_2], where w1w_1 and w2w_2 are commutators of lengths t1t_1 and t2t_2, respectively, satisfying t1+t2=tt_1+t_2=t. Furthermore, xx appears as an entry in at least one of the words w1,w2w_1,w_2. Without loss of generality, assume that xx is an entry of w1w_1. By the induction hypothesis, we have x→w1x \to w_1, which means w1=xg1⋯xgnw_1=x^{g_1}\cdots x^{g_n} for some g1,…,gn∈Gg_1,\ldots,g_n \in G. Thus,

w=[w1,w2]=w1−1w2−1w1w2=((xk−1)gn⋯(xk−1)g1)(xg1w2⋯xgnw2).w=[w_1,w_2]=w_1^{-1}w_2^{-1}w_1w_2=((x^{k-1})^{g_n}\cdots(x^{k-1})^{g_1})(x^{g_1w_2}\cdots x^{g_nw_2}).

This shows that x→wx \to w.

We note that if P∙(G)\mathcal{P}_\bullet(G) is complete, then GG is a generalized torsion group. Furthermore, if GG is a generalized torsion group, then P∙(G)\mathcal{P}_\bullet(G) is complete if and only if P∗(G)\mathcal{P}^*(G) is complete. Obviously, we cannot replace “complete” with “connected” (as shown by G=C2×C2G=C_2\times C_2).

Lemma 2.2. Let GG be a group.

(1) If P∗(G)\mathcal{P}^*(G) is complete (resp. connected), then P∗(H)\mathcal{P}^*(H) is complete (resp. connected) for every direct factor HH of $G.

(2) If P∗(G)\mathcal{P}^*(G) is complete, then P∗(G/N)\mathcal{P}^*(G/N) is complete for every normal subgroup NN of GG.

Proof. (1) Let HH be a direct factor of GG. Thus, there exists a subgroup K≤GK \leq G such that G=H×KG=H\times K. Identify HH with the subgroup H×{1}≤GH\times\{1\}\leq G, and choose (h1,1),(h2,1)∈H×{1}(h_1,1),(h_2,1)\in H\times\{1\} non-trivial elements. Since P∗(G)\mathcal{P}^*(G) is complete, we may assume without loss of generality that (h2,1)→(h1,1)(h_2,1)\to(h_1,1). Thus, there exist elements (x1,y1),…,(xn,yn)∈G(x_1,y_1),\ldots,(x_n,y_n)\in G such that

(h1,1)=(h2,1)(x1,y1)⋯(h2,1)(xn,yn)=(h2x1⋯h2xn,1).(h_1,1)=(h_2,1)^{(x_1,y_1)}\cdots(h_2,1)^{(x_n,y_n)}=(h_2^{x_1}\cdots h_2^{x_n},1).

which implies that h1=h1x1⋯hnxnh_1=h^{x_1}_1\cdots h^{x_n}_n. Consequently, P∙∗(H)\mathcal{P}^*_\bullet(H) is complete. The connectedness part follows directly from the fact that if G=H×KG=H\times K and (h1,k1)(h_1,k_1) is adjacent to (h2,k2)(h_2,k_2), then either h1→h2h_1\to h_2 and k1→k2k_1\to k_2, or h2→h1h_2\to h_1 and k2→k1k_2\to k_1.

(2) Let xN,yNxN,yN be non-trivial distinct elements of G/NG/N. Since P∙∗(G)\mathcal{P}^*_\bullet(G) is complete, we have x→yx\to y or y→xy\to x. Assume, without loss of generality, that x→yx\to y. Then there exist elements g1,…,gk∈Gg_1,\ldots,g_k\in G such that y=xg1⋯xgky=x^{g_1}\cdots x^{g_k}. Projecting this equality onto the quotient G/NG/N, we obtain yN=(xN)g1N⋯(xN)gkNyN=(xN)^{g_1N}\cdots(xN)^{g_kN}, which completes the proof.

Note that in Lemma 2.2 (2), we cannot replace “complete” with “connected”. For instance, the proper generalized power graph of the quaternion group Q8Q_8 is connected, but Q8Q_8 has a quotient isomorphic to the Klein four-group, whose proper generalized power graph is disconnected.

Remark 2.3. As shown in [7], any two finite abelian groups with isomorphic proper power graphs must be isomorphic. However P∙∗(G)≃P∙∗(H)\mathcal{P}^*_\bullet(G)\simeq\mathcal{P}^*_\bullet(H) does not imply G≃HG\simeq H, even when GG and HH are nilpotent. Indeed, consider the quaternion group Q8Q_8 and the dihedral group D8D_8.

Further, the same holds if we replace “nilpotent” with “finite simple”, because the proper generalized power graph of a finite simple group is complete and there exist pairs of non-isomorphic finite simple groups of the same order, such as A8A_8 and PSL⁡3(4)\operatorname{PSL}_3(4).

Completeness

In this section, we present several results concerning the completeness of the (proper) generalized power graph. We begin with the following result, which characterizes all abelian groups whose generalized power graph is complete.

Throughout this paper, the Prüfer pp-group is denoted by Cp∞C_{p^\infty}, where pp is a prime.

Lemma 3.1. Let GG be an abelian group. Then P∙(G)\mathcal{P}_\bullet(G) is complete if only if G≃CpnG\simeq C_{p^n}, where n∈N∪{∞}n\in\mathbb{N}\cup\{\infty\}.

Proof. Since P∙(G)\mathcal{P}_\bullet(G) coincides to power graph of GG, the result follow by [8 Theorem 2.12] and [2 Theorem 1]. □

In the non-abelian case, the situation becomes much more interesting. For example, P∙(G)\mathcal{P}_\bullet(G) is complete if G=S3G=S_3 or A4A_4 (cf. § 1 and Example 3.4), whereas P∙(G)\mathcal{P}_\bullet(G) is not complete for any nonabelian group GG of order 8 (see Figure 2a and Figure 2b in Remark 2.3). We begin by proving the following result.

Proper generalized power graphs of Q8 and D8

Figure 2. Proper Generalized Power Graphs of Q8Q_8 and D8D_8.

Lemma 3.2. If GG is a finitely generated nilpotent group such that P∙∗(G)\mathcal{P}^*_\bullet(G) is complete, then GG must be a finite pp-group.

