Power Graphs: a Survey

This article gives a survey of all results on the power graphs of groups and semigroups obtained in the literature. Various conjectures due to other authors, questions and open problems are also included.


Introduction
Many interesting results on the power graphs have been obtained recently. This article gives a survey of the current state of knowledge on this research direction by presenting all results and open questions recorded in the literature dealing with power graphs. Various conjectures formulated by other authors and several new open problems are also included. This is the first survey devoted to the power graphs.

Definition of power graphs
Let us begin with the definitions of the directed and undirected power graphs, which are the main concepts considered in this survey. In the literature, the power graphs have been defined and considered for groups and semigroups. Section 4 contains concise prerequisites on the terminology of the group and semigroup theories used in the results (cf. [30,79]). It is well known that the operation in every group or semigroup is power-associative, i.e., the powers of each element do not depend on the order in which the operation in the power is performed (see Section 4 for details). Therefore the following definition applies to all groups and semigroups.   The directed power graph could also be called the power digraph, although the latter term has not been used in the literature yet. We include Figures 1 and 2 with the directed power graph − → G (Z 6 ) and undirected power graph G (Z 6 ). For clarity, formulating all results in this article we use only complete terms 'directed power graph' and 'undirected power graph'. The shorter expression 'power graph' is used here only when we refer to both of these concepts simultaneously and in general discussions that may apply to both of these notions.

Historical comments
The concept of a directed power graph was first introduced and considered in [43] in the case of groups. For semigroups, it was first investigated in [47], and then in [44,46]. All of these papers used only the brief term 'power graph' to refer to the directed power graph, with the understanding that the undirected power graph is the underlying undirected graph of the directed power graph.
The undirected power graphs became the main focus of study in [24] and in the subsequent papers by P. J. Cameron et al. [22,23], which introduced the use of the brief term 'power graph' in the second meaning of an undirected power graph. For a group G, the digraph G * (G) was considered in [76] as the main object of study.
Thus, the readers may need to be aware that in the literature the brief term 'power graph' can mean either a 'directed power graph' or an 'undirected power graph' depending on which type of graphs is the main subject of study in each particular paper. For clarity, the present article states all results using only the complete terms 'directed power graph' and 'undirected power graph'.