Proof. First, we show that GG must be a finite group. We proceed by induction on the nilpotency class of GG, denoted by cc. If c=1c=1, the result follows directly from Lemma 3.1. Assume c≥2c\ge2. Combining Lemma 2.2 (2) with the induction hypothesis, we deduce that the central quotient G/Z(G)G/Z(G) is a finite group. By Schur’s Theorem [20 10.1.4], the derived subgroup G′G' is finite. Lemmas 2.2 (2) and 3.1 ensure that the abelian quotient G/G′G/G' is a finite pp-group, which implies that GG is finite. Thus, GG decomposes as a direct product of its Sylow pip_i-subgroups, say G≅P1×⋯×PnG\cong P_1\times\cdots\times P_n. Suppose for a contradiction that n≥2n\ge2 and that P∗(G)\mathcal{P}^*(G) is complete. Consider the non-trivial elements x=(a,1,…,1)x=(a,1,\ldots,1) and y=(1,b,…,1)y=(1,b,\ldots,1) in GG. Since P∗(G)\mathcal{P}^*(G) is complete, xx and yy must be adjacent, say x→yx\to y. Then there exist elements g1,…,gk∈Gg_1,\ldots,g_k\in G such that y=xg1⋯xgky=x^{g_1}\cdots x^{g_k}. Consequently b=1b=1, which is a contradiction. Therefore, n=1n=1, and GG is a pp-group for some prime pp.

Example 3.3. Let D2n=⟨a,b∣an=1,b2=1,bab=a−1⟩D_{2n}=\langle a,b\mid a^n=1,b^2=1,bab=a^{-1}\rangle be the dihedral group of order 2n2n. We show that P∗(D2n)\mathcal{P}^*(D_{2n}) is complete if and only if n=pmn=p^m for an odd prime pp and a positive integer mm. Indeed, for every element ai∈⟨a⟩a^i\in\langle a\rangle, we have ⟨ai⟩=⟨(ai)D2n⟩\langle a^i\rangle=\langle(a^i)^{D_{2n}}\rangle. Consequently, if P∗(D2n)\mathcal{P}^*(D_{2n}) is complete, then the proper power graph P∗(⟨a⟩)\mathcal{P}^*(\langle a\rangle) is also complete. According to [8 Theorem 2.12], this occurs only when n=pmn=p^m for some prime pp and some positive integer mm.

If pp is an odd prime, all elements aib∈D2pma^ib\in D_{2p^m} belong to the same conjugacy class bD2pmb^{D_{2p^m}}, and ⟨bD2pm⟩=D2pm\langle b^{D_{2p^m}}\rangle=D_{2p^m}, which implies the completeness of P∗(D2pm)\mathcal{P}^*(D_{2p^m}).

If p=2p=2, it suffices to observe that the vertices aa and aba^b are not adjacent. Indeed, this follows from the fact that ⟨aD2⋅2m⟩=⟨a⟩\langle a^{D_{2\cdot2^m}}\rangle=\langle a\rangle and ⟨(a2b)D2⋅2m⟩={a2kb∣0≤k<2m−1 and i=0,1}\langle(a^2b)^{D_{2\cdot2^m}}\rangle=\{a^{2k}b\mid0\le k<2^{m-1}\text{ and }i=0,1\}.

Example 3.4. The graph P∗(A4)\mathcal{P}^*(A_4) is complete. Indeed, the four conjugacy classes of the alternating group A4A_4 are: C1={(1)}C_1=\{(1)\}, C2={(12)(34),(13)(24),(14)(23)}C_2=\{(12)(34),(13)(24),(14)(23)\}, C3={(123),(243),(134),(142)}C_3=\{(123),(243),(134),(142)\} and C4={(132),(234),(143),(124)}C_4=\{(132),(234),(143),(124)\}. Observe that (132)(132)(124)=(142)(132)(132)(124)=(142) and (123)(123)(234)(123)=(12)(34)(123)(123)(234)(123)=(12)(34). Thus, Lemma 2.1 (4) implies that C4→C3→C2C_4\to C_3\to C_2. Combining items (1) and (3) from Lemma 2.1, we obtain that P∗(A4)\mathcal{P}^*(A_4) is complete.

Proper Generalized Power Graph of A4

Figure 3. Proper Generalized Power Graph of A4A_4

Lemma 3.5. If GG is a torsion simple group, then P∗(G)\mathcal{P}^*(G) is complete.

Proof. If GG is abelian, we can deduce that GG has prime order. Thus, the result follows from Lemma 3.1. Now, assume that GG is nonabelian. Let xx and yy be arbitrary non-trivial elements of GG. Since GG is simple, we have G=⟨xG⟩G=\langle x^{G}\rangle, which implies that y=x±g1⋯x±gky=x^{\pm g_{1}}\cdots x^{\pm g_{k}} for some g1,…,gk∈Gg_{1},\ldots,g_{k}\in G. Since GG is a torsion group, x−1=xnx^{-1}=x^{n} for some positive integer nn, which shows that x→yx\to y. This completes the proof. □

Remark 3.6. Completeness is hereditary under direct summands and quotients, as shown in Lemma 2.2. However, it is not hereditary under taking subgroups. Indeed, while Lemma 3.5 ensures that P∙∗(A5)\mathcal{P}_{\bullet}^{*}(A_{5}) is complete, the alternating group A5A_{5} contains the Klein four-group V4≅C2×C2V_{4}\cong C_{2}\times C_{2} as a subgroup, whose corresponding graph P∙∗(V4)\mathcal{P}_{\bullet}^{*}(V_{4}) is not complete by Lemma 3.1.

As a consequence of Lemma 3.5, we obtain the following result.

Theorem 3.7. Every finite group can be embedded into a finite group GG such that P∙∗(G)\mathcal{P}_{\bullet}^{*}(G) is complete.

Proof. It is sufficient to show that P∙∗(Sn)\mathcal{P}_{\bullet}^{*}(S_{n}) is complete for all integers n≥5n\ge5. Let x,y∈Snx,y\in S_{n}. We prove that y→xy\to x or x→yx\to y by analysing the following cases:

Case 1: x,y∈Anx,y\in A_{n}

Since AnA_{n} is a nonabelian simple group for n≥5n\ge5, Lemma 3.5 applies.

Case 2: x∈Anx\in A_{n} and y∉Any\notin A_{n}

First, suppose that y2≠1y^{2}\ne1. Observe that y2∈Any^{2}\in A_{n}, and by Lemma 3.5, y2↔xy^{2}\leftrightarrow x. Hence, there exist g1,…,gm∈Sng_{1},\ldots,g_{m}\in S_{n} such that

x=(y2)g1⋯(y2)gm=(yg1)2⋯(ygm)2,x=(y^{2})^{g_{1}}\cdots(y^{2})^{g_{m}}=(y^{g_{1}})^{2}\cdots(y^{g_{m}})^{2},

which shows that y→xy\to x.

Now suppose that y2=1y^{2}=1. If An≤CSn(y)A_{n}\le C_{S_{n}}(y), since AnA_{n} has index 22 in SnS_{n}, we get that either CSn(y)=SnC_{S_{n}}(y)=S_{n}, which implies y∈Z(Sn)={1}y\in Z(S_{n})=\{1\}, or CSn(y)=AnC_{S_{n}}(y)=A_{n}, which implies y∈Any\in A_{n}. Both cases yield a contradiction. Thus, there exists an element z∈Anz\in A_{n} such that 1≠[y,z]=yyz∈An1\ne[y,z]=yy^{z}\in A_{n}. This implies that y→[y,z]y\to[y,z]. By Lemma 3.5, we have y→[y,z]↔xy\to[y,z]\leftrightarrow x. The result follows by applying Lemma 2.1(1).

Case 3: y∈Any\in A_{n} and x∉Anx\notin A_{n}

Repeating the argument used in the proof of Case 2, we obtain x→yx\to y.

Case 4: x,y∉Anx,y\notin A_{n}

Since xy∈Anxy\in A_{n}, by Case 2, we have y→xyy\to xy. Thus, there exist h1,…,hs∈Snh_{1},\ldots,h_{s}\in S_{n} such that

xy=yh1⋯yhs.xy=y^{h_{1}}\cdots y^{h_{s}}.

Let rr be the order of yy. Multiplying the right-hand side of the equality by yr−1y^{r-1}, we get

x=yh1⋯yhsyr−1,x=y^{h_{1}}\cdots y^{h_{s}}y^{r-1},

which proves that y→xy\to x. This completes the proof. □

Theorem 3.8. Let FF be a free group on countably many generators. Then P∙∗(F)\mathcal{P}_{\bullet}^{*}(F) is not complete.

Proof. By Lemma 3.1 we can assume that the rank of GG is at least 22. Assume by contradiction that P∙∗(F)\mathcal{P}_{\bullet}^{*}(F) is complete. By Lemma 2.2(2), P∙∗(F/F′)\mathcal{P}_{\bullet}^{*}(F/F') is complete as well. However, F/F′F/F' is isomorphic to Cartesian product of copies of Z\mathbb{Z} what is a contradiction (by Lemma 3.1). □

Corollary 3.9. If GG has any factor direct or quotient which is isomorphic to free group on countably many generators, then P∙∗(G)\mathcal{P}_{\bullet}^{*}(G) is not complete.

Proof. This follows immediately from the above theorem together with Lemma 2.2. □

The next lemma is probably well-known but we do not have a reference to it.

Lemma 3.10. A finite group GG satisfies exactly one of the following conditions:

  • GG is metacyclic; or

  • GG is non-metacyclic and contains a subgroup isomorphic to either Cp×CpC_p \times C_p (for some prime pp) or quaternion group of order eight Q8Q_8.

Proof. Assume that GG is non-metacyclic and that GG contains no subgroup isomorphic to Cp×CpC_p \times C_p for any prime pp. Since GG is non-metacyclic, it follows from [11 Theorem 7.6.2] that some Sylow pp-subgroup of GG is non-cyclic. Let PP be such a subgroup. This PP possesses a central subgroup of order pp. Furthermore, this central subgroup is necessarily unique, otherwise GG would contain a copy of Cp×CpC_p \times C_p. Hence, by Burnside’s theorem (cf. [20 5.3.6]), PP admits a subgroup isomorphic to the quaternion group Q8Q_8.

Recall that a Dedekind group is a group in which all subgroups are normal. We show the following result.

Proposition 3.11. (1) If GG is a Dedekind group, then P∙(G)\mathcal{P}_{\bullet}(G) is complete if only if G≃CpnG \simeq C_{p^n}, where n∈N∪{∞}n \in\mathbb{N} \cup\{\infty\}.

(2) If GG is a finite group which P∙(H)\mathcal{P}_{\bullet}(H) is complete for every proper subgroup and quotient HH of GG, then:

(a) GG is cyclic of order pqpq for some different primes pp and qq;

(b) GG is cyclic of order power of a prime;

(c) GG is metacyclic and is an extension of CpaC_{p^a} by CqbC_{q^b}, for some primes p,qp,q and positive integers a,ba,b.

Proof. (1) By [20 5.3.7], a group GG is Dedekind if and only if it is either abelian or isomorphic to the direct product Q8×AQ_8 \times A, where AA is a torsion abelian group. If GG is abelian, then the result follows by Lemma 3.1. Now assume that G=Q8×AG = Q_8 \times A. It follows from Lemma 2.2 (1) that if P∙(G)\mathcal{P}_{\bullet}(G) is complete, then P∙(Q8)\mathcal{P}_{\bullet}(Q_8) must also be complete. However, P∙(Q8)\mathcal{P}_{\bullet}(Q_8) is not complete. Therefore, in this case, P∙(G)\mathcal{P}_{\bullet}(G) cannot be complete.

(2) If GG is non-metacyclic, then by Lemma 3.10, GG contains a subgroup isomorphic to either Cp×CpC_p \times C_p or Q8Q_8. Thus, combining Item (1), Lemma 3.1 and Lemma 2.2 (2), we get a contradiction. Now, assume that GG is metacyclic. If GG is cyclic, then G≃CnG \simeq C_n, where n=p1n1⋯prnrn = p_1^{n_1} \cdots p_r^{n_r} and p1,…,prp_1, \ldots, p_r are pairwise distinct primes. It follows from Lemma 3.1 that GG is either of prime-power order or of order pqpq for distinct primes pp and qq. Assuming that GG is not cyclic, there exists N⊲GN \lhd G be such that both NN and G/NG/N are non-trivial cyclic. By Lemma 3.1, NN and G/NG/N have prime-power orders, say pap^a and qbq^b, respectively. This completes the proof.

The structure of the pp-subgroups of a finite group can reveal significant information about the group itself. For instance, Lucchini proved in [18] that if every pp-subgroup of a group GG can be generated by at most rr elements for every prime pp dividing the order of GG, then every subgroup of GG can be generated by at most r+1r+1 elements. This insight naturally motivates the following problem.

Problem 3.12. Let GG be a finite group and let P\mathcal{P} be a graph property. If P∙(H)\mathcal{P}_{\bullet}(H) satisfies P\mathcal{P} for every pp-subgroup HH of GG and every prime pp dividing the order of GG, what can be said about the structure of GG?

We establish the following result.

Proposition 3.13. Let GG be a group satisfying the hypothesis of Problem 3.12, where P\mathcal{P} is the property of being complete. Then GG is metacyclic.

Proof. If GG is non-metacyclic, then by Lemma 3.10, GG contains a subgroup isomorphic to either Cp×CpC_p \times C_p for some prime pp or the quaternion group Q8Q_8. However, this contradicts the hypothesis in view of Lemmas 3.11 (1) and 3.1. The proof is complete.

We end this section with the following problem.

Problem 3.14. Can we characterize all finite groups GG for which P∙(G)\mathcal{P}_{\bullet}(G) is complete?

Connectivity

Let G=⟨a⟩×⟨b⟩G=\langle a\rangle\times\langle b\rangle be the elementary abelian pp-group for some prime pp. It is straightforward to verify that the element (a,b)(a,b) is an isolated vertex in the graph P∙∗(G)\mathcal{P}^{*}_{\bullet}(G). Hence, this graph is disconnected. However, the following result shows that every proper generalized power graph of a finite group whose center is not a pp-group is connected.

Theorem 4.1. If GG is a finite group such that Z(G)Z(G) is not a pp-group, then P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is connected with diameter at most 4.