Preliminaries and background information
We use standard terminology and refer the readers to [25,30,36,37,60,78,79] for more details, further bibliography and explanations.
Throughout Z and N denote the set of all integers and all positive integers, respectively, and N 0 = N ∪ {0} stands for the set of all natural numbers. The Euler totient function of a positive integer n is denoted by φ(n).
All graphs considered in this paper are assumed to be simple. In particular, they do not have loops. If D = (V, E) is a directed graph, then the underlying undirected graph of D is the undirected graph with the same set V of vertices, where two distinct vertices u and v are adjacent if and only if (u, v) ∈ E or (v, u) ∈ E.
Subsequent sections of this survey article are organised according to the graph theoretical properties being considered, and the definitions of these properties are reviewed in the following sections. The rest of this section contains concise preliminaries concerning groups and semigroups.
An operation or a binary operation on a set S is a mapping from the direct product S × S to S. If • is a binary operation on a set S and x, y ∈ S, then the image of the pair (x, y) under • is denoted by x • y, which is called the infix notation for the operation •. In additive terminology and notation the operation • is denoted by + and is called addition, so that the result of the operation is written down as x + y. In multiplicative terminology and notation the operation • is referred to as multiplication and is denoted by · so that the result of the operation is recorded as xy = x · y. In this survey we use multiplicative notation to discuss sets with a binary operation. A set with a binary operation is called a magma [19,29] or a groupoid [87].
An operation · on the set S is associative if it satisfies the following associative law: x(yz) = (xy)z for all x, y, z ∈ S. A semigroup is a set S equipped with an associative binary operation ·. For example, the set of positive integers is a semigroup with respect to addition, and at the same time it is a semigroup with respect to multiplication. If we want to distinguish between these two semigroups on the same set, we can indicate the corresponding operation explicitly by using the symbols (N, +) or (N, ·), respectively.
Let M = (M, ·) be a magma. A subset S of M is called a submagma of M if it is closed for the operation of M , i.e., if x, y ∈ S implies xy ∈ S. The intersection of all submagmas of M that contain S is the smallest submagma of M containing S. It is denoted by S . An operation · on M is said to be power-associative if the submagma x is associative, for every x ∈ M . In particular, every semigroup is power-associative.
An element e of a semigroup S is called an identity or an identity element of S if xe = ex = x for all x ∈ S. It is well known and easy to verify that a semigroup cannot have two distinct identity elements. This means that if a semigroup has an identity, then the identity is unique.
Let S be a semigroup with identity e, and let x ∈ S. An element y ∈ S is called an inverse of x if xy = yx = e. It is also well known and easy to prove that if an element x has an inverse, then the inverse is unique. The inverse of x is denoted by x −1 , as no ambiguity can arise. A group is a semigroup with identity where every element has an inverse.
Let G be a group. The cardinality |G| is also called the order of G. For n ∈ N, the cyclic group of order n can be defined as the group Z n = Z/nZ of residues modulo n.
A subgroup of G is a subset of G that is a group with the same identity and with respect to the same operation. In particular, every subgroup H of G is closed for the operation of the group, which means that xy ∈ H for all x, y ∈ H. Besides, if H is a subgroup of G, then x −1 ∈ H for all x ∈ H. A subgroup H of G is said to be normal if g −1 hg ∈ H for all h ∈ H, g ∈ G. If H is a subgroup of G, then we write H < G. If H is a normal subgroup of G, then we write H ✁ G. If H is a maximal subgroup of G, then we write H < max G.
For any subset T of G, the intersection of all subgroups of G that contain T is the smallest subgroup of G containing T . It is denoted by T and is called the subgroup generated by T in G. In particular, the set g = {g n | n ∈ Z} is the cyclic group generated by g in G.
The order of the element x in a group or a semigroup is defined as the cardinality of x . The order of x is denoted by o(x). A group is said to be torsion if all of its elements have finite order.
The set of all elements of G of order k is denoted by Ω k (G). The set of the orders of all elements of G is denoted by π(G) and is called the spectrum of G. The exponent of a finite group G is the least common multiple of the orders of all elements of G. It is denoted by Exp(G).
Let p be a prime. Then all elements with orders equal to a power of p form a subgroup of G called a primary component of G, or a p-primary component of G. The p-primary component of a group G is denoted by G p . A group G is called a p-group if G = G p , i.e., if the order of every element of G is a power of p.
A group G is said to be an EPO-group, if the orders of all nontrivial elements of G are prime. An EPPO-group is a group where the order of every nontrivial element is a power of a prime.
The dihedral group D 2n is defined by the following presentation The semi-dihedral group SD 2 n is presented by The symmetric group S n on a finite set of n letters is the group consisting of all permutations of these n letters with respect to the operation of composition of permutations, regarded as functions.
Let S = (S, ·) be a semigroup. An equivalence relation ̺ on S is called a congruence if it is compatible with the operation · of S, that is if (x, y) ∈ ̺ implies (xz, yz) ∈ ̺ and (zx, zy) ∈ ̺ for all x, y, z ∈ S. If ̺ is a congruence, then we can define an operation on the set S/̺ of all equivalence classes of ̺ by the rule [x][y] = [xy], where [x] stands for the equivalence class containing x. This means that the product of two classes is the class containing the product of representatives of these classes. This definition makes sense, because it is easy to verify that the resulting class [xy] depends only on the whole classes [x] and [y], but does not depend on the choice of particular representatives x and y. It follows that the set S/̺ is a semigroup too. It is called the quotient semigroup of S modulo ̺. Now let S be a semigroup, and let T be a subset of S. Then the intersection of all subsemigroups of S containing T is the smallest subsemigroup of S containing T . It is denoted by T and is called the subsemigroup of S generated by T . In particular, for any x ∈ S, the subsemigroup x coincides with the set {x n | n ∈ N}. The ideal (left ideal or right ideal) generated by T in S is the smallest subsemigroup I containing T and such that GI ∪ IG ⊆ I (respectively, GI ⊆ I, IG ⊆ I).
Let S be a semigroup with an ideal I. Then the Rees quotient semigroup S/I is the semigroup with zero θ obtained from S by identifying all elements of the ideal I with θ. If S has zero θ and I = {θ}, then S/I = S.
An element of a semigroup is said to be periodic if it has a finite order. A semigroup entirely consisting of periodic elements is called a periodic semigroup.
An operation on a set S is commutative if it satisfies the commutative law, i.e., if xy = yx for all x, y ∈ S. An abelian group is a group with commutative operation. A semigroup satisfying the commutative law is said to be commutative.
For a prime p, a group G is called an elementary abelian p-group if G is finite, abelian and every nontrivial element of G has order p. The quasicyclic p-group is the infinite group denoted by C p ∞ that has generators g 1 , g 2 , . . . such that g p 1 = e and g p i = g i−1 , for all i > 1, where e is the identity of the group. Quasicyclic groups are the only infinite abelian groups all subgroups of which are finite.
Let G be a group with a subgroup H. A right (left) coset of the subgroup H is a set of the form Hg (respectively, gH), where g ∈ G. The set of left cosets has the same cardinality as the set of right cosets, and this cardinality is called the index of H in G. If N is a normal subgroup, then the quotient group G/N is the set of all right cosets of N in G with the following operations: The center of a semigroup S is the set Z(S) = {z ∈ S | ∀x ∈ S : xz = zx} consisting of those elements that commute with all elements of S. The center of every group G is a normal subgroup of G. A group G is said to be center-by-finite if the quotient group G/Z(G) is finite.
Rees matrix semigroups are very well known in semigroup theory and play crucial roles in describing the structure of semigroups (cf. [30]). For examples of applications of these concepts, let us also refer to [1,55,56,54,57]. Suppose that G is a group, and I, Λ are nonempty sets. Let P = [p λi ] be a (Λ × I)-matrix with entries p λi ∈ G 0 , for all λ ∈ Λ, i ∈ I. The Rees matrix semigroup M 0 (G; I, Λ; P ) over G with sandwich-matrix P consists of zero θ and all triples (g; i, λ), where g ∈ G 0 , i ∈ I, λ ∈ Λ, where all triples (θ; i, λ) are identified with θ, and where multiplication is defined by the rule (g 1 ;