Proof. First, assume that GG is a finite abelian group. Since P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) coincides with the proper power graph of GG when GG is abelian, the connectedness statement follows from [10 Theorem 2.4]. For the reader’s convenience, we provide a detailed proof below.

Let pp be a prime dividing the order of GG. By the Fundamental Theorem of Finite Abelian Groups, we can write G=P×QG=P\times Q, where PP is a pp-group and QQ is a non-trivial p′p'-group of order qq.

Consider (x,y),(x1,y1)∈G(x,y),(x_{1},y_{1})\in G satisfying

∣x∣=pn≠1,∣y∣≠1,∣x1∣=pn1≠1,∣y1∣≠1.\lvert x\rvert=p^{n}\ne1,\qquad\lvert y\rvert\ne1,\qquad\lvert x_{1}\rvert=p^{n_{1}}\ne1,\qquad\lvert y_{1}\rvert\ne1.

By Bezout’s lemma, there exist integers rr and ss such that rpn+sq=1rp^{n}+sq=1. Consequently,

(x,y)rpn=(xrpn,yrpn)=(1,yrpn)=(1,y1−sq)=(1,y),(x,y)^{rp^{n}}=(x^{rp^{n}},y^{rp^{n}})=(1,y^{rp^{n}})=(1,y^{1-sq})=(1,y),
(x,y)sq=(xsq,ysq)=(xsq,1)=(x1−rpn,1)=(x,1).(x,y)^{sq}=(x^{sq},y^{sq})=(x^{sq},1)=(x^{1-rp^{n}},1)=(x,1).

Thus, the vertex (x,y)(x,y) is connected to (x,1)(x,1) and (1,y)(1,y). Similarly, (x1,y1)(x_{1},y_{1}) is connected to (x1,1)(x_{1},1) and (1,y1)(1,y_{1}), and (x1,y)(x_{1},y) is connected to (x1,1)(x_{1},1) and (1,y)(1,y). Hence, we have the path

(x,y)→(1,y)←(x1,y)→(x1,1)←(x1,y1),(x,y)\mathbin{\to}(1,y)\mathbin{\leftarrow}(x_{1},y)\mathbin{\to}(x_{1},1)\mathbin{\leftarrow}(x_{1},y_{1}),

which yields a path of length four between (x,y)(x,y) and (x1,y1)(x_{1},y_{1}), which concludes the proof for the abelian case.

Now, assume that GG is a finite nonabelian group. Let xx and yy be non-trivial elements of GG. If [x,y]≠1[x,y]\ne1, then by Lemma 2.1(8) we have x→[x,y]←yx\mathbin{\to}[x,y]\mathbin{\leftarrow}y, which yields a path of length two between xx and yy. Thus, we may assume that xx and yy commute. Consider the subgroup H=⟨x,y,Z(G)⟩H=\langle x,y,Z(G)\rangle, which is abelian. By assumption, HH is not a pp-group, implying that P∙∗(H)\mathcal{P}^{*}_{\bullet}(H) is connected with diameter at most 4. This completes the proof. □

Next, we establish the following result.

Theorem 4.2. (1) If GG is a torsion group with trivial center, then P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is connected with diameter at most 4.

(2) If GG is a finite group with a cyclic center, then P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is connected with diameter at most 4.

Proof. (1) Let xx and yy be non-trivial elements of GG. If xx and yy do not commute, then by Lemma 2.1(8) we get x→[x,y]←yx\mathbin{\to}[x,y]\mathbin{\leftarrow}y. Suppose now that xx and yy commute. Since neither element is central, there exist elements x1,y1∈Gx_{1},y_{1}\in G such that xx does not commute with x1x_{1} and yy does not commute with y1y_{1}. By Lemma 2.1(8) again, if either xx does not commute with y1y_{1} or yy does not commute with x1x_{1}, then xx and yy are connected by a path of length at most 4. Otherwise, xx commutes with y1y_{1} and yy commutes with x1x_{1}. In this case, applying commutator calculus (cf. [20 5.1.5 (ii)]), we obtain the following path

y→[y,x1y1]←x1y1→[x,x1y1]←x.y\mathbin{\to}[y,x_{1}y_{1}]\mathbin{\leftarrow}x_{1}y_{1}\mathbin{\to}[x,x_{1}y_{1}]\mathbin{\leftarrow}x.

(2) If Z(G)Z(G) is not a pp-group, then Theorem 4.1 applies and the proof is complete. Hence, we may assume that Z(G)Z(G) is a pp-group. Repeating the above argument, we obtain a path of length at most 5 between any two non-central elements. Furthermore, since the center of GG is cyclic, say Z(G)=⟨z⟩Z(G)=\langle z\rangle, any two elements of Z(G)Z(G) are connected by a path of length at most 2. Suppose that there exists an element g∈Gg\in G whose order is coprime to pp. The subgroup ⟨g,Z(G)⟩\langle g,Z(G)\rangle is not a pp-group. By Theorem 4.1, there exists a path of length at most 4 connecting gg to any element of Z(G)Z(G). Therefore, it remains to consider the situation where GG itself is a pp-group. In this case, let x,y∈G∖{1}x,y\in G\setminus\{1\} be such that xx is non-central and yy is central, and set N=⟨xG⟩N=\langle x^G\rangle, the normal closure of xx. If N=GN=G, then by Lemma 2.1 (5) every element of GG is adjacent to xx in P∙∗(G)\mathcal{P}^{*}_{\bullet}(G). If N≠GN\ne G, since GG is nilpotent, we have N∩Z(G)≠{1}N\cap Z(G)\ne\{1\}. Choosing a non-trivial element ziz^i in this intersection, we obtain the following path connecting xx and yy

x→zi←z→y.x\to z^i\leftarrow z\to y.

This completes the proof.

Note that the previous theorem ensures that the proper generalized power graph of groups such as the Gupta-Sidki, Grigorchuk, and Tarski monster groups is connected.

Problem 4.3. Can the “finite group” hypothesis be replaced with “torsion group” in Item (2) of Theorem 4.2?

In view of the two preceding theorems, it remains to consider the case where the center is a non-cyclic pp-group. To this end, we present a finite group GG with such a center for which P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is disconnected.

Example 4.4. Let G=D8×C2G=D_8\times C_2 be the direct product of the dihedral group D8=⟨a,b∣a4=1,b2=1,bab=a−1⟩D_8=\langle a,b\mid a^4=1,b^2=1,bab=a^{-1}\rangle and the cyclic group C2=⟨c∣c2=1⟩C_2=\langle c\mid c^2=1\rangle. We claim that (1,c)(1,c) is an isolated vertex of P∙∗(G)\mathcal{P}^{*}_{\bullet}(G). Indeed, the center of GG, given by Z(G)={(1,1),(a2,1),(1,c),(a2,c)}Z(G)=\{(1,1),(a^2,1),(1,c),(a^2,c)\}, is isomorphic to the Klein four-group. It is straightforward to observe that if (x,y)(x,y) is adjacent to (1,c)(1,c), then (x,y)→(1,c)(x,y)\to(1,c). In this case, there exist elements (g1,h1),…,(gn,hn)∈G(g_1,h_1),\ldots,(g_n,h_n)\in G such that

(1,c)=∏i=1n(x,y)(gi,hi)=(∏i=1nxgi,yn).(1,c)=\prod_{i=1}^{n}(x,y)^{(g_i,h_i)}=\left(\prod_{i=1}^{n}x^{g_i},y^n\right).