Eulerian graphs
A walk is a list v 0 , e 1 , v 1 , . . . , e k , v k of vertices and edges such that, for 1 ≤ i ≤ k, the edge e i has endpoints v i−1 and v i . A trail is a walk without repeated edges. An Eulerian circuit of an undirected graph is a closed walk that traverses every edge of the graph exactly once. An Eulerian graph is a graph with an Eulerian circuit. Theorem 6.1. ( [24]) Let n ≥ 3. Then the graph G (U n ) is not Hamiltonian for n = 2 m p 1 · · · p k , where p 1 , . . . , p k are distinct Fermat primes, m and k are nonnegative integers, m ≥ 2 for k = 0, 1, and k ≥ 2 for m = 0, 1.
It was conjectured in [24, p. 426] that the undirected power graph G (U n ) is Hamiltonian for all values n ≥ 3 with the exception of n = 2 m p 1 · · · p k , where p 1 , . . . , p k are distinct Fermat primes, m and k are nonnegative integers, m ≥ 2 for k = 0, 1 and k ≥ 2 for m = 0, 1. The following collection of counterexamples to this conjecture was given in [77,Lemma 3].
where p is a Fermat prime, then the undirected power graph G (U n ) does not have a Hamiltonian cycle.

Complete graphs
An undirected graph D = (V, E) is said to be complete if any two distinct vertices of D are adjacent.
The number of edges of the undirected power graph G (Z n ) is given by ) Among all finite groups of a given order, the cyclic group of that order has the largest number of edges in its directed power graph.
The undirected power graph G (Z p n ) has the maximum number of edges among all undirected power graphs of p-groups of order p n .
A conjecture was recorded in [70, Conjecture 2] that Theorem 8.3 remains valid for finite groups of any order. This conjecture was proved in [75]. Several criteria for the nonsimplicity of G have been obtained in [76].

Chromatic number
The chromatic number χ(G) of an undirected graph G is the smallest positive integer k such that the vertices of G can be coloured in k colours so that no two adjacent vertices share the same colour. A graph G is said to be planar if it can be drawn on a plane without any crossing of its edges. (i) n = p m , for some prime p, m ∈ N and n ≥ 5; (ii) n = 2p, for some odd prime p.
A full exponent group is a group G such that Exp(G) = o(x), for an element x ∈ G.

Conjecture 1. ([70]) Theorem 9.3 remains valid for all finite groups.
A complete list of the chromatic numbers of the undirected power graphs of nonabelian groups of order up to 14 are determined and listed in [26]. 1 · · · p a k k , where p 1 < · · · < p k are prime numbers and a 1 , . . . , a k ∈ N 0 . Then the clique number and chromatic number of the undirected power graph G (Z n ) are determined by the following formula In order to describe the power graphs of all finite abelian groups, we take any finite abelian group G and any elements a, b in G, and introduce the following notation. Denote the primary components of G by G p 1 , . . . , G pn , and express each G p i as a direct product of cyclic groups . . , n, denote the projections of a and b on G p i , by a i and b i , respectively. Choose generators g i,j,k in the cyclic groups of   The undirected power graph G (SD 2 n ) is a union of a complete graph of order 2 n and 2 n copies of K 2 that share the identity vertex. For n > 3, this graph is neither Eulerian, nor Hamiltonian, nor nonplanar. Moreover, χ(G (SD 2 n )) = ω(G (SD 2 n )) = α(G (SD 2 n )) = 2 n . Theorem 11.2. ( [70]) Suppose that n = p b 1 1 · · · p b k k is the prime power decomposition of n and m = b 1 + · · · + b k . Then α(G (Z n )) is the coefficient of the middle term or the two middle terms of m j=1 (1 + x + · · · + x b j ).
It is easily seen that the undirected power graph of a p-group G is a union of complete graphs of order p sharing the identity vertex if and only if G has exponent p. The following theorem describes all finite groups with this property.

Isomorphism
The following theorem shows that undirected power graphs can characterize all finite simple groups.
Theorem 12.1. ([70]) Let G 1 be a finite group that is a simple group, or a cyclic group, or a symmetric group, or a dihedral group, or a generalised quaternion group. If G 2 is a finite group such that G (G 1 ) ∼ = G (G 2 ), then G 1 ∼ = G 2 .
Theorem 12.2. ([23]) Let G 1 and G 2 be finite abelian groups such that G (G 1 ) ∼ = G (G 2 ). Then It is explained in [23] that Theorem 12.2 does not generalize to infinite abelian groups. Indeed, let C p ∞ be the group Q p /Z, where p is a prime and Q p is the group of rational numbers with p-power denominators. Then G (C p ∞ ) is a countably infinite complete graph, independent of the chosen prime.
It is also impossible to drop the condition that G be abelian from Theorem 12.2 either. Indeed, let G be a finite group of exponent 3, that is, satisfying x 3 = e for all x ∈ G. Then the undirected power graph G (G) consists of (|G| − 1)/2 triangles sharing the identity vertex. The elementary abelian group (the direct product of cyclic groups of order 3) has exponent 3. However, there are nonabelian groups of exponent 3 as well. The smallest one is the group of order 27 with presentation where [x, y] stands for the commutator x −1 y −1 xy. It is noted in [23] that a short calculation with GAP [85] and its package GRAPE [83] revealed that there are two pairs of groups of order 16 with isomorphic undirected power graphs, and two pairs of groups of order 27. As the order of the group increases, isomorphic power graphs become more common. For example, for order 32 there are one quadruple, two triples, and eight pairs. It is also possible to use Magma [19] for experimental verification of properties of this sort. It is noted in [23] that this conjecture has been verified computationally using GAP [85] for all groups of order up to 511. The converse to the conjecture fails for groups of order 16. (b) G is cyclic of order n that is not a power of a prime, and S consists of the identity and the generators of G, so that |S| = 1 + φ(n); (c) G is generalised quaternion, and S contains the identity and the unique involution in G, so that |S| = 2.