Consequently, y=cy=c, nn is odd, and ∏i=1nxgi=1\prod_{i=1}^{n}x^{g_i}=1. Since ∣D8∣=8|D_8|=8, the element xx must be non-central. Thus, xx belongs to one of the following conjugacy classes:

bD8={b,a2b},(ab)D8={ab,a3b},andaD8={a,a3}.b^{D_8}=\{b,a^2b\},\qquad(ab)^{D_8}=\{ab,a^3b\},\qquad\text{and}\qquad a^{D_8}=\{a,a^3\}.

However, any two elements within the same conjugacy class commute. Thus, no such element xx exists, which implies that P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is disconnected. Direct calculations show that the graph P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) can be represented as follows:

Figure 4 (PDF p. 10).

Theorem 4.5. Let n≥2n \ge2 be an integer and G=SLn(F)G = SL_n(F) be the special linear group of degree nn over a field FF. Then P∗(G)\mathcal{P}^{*}(G) is connected.

Proof. Assume that n>2n > 2 or ∣F∣>3|F| > 3, and let A∈GA \in G be a non-scalar matrix. By [23], every element BB in GG is a product of conjugates of AA, and so A→BA \to B. Thus, in order to complete the proof, it is sufficient to show that P∗(G)\mathcal{P}^{*}(G) is connected when G=SL(2,2)G = SL(2,2) and G=SL(2,3)G = SL(2,3). The first case, G=SL(2,2)G = SL(2,2), follows from Theorem 4.2 (1) because SL(2,2)≅S3SL(2,2) \cong S_3. Now, we will deal with the case G=SL(2,3)G = SL(2,3). Here, GG has order 24 and contains a normal subgroup isomorphic to the quaternion group Q8Q_8. By Theorem 4.2 (2) (or the figure in Remark 2.3), P∗(Q8)\mathcal{P}^{*}(Q_8) is connected. Further, GG has exactly four Sylow 3-subgroups, generated respectively by the matrices

A=(1101),B=(1011),C=(0212),D=(2210).A = \begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix}, \qquad B = \begin{pmatrix} 1 & 0 \\ 1 & 1 \end{pmatrix}, \qquad C = \begin{pmatrix} 0 & 2 \\ 1 & 2 \end{pmatrix}, \qquad D = \begin{pmatrix} 2 & 2 \\ 1 & 0 \end{pmatrix}.

Since all Sylow 3-subgroups are conjugate, all elements of order 3 are connected. Consider arbitrarily MM a non-trivial element in GG. Note that G3≤Q8G^3 \leq Q_8. If M3≠1M^3 \ne1, then MM is connected to all elements in Q8Q_8. If M3=1M^3 = 1, then MM is contained in some 3-Sylow subgroup. Further, the commutator [A,B]=(01−11)[A,B] = \begin{pmatrix} 0 & 1 \\ -1 & 1 \end{pmatrix} has order 4, and thus [A,B][A,B] is connected to all elements in Q8Q_8. By Lemma 2.1 (8), C→A→[A,B]←B←DC \to A \to[A,B] \leftarrow B \leftarrow D. This shows that every non-trivial element of GG is connected to all non-trivial elements of Q8Q_8. This completes the proof. □

Theorem 4.6. (1) Let GG be a finite group of odd order satisfying the hypothesis of Problem 3.12, where P\mathcal{P} is the property of being connected. Then GG is metacyclic.

(2) If GG is a finite group such that P∗(H/N)\mathcal{P}^{*}(H/N) is connected for every proper subgroup HH of GG and every N◃HN \triangleleft H, then GG is metacyclic.

Proof. (1) Assume that GG is a non-metacyclic group. Since GG has odd order, Lemma 3.10 implies that GG contains a subgroup isomorphic to Cp×CpC_p \times C_p for some prime pp. However, because P∗(Cp×Cp)\mathcal{P}^{*}(C_p \times C_p) is not connected, this yields a contradiction.

(2) Suppose that GG is a non-metacyclic group. Lemma 3.10 implies that GG contains a subgroup isomorphic to Cp×CpC_p \times C_p for some prime pp or to a generalized quaternion group. First, assume that GG is a pp-group for some prime pp. If p=2p = 2, our assumption implies that GG has a unique subgroup of order 2. By [20 5.3.6], GG is either cyclic or a generalized quaternion group, yielding a contradiction in both cases. If pp is odd, Lemma 3.10 implies that GG contains a subgroup isomorphic to Cp×CpC_p \times C_p, which is again a contradiction.

Now, assume that GG is not a pp-group. By Item (1), we may assume that GG has even order. Furthermore, it follows from [20 5.3.6] that every Sylow subgroup of odd order in GG is cyclic, and every Sylow 22-subgroup is either cyclic or a generalized quaternion group. Let PP be a Sylow 22-subgroup of GG. If PP is cyclic, then [11 Theorem 7.6.2] implies that GG is metacyclic, a contradiction. Consequently, PP is a generalized quaternion group. But P/P′≃C2×C2P/P' \simeq C_2 \times C_2, which is a contradiction. This completes the proof.

Planarity

In this section, we provide necessary and sufficient conditions for the planarity of the proper generalized power graph P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) of a finite group GG. More precisely, we prove the following result.

Theorem 5.1. Let GG be a non-trivial finite group. Then P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is planar if and only if one of the following conditions holds:

  1. G≃C3mG \simeq C_3^m, for some integer m≥1m \ge1;

  2. G≃C5mG \simeq C_5^m, for some integer m≥1m \ge1;

  3. G≃C2m×C3nG \simeq C_2^m \times C_3^n, where either m=1m = 1 or n=1n = 1;

  4. G≃C2m×C4nG \simeq C_2^m \times C_4^n, for some non-negative integers mm and nn, not both zero;

  5. GG is a non-abelian group of exponent 44 that is a central extension of C2C_2 by C2mC_2^m for some positive integer mm.

To prove the Theorem 5.1, we first establish several auxiliary lemmas.

Lemma 5.2. The graph P∙∗(C6×C6)\mathcal{P}^{*}_{\bullet}(C_6 \times C_6) is non-planar.

Proof. Consider the additive group C6×C6C_6 \times C_6. Let a1=(3,0)a_1 = (3, 0), a2=(0,3)a_2 = (0, 3), and a3=(3,3)a_3 = (3, 3) be elements of order 22, and let b1=(0,2)b_1 = (0, 2), b2=(0,4)b_2 = (0, 4), and b3=(2,0)b_3 = (2, 0) be elements of order 33. For each i,j∈{1,2,3}i,j \in\{1,2,3\}, define cij=ai+bjc_{ij} = a_i + b_j. Note that 4cij=bj4c_{ij} = b_j and 3cij=ai3c_{ij} = a_i, which implies that for each pair (i,j)(i,j) there exists a path bj←cij→aib_j \leftarrow c_{ij} \rightarrow a_i. Consequently, the graph P∙∗(C6×C6)\mathcal{P}^{*}_{\bullet}(C_6 \times C_6) contains a subdivision of K3,3K_{3,3}. By Kuratowski’s theorem [6 Theorem 9.10, p. 153], P∙∗(C6×C6)\mathcal{P}^{*}_{\bullet}(C_6 \times C_6) is non-planar, which completes the proof. □

Let π\pi be a set of prime numbers. Recall that a group GG is called a π\pi-group if every non-trivial element in GG has an order whose prime factors belong to π\pi.

Lemma 5.3. Let GG be a non-trivial finite group such that P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is planar. Then GG is a metabelian {2,3,5}\{2,3,5\}-group and every conjugacy class of GG has size at most 44.

Proof. Suppose there exists a prime p≥7p \ge7 dividing the order of GG. Then, by Cauchy’s Theorem, GG contains a cyclic subgroup CC of order pp. Consequently, P∙∗(C)≃Kp−1\mathcal{P}^{*}_{\bullet}(C) \simeq K_{p-1}. Since p−1≥6p - 1 \ge6, it follows that P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) contains K5K_5 as a subgraph. By Kuratowski’s Theorem, P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is non-planar, which implies that GG must be a {2,3,5}\{2,3,5\}-group.

Now let xx be a non-trivial element of GG. If ∣xG∣≥5|x^G| \ge5, then by Lemma 2.1 (3), P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) again contains K5K_5 as a subgraph, contradicting planarity. Hence, we have that every conjugacy class of GG has size at most 44. Under this assumption, it follows from [22 Theorem 5.6 and Section 6] that the derived subgroup G′G' is abelian, completing the proof. □

Lemma 5.4. Let GG be a finite group containing an element aa with at least three conjugates. If P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is planar, then ⟨aG⟩=aG∪{1}\langle a^G\rangle= a^G \cup\{1\}. Furthermore, ⟨aG⟩\langle a^G\rangle is isomorphic to V4V_4, the Klein four-group.

Proof. First, suppose that aa is adjacent to some element b≠1b \ne1 with b∉aGb \notin a^G. We claim that bb is a central element of order 22. Indeed, if b∉Z(G)b \notin Z(G), then ∣bG∣≥2|b^G| \ge2. By Lemma 2.1 (3) and (4), every element in aGa^G is adjacent to every element in bGb^G and thus K5K_5 is a subgraph of P∙∗(G)\mathcal{P}^{*}_{\bullet}(G), which contradicts Kuratowski’s theorem. Therefore, b∈Z(G)b \in Z(G). Furthermore, if b2≠1b^{2} \ne1, then every element in aGa^{G} is adjacent to every element in ⟨b⟩\langle b\rangle, which again yields that K5K_{5} is a subgraph of P∗(G)\mathcal{P}^{*}(G), reaching a contradiction.

Now, suppose that there exist distinct non-trivial elements b1,b2,…,bn∉aGb_{1}, b_{2}, \ldots, b_{n} \notin a^{G} with n≥3n \ge3, adjacent to aa. Lemma 2.1(3) and (4) imply that K3,3K_{3,3} is a subgraph of P∗(G)\mathcal{P}^{*}(G), again contradicting planarity. Thus, the order of ⟨aG⟩\langle a^{G}\rangle can only be 44, 55, or 66. If ∣⟨aG⟩∣=6|\langle a^{G}\rangle| = 6, then there exist two central involutions b1,b2∈⟨aG⟩b_{1}, b_{2} \in\langle a^{G}\rangle such that b1,b2∉aGb_{1}, b_{2} \notin a^{G}. Since 33 divides ∣⟨aG⟩∣|\langle a^{G}\rangle|, the subgroup ⟨aG⟩\langle a^{G}\rangle contains an element of order 33; because b1b_{1} and b2b_{2} are involutions, aa must have order 33. However, no group of order 66 contains two central involutions (or three elements of order 33), so ∣⟨aG⟩∣≠6|\langle a^{G}\rangle| \ne6. If we suppose that ∣⟨aG⟩∣=5|\langle a^{G}\rangle| = 5, there exists a central involution b∈⟨aG⟩b \in\langle a^{G}\rangle, which contradicts Lagrange’s theorem. Thus, the order of ⟨aG⟩\langle a^{G}\rangle must be 44. Since all elements in aGa^{G} are conjugate and therefore have the same order, it follows that ⟨aG⟩≅V4\langle a^{G}\rangle\cong V_{4}, which completes the proof. □

Lemma 5.5. Let GG be a finite group in which no element has more than 44 conjugates, and some element has exactly 44 conjugates. If P∗(G)\mathcal{P}^{*}(G) is planar, then GG has some element of order 55.

Proof. It is sufficient to show that GG contains a subgroup of order 55. Let xx be an element with exactly 44 conjugates, and consider H=⟨xG⟩H = \langle x^{G}\rangle, the normal closure of xx. It is clear that ∣H∣≥5|H| \ge5. Suppose, for a contradiction, that ∣H∣≥6|H| \ge6. Then there exists a non-trivial element y∈H∖xGy \in H \setminus x^{G}. By Lemma 2.1(4), yy is adjacent to all elements of xGx^{G}, implying that P∗(G)\mathcal{P}^{*}(G) contains K5K_{5} as a subgraph. By Kuratowski’s Theorem [6], P∗(G)\mathcal{P}^{*}(G) is non-planar, contradicting our assumption. Thus, ∣H∣=5|H| = 5, completing the proof. □

Lemma 5.6. Let GG be a non-trivial finite group.

  1. If GG is an extension of C3C_{3} by C5mC_{5}^{m}, m≥1m \ge1, then GG is abelian. Consequently, P∗(G)\mathcal{P}^{*}(G) is non-planar.

  2. If GG is an extension of NN by C5mC_{5}^{m}, m≥1m \ge1, where ∣N∣=2n|N| = 2^{n}, n≤3n \le3, then P∗(G)\mathcal{P}^{*}(G) is non-planar.

Proof. We only present the proof of (1), since (2) follows similarly. By Sylow’s theorem, GG contains a unique Sylow 55-subgroup, which is therefore normal. Since its Sylow subgroups are normal and abelian with coprime orders, GG is an abelian group containing an element of order 1515. The result then follows from [9 Theorem 4.2]. □

Proof of Theorem 5.1. First, suppose that P∗(G)\mathcal{P}^{*}(G) is planar. By Lemma 5.3 we may assume henceforth that GG is a metabelian {2,3,5}\{2,3,5\}-group and every conjugacy class of GG has size at most 44. We prove this direction of the theorem by considering each case separately.