Connected graphs
Let us define a binary relation ̺ on S by a̺b ⇔ a m = b m , for some m ∈ N. A description of all idempotents of the semigroup (Z m , ·) of all integers modulo m with respect to multiplication is given in [24, §3]. In particular, it is shown that the graph G ((Z m , ·)) is disconnected for all m > 1 and that it is Eulerian if and only if n = 2. The minimum number κ(D) of vertices of D which need to be removed to make the remaining graph disconnected is called the connectivity of D. If G is finite group, then we define For any element x of G, put r(x) = ∪ y∈M (G)− x ( x ∩ y ).

Theorem 13.3. ([70]) Let G be a finite group that is not a cyclic group. Then
Theorem 13.4. ( [27]) For any finite cyclic group Z n , the connectivity κ(G (Z n )) satisfies the following conditions: (i) κ(G (Z n )) = n − 1 when n = 1 or n = p m for a prime p and m ∈ N.
(ii) κ(G (Z n )) = φ(n) + 1 when n is not a power of a prime number. The equality holds for n = pq, where p, q are distinct primes.

2-connected graphs
A cut vertex in an undirected graph is a vertex whose removal increases the number of connected components of the graph. A graph is said to be 2-connected if it does not have a cut vertex. It is noted in [77] that the converse to Theorem 14.2 is not true. Indeed, the group G = Z 5 × S 3 is not nilpotent, but its undirected power graph is 2-connected. Besides, the dihedral group D 2p , where p is a prime, is a solvable group such that the undirected power graph G (D 2p ) is not 2-connected. Furthermore, the following proposition holds.

Diameter
The distance d(x, y) between vertices x and y of a connected graph D = (V, E) is the length of a shortest path connecting them. The maximum possible distance in D is called the diameter of D and is denoted by d(D). For example, in [10] a voltage assignment approach was used to construct large digraphs with small diameter.
The deletion of a single edge or vertex from a graph can change its diameter. Since the identity element of a group is adjacent to all other vertices of the undirected power graph of the group, one can delete any vertex, which is not the identity of the group, from the graph without changing its diameter. It would be interesting to determine when the same property holds for the identity vertex too. The following theorem gives a complete description of the structure of finite groups satisfying this property.

Girth
The girth of an undirected graph G is denoted by g(G) and is defined as the length of a shortest cycle in G.

Combinatorial properties expressed as unavoidable regularities
Combinatorial properties of groups and semigroups defined in terms of all infinite subsets enjoying certain unavoidable regularities have been investigated by many authors, and a survey of this direction of research has been given in [33], (cf. [31,32,41,53]).
The following combinatorial property of this sort has been defined in terms of directed power graphs in [43]. Let D = (V, E) be a directed graph. A semigroup S is said to be power Dsaturated if and only if, for every infinite subset T of S, the directed power graph of S has a subgraph isomorphic to D with all vertices in T .
A directed graph is said to be acyclic if it has no directed cycles. Denote by T ∞ the transitive tournament with the vertex set N and the edge set If S is finite or if D is a null graph, then it is obvious that S is power D-saturated. Therefore the following theorems deal only with infinite groups or semigroups and with directed graphs which have edges. For a skew field K, the set of all n × n matrices with entries in K is denoted by M n (K). A matrix is said to be monomial if every row and column contains at most one nonzero entry. If G is a group, then the set of all n × n monomial matrices over G 0 = G ∪ {0} forms a semigroup denoted by M n (G). Thus a matrix semigroup is a subsemigroup of M n (K) or M n (G), for some n, K and G. (ii) D is acyclic and S 0 has a finite ideal series where every infinite Rees quotient I j /I j−1 is a Rees matrix semigroup that has a finite sandwich matrix with entries in a center-by-finite torsion group H j such that each primary component of the center of H j is finite or quasicyclic, the center of H j has only a finite number of primary components, and the index of the center is not divisible by p for each quasicyclic p-subgroup of H j .

Open questions
A Moore graph or digraph is a graph or digraph that meets the Moore bound or directed Moore bound, respectively. Let us refer to the survey [69] and articles [11,12,13,14,15,16,17,81,82,88] for more information and examples of previous results pertaining to the Moore graphs and Moore bound. Problem 2. Does there exist any undirected or directed power graph of a group or a semigroup, which is a Moore graph or digraph? Determine for which positive integers k there exist directed power graphs that are at distance k from achieving the optimality in terms of the Moore bound. Furthermore, for each positive integer k, describe all (a) groups and (b) semigroups whose directed or undirected power graphs are at distance k from achieving the Moore bound.
Let us refer to the survey [21] for previous results on graceful labellings.
Problem 3. Describe all groups and semigroups whose undirected power graphs have graceful labellings.
Edge-magic total labellings of graphs were considered, for example, in [86]. We refer to the survey [6] and papers [7,8] for information on the edge antimagic labellings of graphs.
Problem 5. For each undirected power graph determine the number of edge antimagic labellings of the graph.
Let us refer to [9,84] for preliminaries on super edge-antimagic labellings of graphs. Problem 6. For each undirected power graph of a group or a semigroup, determine the number of super edge-antimagic labellings of the graph.
The readers are referred to [28] for preliminaries on group distance magic labellings and group distance antimagic labellings.
Problem 7. Investigate group distance magic labellings and group distance antimagic labellings of power graphs.
The readers are referred to [18] for background information on the Ramsey numbers of graphs. Let us refer to [3,4,5] for more information concerning the face antimagic labellings of planar graphs.

Problem 9. For each planar undirected power graph of a group or a semigroup, determine the number of face antimagic labellings of the graph.
It is also essential to investigate the relations of power graphs and other important classes of graphs associated to groups and semigroups. One of the largest classes of graphs of this sort is that of Cayley graphs. Let G be a semigroup, and let S be a nonempty subset of G. As in [42], we say that the Cayley graph Cay(G, S) of G relative to S is defined as the directed graph with vertex set G and edge set E(S) consisting of those ordered pairs (x, y) such that sx = y for some s ∈ S. The set S is called the connection set of the Cayley graph Cay(G, S). The Cayley graph Cay(G, S) is also called the Cayley digraph. A Cayley graph Cay(G, S) is said to be undirected if, (u, v) ∈ E(S) implies (v, u) ∈ E(S). On the other hand, the underlying undirected graph ← − → Cay(G, S) of Cay(G, S) can be called the undirected Cayley graph of G relative to S.
Problem 10. Describe all directed and undirected power graphs of groups and semigroups that can be represented as Cayley graphs and, respectively, undirected Cayley graphs of groups or semigroups.
It would be also interesting to investigate the power graphs of ring constructions (cf. [36]). The most important ring constructions are the group rings, crossed products, group-graded rings and semigroup-graded rings (cf. [36,40,73,74]). Let S be a semigroup. An associative ring R is said to be S-graded if R = s∈S R s is a direct sum of additive groups R s , and R s R t ⊆ R st , for all s, t ∈ S. The set H(R) = ∪ s∈S R s of homogeneous elements of R is a semigroup with respect to multiplication. If G is a group and R = g∈G R g is G-graded, then R is called a group-graded ring. If e is the identity of the group G, then the component R e is a subring of R. Group rings and crossed products are special cases of group-graded rings, and so the set H(R) is defined for these constructions too. Let us formulate the following general problems. x if x, y ∈ V and (x, y) ∈ E, 0 otherwise.
Problem 13. Describe all directed graphs whose graph algebras are power-associative.
The following problem makes sense for directed graphs and undirected graphs with appropriate adjustments.

Problem 14.
For each power-associative graph algebra Alg(D), determine the girth, diameter, connectivity, and the chromatic number of its directed power graph − → G ( Alg(D)) and undirected power graph G ( Alg(D)).
Problem 15. Find conditions necessary and sufficient for the directed power graph − → G ( Alg(D)) and undirected power graph G ( Alg(D)) to be planar.
Problem 16. Describe all directed graphs D such that the graph algebra Alg(D) is powerassociative and its undirected power graph has a graceful labelling.