Case 1: Suppose that every conjugacy class has cardinality 11. In this case, GG is abelian. If GG has a non-trivial Sylow 55-subgroup, then by Theorem 4.24.2 of [9], we have G≅C5mG \cong C_{5}^{m} for some positive integer mm (since the existence of any other Sylow subgroup would force GG to contain a cyclic subgroup of order greater than 66). Otherwise, G=P2×P3G = P_{2} \times P_{3}, where P2P_{2} and P3P_{3} are Sylow pp-subgroups for p=2,3p = 2, 3, respectively (each possibly trivial). If P2=1P_{2} = 1, then G≅C3nG \cong C_{3}^{n} for some n≥1n \ge1. If P3=1P_{3} = 1, then GG has exponent dividing 44. If both are non-trivial, Theorem 4.24.2 of [9] yields G≅C2m×C3nG \cong C_{2}^{m} \times C_{3}^{n} for positive integers m,nm,n. By Lemma 5.2, if m≥2m \ge2 and n≥2n \ge2, then P∗(G)\mathcal{P}^{*}(G) is non-planar. Thus, either m=1m = 1 or n=1n = 1.

Case 2: Suppose that every conjugacy class has cardinality at most 22, and there exists an element with exactly two conjugates. By Theorem 5.65.6 of [22], the derived subgroup G′G' is cyclic of order 22, say G′=⟨z⟩G' = \langle z\rangle. Since zg∈G′={1,z}z^{g} \in G' = \{1,z\} for all gg in GG, we get that G′≤Z(G)G' \le Z(G). Thus, we conclude that GG is nilpotent, and therefore it decomposes as the direct product of its Sylow subgroups. By [9 Theorem 4.2], the Sylow 55-subgroup of GG is trivial. Furthermore, [22 Theorem 5.6] also implies that the Sylow 33-subgroup of GG is trivial.

Since GG is nonabelian, it follows from [9 Theorem 4.2] that GG has exponent exactly 44. Thus, GG is an extension of C2C_{2} by C2m×C4nC_{2}^{m} \times C_{4}^{n}, for some integers mm and nn. Next, we show that n=0n = 0. To this end, suppose by contradiction that n≥1n \ge1. Firstly, note that if xx is a non-central element of GG, then xg=x[x,g]x^{g}=x[x,g] for all g∈Gg \in G. Since ∣xG∣=2|x^{G}|=2 and G′={1,z}G^{\prime}=\{1,z\}, it follows that xG={x,xz}x^{G}=\{x,xz\}.

Now, set T={g∈G∣g2∉G′}T=\{g \in G \mid g^{2}\notin G^{\prime}\}. Since n≥1n \geq1, we have T≠∅T \ne\emptyset. Note that G∖TG \setminus T is a proper subgroup of GG, which implies that G=⟨T⟩G=\langle T\rangle. If T⊆Z(G)T \subseteq Z(G), then GG is abelian, a contradiction. Thus there exist y∈T∖Z(G)y \in T \setminus Z(G). By previous paragraph, we have that yG={y,yz}y^{G}=\{y,yz\} and

⟨yG⟩=⟨y,z⟩={1,z,y,y2,y3,yz,y2z,y3z}.\langle y^{G}\rangle=\langle y,z\rangle=\{1,z,y,y^{2},y^{3},yz,y^{2}z,y^{3}z\}.

Note that since yy is noncentral, y3=y−1y^{3}=y^{-1}, yzyz and y3zy^{3}z are also noncentral. Thus, by the same argument as above, we have that (yz)G={yz,y}(yz)^{G}=\{yz,y\}, (y3)G={y3,y3z}=(y3z)G(y^{3})^{G}=\{y^{3},y^{3}z\}=(y^{3}z)^{G}. Therefore, we conclude that

⟨yG⟩=⟨(yz)G⟩=⟨(y3)G⟩=⟨(y3z)G⟩=⟨y,z⟩.\langle y^{G}\rangle=\langle(yz)^{G}\rangle=\langle(y^{3})^{G}\rangle=\langle(y^{3}z)^{G}\rangle=\langle y,z\rangle.

Consequently, K5K_{5} is a subgraph of P∙∗(G)\mathcal{P}^{*}_{\bullet}(G), a contradiction. Thus n=0n=0, which implies that GG has exponent 4 and is an extension of C2C_{2} by C2mC_{2}^{m} for some integer mm.

Case 3: Suppose that every conjugacy class in GG has cardinality at most 3, and there exists an element with exactly 3 conjugates. By [22 Theorem 5.6], the derived subgroup G′G^{\prime} is cyclic of order 3.

Let S={x∈G∣∣xG∣=3}S=\{x \in G \mid|x^{G}|=3\} be the set of all elements in GG whose conjugacy classes have size 3. For each x∈Sx \in S, set Hx=⟨xG⟩H_{x}=\langle x^{G}\rangle, and define the subgroup H=∏x∈SHxH=\prod_{x\in S}H_{x}. Consider the quotient group G/HG/H. Note that every conjugacy class of G/HG/H has size at most 2. Further, its derived subgroup (G/H)′≅G′/H∩G′(G/H)^{\prime}\cong G^{\prime}/H\cap G^{\prime} has order dividing 3.

Thereby, if there exists an element in G/HG/H with exactly 2 conjugates, then [22 Theorem 5.6] ensures that (G/H)′(G/H)^{\prime} is cyclic of order 2, a contradiction. Hence, G/HG/H must be abelian, so (G/H)′={1}(G/H)^{\prime}=\{1\}, which implies G′⊆HG^{\prime}\subseteq H. However, by Lemma 5.4, HH is a 2-group, yielding a final contradiction.

Case 4: Suppose that every conjugacy class in GG has cardinality at most 4 and there exists an element with exactly 4 conjugates. According to Section 6 of [22], the derived subgroup G′G^{\prime} is isomorphic to C4C_{4}, C22C_{2}^{2}, or C23C_{2}^{3}. Moreover, Lemma 5.5 implies that 5 divides the order of GG. Hence, by Theorem 4.2 of [9], we obtain G/G′≃C5mG/G^{\prime}\simeq C_{5}^{m} for some integer m≥1m \geq1. However, Lemma 5.6 [2] guarantees that GG is abelian and contains a cyclic subgroup of order 10, which contradicts Theorem 4.2 of [9].

Now, we provide a proof of the converse. If GG satisfies one of the first three conditions, then [9 Theorems 4.3 and 4.5] ensure that P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is planar. Theorem 2 of [2] implies that P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) is planar if GG satisfies the conditions of item (4). Finally, assume that GG satisfies the conditions of item (5). Thus, there exists a central involution zz satisfying G/⟨z⟩G/\langle z\rangle is elementary abelian 2-group. Since GG is nonabelian, we get G′=⟨z⟩G^{\prime}=\langle z\rangle. Consequently, for any pair of elements x,g∈Gx,g \in G we get xg=x[x,g]x^{g}=x[x,g] which implies that xG={x,xz}x^{G}=\{x,xz\}.

Assume that y→xy\to x. We show that x∈{z,yz}x\in\{z,yz\}. Indeed, there exist elements g1,…,gn∈Gg_{1},\ldots,g_{n}\in G satisfying

x=∏i=1nygi∈⟨yG⟩=⟨y,yz⟩=⟨y,z⟩.x=\prod_{i=1}^{n}y^{g_{i}}\in\langle y^{G}\rangle=\langle y,yz\rangle=\langle y,z\rangle.

Since zz is a central involution, we have either x=yjx=y^{j} or x=yjzx=y^{j}z for some integer jj. Furthermore, since y2∈⟨z⟩={1,z}y^{2}\in\langle z\rangle=\{1,z\}, it is straightforward to see that in both cases x∈{z,yz}x\in\{z,yz\}. This shows that if yy is not an isolated vertex, then its unique adjacent vertices are zz and yzyz. Up to isolated vertices, the graph P∙∗(G)\mathcal{P}^{*}_{\bullet}(G) can be represented as shown in the figure below.

Graph with vertices $z$, $y_1$, $y_1z$, $y_2$, $y_2z$, $y_3$, and $y_3z$

Figure 5.

This completes the proof.

IA

Artificial intelligence (https://gemini.google.com/app) was used to assist with routine typesetting, figure generation, grammar checking, and phrasing. Apart from these minor editorial and illustrative uses, the text and results of this paper were entirely authored by the human authors.

Disclosure statement

The authors have no conflicts of interest to declare that are relevant to the content of this article.

Data sharing

Not applicable to this article as no datasets were generated or analyzed during the current study.

Funding

This work was partially supported by FAPEMIG (grant number RED-00133-21).

Acknowledgement

The second author acknowledges the Departamento de Matemática of the Universidade Federal de Minas Gerais for its excellent hospitality while this research was conducted.

(D. Silveira) Departamento de Ciências Exatas e Aplicadas, Universidade Federal de Ouro Preto, 35931-008 Ouro Preto, MG, Brazil

Departamento de Matemática, Universidade Federal Minas Gerais, 31270-901 Belo Horizonte, MG, Brazil

Email address: [email protected]

(Y. Contreras-Rojas) Departamento de Matemática, Universidade Estadual de Maringá 87020-900 Maringá, PR, Brazil

Email address: [email protected]

References

  1. [1]J. Abaway, A. Kelarev, and M. Chowdhury, Power graphs: a survey, Electron. J. Graph Theory Appl. 1 (2013), no. 2, 125–147.
  2. [2]A. Doostabadi, A. Erfanian, and A. Jafarzadeh, Some results on the power graphs of finite groups, Sci. Asia 41 (2015), no. 1, 73–78.DOI
  3. [3]R. Bastos and L. Mendonça, Generalized torsion elements in infinite groups, Proc. Amer. Math. Soc. 153 (2025), no. 12, 5171–5181.DOI
  4. [4]C. C. Schneider, and D. Silveira, Generalized torsion elements in groups, Arch. Math. (Basel) 122 (2024), no. 2, 121–131.DOI
  5. [5]R. Bastos and D. Silveira, Finiteness conditions and generalized torsion elements in groups, Comm. Algebra 54 (2026), no. 4, 1687–1697.DOI
  6. [6]J. A. Bondy and U. S. R. Murty, Graph Theory, Graduate Texts in Mathematics, vol. 244, Springer, London, 2008.
  7. [7]P. J. Cameron and S. Ghosh, The power graph of a finite group, Discrete Math. 311 (2011), no. 13, 1220–1222.DOI
  8. [8]I. Chakrabarty, S. Ghosh, and M. K. Sen, Undirected power graphs of semigroups, Semigroup Forum 78 (2009), no. 3, 410–426.DOI
  9. [9]G. Dhawlath and V. Raja, Some properties of proper power graphs in finite Abelian groups, Discrete Math. Algorithms Appl. 17 (2025), no. 8, 2550007.
  10. [10]A. Doostabadi and F. Farrrokhi D. Ghouchan, On the connectivity of proper power graphs of finite groups, Comm. Algebra 43 (2015), no. 10, 4305–4319.arxiv.org/abs/1401.6760
  11. [11]D. Gorenstein, Finite Groups, 2nd ed., AMS Chelsea Publishing, Providence, RI, 2007.
  12. [12]T. Ito, On a group whose generalized torsion elements are torsion elements, Comm. Algebra 52 (2024), no. 4, 1271–1276.arxiv.org/abs/2303.05726
  13. [13]T. Ito, K. Motegi, and M. Teragaito, Generalized torsion and decomposition of 3-manifolds, Proc. Amer. Math. Soc. 147 (2019), no. 11, 4995–5008.arxiv.org/abs/1811.07532
  14. [14]T. Ito, K. Motegi, and M. Teragaito, Generalized torsion for hyperbolic 3-manifold groups with arbitrary large rank, Bull. Lond. Math. Soc. 55 (2023), no. 3, 1203–1209.arxiv.org/abs/2112.00418
  15. [15]J. Kato and D. V. P. Silva, Generalized torsion in triangular matrix groups, preprint, arXiv:2606.10168, 2026.arxiv.org/abs/2606.10168
  16. [16]A. V. Kelarev and S. J. Quinn, A combinatorial property and power graphs of groups, Contributions to General Algebra, vol. 12, Verlag Johannes Heyn, Klagenfurt, 2000, pp. 195–200.
  17. [17]A. Kumar, L. Selvagonesh, P. J. Cameron, and T. T. Chelvam, Recent developments on the power graph of finite groups – a survey, AKCE Int. J. Graphs Comb. 18 (2021), no. 2, 65–94.DOI
  18. [18]A. Lucchini, A bound on the number of generators of a finite group, Arch. Math. (Basel) 53 (1989), no. 4, 313–317.DOI
  19. [19]A. R. Moghaddamfar, S. Rahbarivan, and W. J. Shi, Certain properties of the power graph associated with a finite group, J. Algebra Appl. 13 (2014), no. 7, 1450004.
  20. [20]D. J. S. Robinson, A Course in the Theory of Groups, 2nd ed., Graduate Texts in Mathematics, vol. 80, Springer-Verlag, New York, 1996.
  21. [22]J. Wiegold, Groups with boundedly finite classes of conjugate elements, Proc. Roy. Soc. London Ser. A 238 (1957), no. 1214, 389–401.DOI
  22. [23]D. Witte, Products of similar matrices, Proc. Amer. Math. Soc. 126 (1998), no. 4, 1005–1015.DOI

Paper details

vlnc:W_tszwbndt
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents