R ESEARCH ARTICLE ScienceAsia 41 (2015): 73–78

doi: 10.2306/scienceasia1513-1874.2015.41.073

Some results on the power graphs of finite groups

Doostabadi Alireza, Erfanian Ahmad∗ , Jafarzadeh Abbas
Ferdowsi University of Mashhad,
Department of Mathematics and Centre of Excellence in Analysis on Algebraic Structures, Mashhad, Iran
∗
Corresponding author, e-mail: [email protected]
Received 24 Apr 2013
Accepted 21 Feb 2014

ABSTRACT: We classify planar graphs and complete power graphs of groups and show that the only infinite group
with a complete power graph is the Prüfer group Z p∞ . Clique and chromatic numbers and the automorphism group of
power graphs are investigated. We also prove that the reduced power graph of a group G is regular if and only if G is
a cyclic p-group or exp(G) = p for some prime number p.

KEYWORDS: reduced power graph, planar graph, perfect graph, regular graph, automorphism group of a graph

INTRODUCTION Γ is defined as V (Γ ) = V (Γ ) and two vertices are

adjacent in Γ if and only if they are not adjacent in
For a semigroup or group, we can associate a Γ . A proper colouring of a graph Γ is a map from
graph in different ways to describe algebraic struc- V (Γ ) into some finite set of colours such that no two
tures of the group or semigroup using the prop- adjacent vertices are assigned the same colour. If
erties of graph. Bosak 1 studied certain graphs Γ can be properly coloured with a set of k colours,
over semigroups. The directed power graph of a then we say that Γ can be properly k-coloured. The
semigroup was defined by Kelarev and Quinn 2, 3 . minimum value of k for which Γ can be k-coloured
Recently, Chakrabarty, Ghosh and Sen 4 investigated is called the chromatic number of Γ and we denote
the power graph of semigroups and characterized it by χ(Γ ). The degree of vertex v in Γ is the number
the class of semigroups with a connected or com- of edges incident to v and we denote it by degΓ (v)
plete power graph. Also, Cameron 5 proved that or simply deg(v). A graph automorphism ϕ is a
for finite groups, the undirected power graph de- permutation on V (Γ ) such that u, v are adjacent if
termines the directed power graph up to an iso- and only if ϕ(u), ϕ(v) are adjacent. The set of all
morphism and two finite groups with isomorphic graph automorphisms of Γ is denoted by Aut(Γ ). A
undirected power graphs have the same number of clique is an induced subgraph of Γ that is complete.
elements of each order. The maximum order of a clique in Γ is called the
For a given group G we may define the directed clique number of Γ and is denoted by ω(Γ ).
power graph P~ (G) as a directed graph with vertex
set G in which there is an arc from x to y if PLANAR AND COMPLETE POWER GRAPHS
x 6= y and y = x m for some positive integer m. In this section, we shall characterize all groups with
The indegree (outdegree), denoted by indegΓ~ (u) a complete or planar power graph.
(respectively, outdegΓ~ (u)) of a vertex u in a directed
Definition 1 A group G is called torsion group if the
graph Γ~ is the number of arcs (v, u) (respectively,
order all of elements of G is finite.
(u, v)) in Γ~ . The undirected power graph P (G) is
defined in such a way that two distinct vertices x Lemma 1 Let G be a torsion group. If x and y are
and y are adjacent if one of them is a power of adjacent vertices in P (G), then either |x| divides | y|
the other one. For any graph Γ , we denote the set or | y| divides |x|. The converse is true whenever G is
of vertices and the edges of Γ by V (Γ ) and E(Γ ), a cyclic group.
respectively. If X ⊆ G then we can also define P (X )
as the induced subgraph on X of P (G). Moreover, For a finite nontrivial group G, Chakrabarty
|Γ | stands for the number of vertices of Γ and is et al proved that P (G) is complete if and only if
called the order of Γ . Two distinct vertices v1 , v2 G is a cyclic group of prime power order 4 . In the
are adjacent if they are joined by an edge in Γ and next theorem, we investigate a result for infinite
we denote this by v1 ∼ v2 . The complement graph groups. Recall that a group G is called locally cyclic

www.scienceasia.org
74 ScienceAsia 41 (2015)

if every finitely generated subgroup of G is cyclic.

Equivalently, G is locally cyclic if 〈x, y〉 is cyclic for
every pair x, y of elements of G. In particular, every
locally cyclic group is abelian. It is a well-known
result that a group is locally cyclic if and only if it is
isomorphic to a subgroup of Q or Q/Z (see Ref. 6).

Definition 2 The power graph P (G) of a group G,

is the graph whose vertex set is the group G such
that two distinct elements are adjacent if one is a
power of the other.
Fig. 1 exp(G \ A) = 2.
Theorem 1 Let G be an infinite group. Then P (G)
is complete if and only if G ∼
= Z p∞ for some prime p.

Proof : Since every proper subgroup of the Prüfer

group Z p∞ is a finite cyclic p-group, for every pair
of elements of G, one is a power of the other, which
implies that P (Z p∞ ) is an infinite complete graph.
Now suppose that P (G) is a complete graph
for some infinite group G. Clearly G is an abelian
torsion group, which is also a p-group. Since P (G)
is a complete graph, every pair x, y of G generates a
cyclic subgroup. Hence G is a locally cyclic p-group,
which implies G ∼ = Z p∞ . 2 Fig. 2 exp(G \ A) = 4.
In the following theorem, we give a necessary and
sufficient condition for the power graph P (G) to
be planar. Recall that by a well-known theorem exp(G \ A) = 2 or 4. If exp(G \ A) = 2, then P (G \ A)
of Kuratowski, a graph Γ is planar if it has no is a star graph and so P (G) is a graph containing
subdivision of the graphs K3,3 and K5 (see Ref. 7). some triangles and the above star graph with the
same identity (Fig. 1).
Theorem 2 Let G be a group. Then P (G) is planar Now suppose that exp(G \ A) = 4. Consider
if and only if G is a torsion group and πe (G) ⊆ B = {g ∈ G : |g| = 4}, C = {g ∈ G : |g| = 2} and
{1, 2, 3, 4}, where πe (G) = {|x| : x ∈ G}. D = C \ B {2} , where B {2} = {g 2 : g ∈ B}. For each
g ∈ B {2} , put B g = {g 0 ∈ B : g 02 = g}. Then we can
Proof : First assume that P (G) is a planar graph. We easily see that P (B g ∪ {1, g}) consists of some (not
claim that G is torsion group. If G has an element necessarily finitely many) subgraphs K4 with two
of infinite order, then we can easily find subgraphs common vertices, namely the identity and g, and
K3,3 or K5 in P (G), which is a contradiction. Now, two non-common vertices g 0 , g 0−1 ∈ B g . Thus P (G)
suppose on the contrary that there exists an element involves a star graph with the identity as midpoint,
x ∈ G such that |x| ¾ 5. If |x| is divisible by a some triangles and some complete graphs K4 as
prime p ¾ 5, then we have P (〈x |x|/p 〉) ∼= K p which above, all of them having the identity in common.
is impossible. Thus |x| = 2m 3n , where m and n are Hence P (G) is planar (Fig. 2). 2
non-negative integers. If n = 0, m ¾ 3 or m = 0,
n ¾ 2, then P (〈x〉) ∼ = K2m or K3n . Hence P (G) Corollary 1 If G is a group with a planar power
contains a subgraph isomorphic to K5 . Hence we graph, then χ(P (G)) = max πe (G).
may assume that m, n ¾ 1 and 6 divides |x| and
consequently P (〈x |x|/6 〉) has a subgraph isomorphic Proof : If P (G) is planar, then by Theorem 2,
to K3,3 , which is again a contradiction. Hence max πe (G) ¶ 4. If max πe (G) = 4, then P (G)
πe (G) ⊆ {1, 2, 3, 4}. Conversely, assume that G is is a graph as in Fig. 2 with a complete subgraph
a torsion group and πe (G) ⊆ {1, 2, 3, 4}. If x is an of order 4, which implies that χ(P (G)) = 4. If
element of order 3, then it is not adjacent to any max πe (G) = 3, then P (G) contains complete sub-
element of order 2 or 4 by Lemma 1. Therefore graphs of order 3 that have the identity in com-
P (〈x〉) ∼= K3 . Put A = {g ∈ G : |g| = 3}. Then mon. Thus χ(P (G)) = max πe (G) = 3. In the case

www.scienceasia.org
ScienceAsia 41 (2015) 75

max πe (G) = 2, P (G) is a star graph and clearly We can easily compute the degree of vertex x 0 in
χ(P (G)) = 2. 2 P ∗ (G): that is, (n/p1 ) − 2 + kϕ(n) where k is the
As a direct consequence of the above theorem, one number of maximal cyclic subgroups which contain
can deduce that the subgroups and quotients of a x 0 and ϕ is the Eulerian function. Since P ∗ (G) is
group with a planar power graph have planar power regular, (n/p1 ) − 2 + kϕ(n) = n − 2 which implies
graphs. The following theorem is a consequence that k(p2 − 1)(p3 − 1) · · · (pm − 1) = p2 p3 . . . pm . Also
of a theorem given in Ref. 8 which determines the p2 > 2 (p2 6= 1). If prime number q divides p2 − 1,
structure of all groups whose power graphs are then q | p2 p3 . . . pm . Thus q = pi for some 2 ¶ i ¶ m
planar. which is a contradiction. Hence G is a finite p-
group and P ∗ (G) is n − 2 regular. If n = p, then
Theorem 3 Let G be a group. Then P (G) is planar exp(G) = p. Suppose that n = p t > p. Let 〈x〉 and
if and only if G is locally finite and has one of the 〈 y〉 be distinct maximal cyclic subgroups of order n.
following structures. The regularity P ∗ (G) implies that these subgroups
(i) G is a group of exponent 2, 3, or 4. are disjoint, otherwise the element g ∈ 〈x〉∩〈 y〉 has
(ii) There is a normal elementary abelian 3-subgroup degree greater of n − 2 in P ∗ (G).
N of G such that G/N is isomorphic to a subgroup Suppose the maximal cyclic subgroups of G are
of Q 8 . not unique. Let z ∈ Z(G) be of prime order p. Then
(iii) There is a normal elementary abelian 2-sub- there exists a maximal cyclic subgroup 〈x〉 such that
group N of G such that G = N S where S ∼ = S3 . z∈ / 〈x〉. Now 〈z x〉 is a maximal cyclic group such
(iv) There is a normal 2-subgroup N of G of nilpo- that 〈z x〉 =6 〈x〉 and 〈z x〉 ∩ 〈x〉 = 6 〈1〉 which is a
tency class 2 such that |G/N | = 3. contradiction. Hence the maximal cyclic subgroup
of G is unique. Thus G is a cyclic group. 2
Example 1 For the quaternion group Q 8 , we can
see that its power graph is planar and it is clear that CLIQUE NUMBER AND CHROMATIC NUMBER
Q 8 is a group of exponent 4.
It is clear that χ(Γ ) ¾ ω(Γ ) for every graph Γ and
REGULARITY OF A POWER GRAPH WITHOUT a graph Γ is perfect if χ(Γ1 ) = ω(Γ1 ) for all induced
IDENTITY ELEMENT subgraphs Γ1 of Γ . From the strong perfect graph
theorem given in Ref. 9, a finite graph Γ is perfect if
The graph Γ is regular if the all vertices have the and only if neither Γ nor Γ contains an odd cycle of
same degree. We know that for a nontrivial finite length at least 5 as an induced subgraph. Utilizing
group G of order n the degree of the identity element this fact, we prove that the power graph of each
in the graph P (G) is n − 1. Thus P (G) is a regular group is perfect.
graph if and only if P (G) is a complete graph Kn−1 .
It is known that the power graph of G is complete Theorem 5 The power graph of a finite group is
if and only if G is a finite cyclic p-group, where p is perfect.
prime. The reduced power graph of a finite group Proof : Suppose that P (G) contains an induced
G is obtained when we remove the identity element cycle subgraph C of odd length at least 5. Let C~
from the vertex set and is denoted by P ∗ (G). be a related directed subgraph to C in the direct
power graph P~ (G). Then C~ must have a strong
Theorem 4 Let G be a finite group. The reduced
directed path of length two which makes a chordal
power graph P ∗ (G) is regular if and only if G is
in C which is a contradiction. Now suppose that C
isomorphic to the cyclic p-group or exp(G) = p, where
is an induced cycle subgraph of odd length at least 5
p is prime.
in P (G). If C is a cycle of length 5, then C is also a
Proof : Suppose that P ∗ (G) is regular and let 〈x〉 cycle of length 5 in P (G) which is not possible. Now
be the maximal cyclic subgroup of G of order n. It assume that C is a cycle of length at least 7. Then we
is clear that degP ∗ (G) (x) = n − 2. Since P ∗ (G) is can easily find a triangle in C. On the other hand,
regular, we infer that all the maximal cyclic sub- we should have indegC~ (x) = 0 or outdegC~ (x) = 0
groups of G have the same order. Every element for every vertex x in V (C) which is not true for the
of finite group G is in a maximal cyclic subgroup. above triangle. Thus there is no odd cycle of length
Hence exp(G) = |〈x〉| = n. By the contradiction, at least 5 as an induced subgraph in P (G). 2
assume that G is not a cyclic p-group. Then we can
a a Theorem 6 Let G be a finite group. Then
write n = p11 p22 . . . pm
am
where p1 < p2 < · · · 〈pm m〉1
and pi s are prime. Let x 0 ∈ 〈x〉 of order n/p1 .
0
ω(P (G)) = max{ω(P (Zn )) : n ∈ πe (G)}.

www.scienceasia.org
76 ScienceAsia 41 (2015)

‚ Œ ‚ Œ
Proof : Suppose that C = {x 1 , . . . , x m } ⊆ V (P (G)) n n
+ϕ +···+ϕ
induces a complete subgraph. Clearly 〈C〉 is an λ
p1 1 p2
λ
p1 1 p2 2
λ
abelian subgroup. Let Ci = {S pi (〈x〉) : x ∈ C}, where ‚ Œ
S pi (〈x〉) is the Sylow pi -subgroup of 〈x〉. Also, n
+···+ϕ +···
let |S pi (〈x ki 〉)| = max{|S pi (〈x〉)| : x ∈ C}. Since λ λ
p1 1 p2 2 . . . pr
〈S pi (x s ), S pi (x t )〉 is cyclic for every 1 ¶ s, t ¶ m, it ‚ Œ
is easy to see that 〈Ci 〉 = S pi (〈x ki 〉) is a cyclic group. n
+ϕ λ λ λ −1
+ ϕ(1),
Therefore 〈C〉 = 〈C1 , . . . , Cm 〉 is a cyclic group and p1 1 p2 2 . . . p r r
the result follows. 2
In what follows, we shall give a formula for the where ϕ is the Eulerian function.
clique number of a finite cyclic group.
Proof : Let Y be a clique in P (Zn ). We first ob-
λ
Lemma 2 Let n = p1 1 . . . pλr r be a natural number serve that if y ∈ V (Y ), then r y ∈ V (Y ) whenever
and m = λ1 + · · · + λ r . Let S be the set of all (m + 1)- gcd(r, | y|) = 1. Hence the elements of V (Y ) can
tuples (d0 , d1 , . . . , dm ) such that n = d0 > d1 > · · · > be partitioned into sets each of which contains
dm = 1 is a chain of divisors of n and di−1 /di is a elements of the same order. Thus V (Y ) = Yh1 ∪
prime for all i = 1, . . . , m. Let f : S −→ N be a map Yh2 ∪ · · · ∪ Yhk , where Yhi possesses of all elements
defined by f (d0 , . . . , dm ) = ϕ(d0 )+· · ·+ϕ(dm ). Then of order hi and |Yhi | = ϕ(hi ). By Lemma 1 and
f takes its maximum value at (d0 , . . . , dm ) if and only the fact that Y is a complete subgraph of P (Zn ),
if di−1 /di ¶ di /di+1 for all 0 < i < m, and it is unique it follows that for each i, j ¶ m, either hi | h j or
with this property. h j | hi . Without loss of generality, we assume that
h1 | · · · | hk . On the other hand, for any chain of
Proof : Suppose on the contrary that there exists D = positive divisors l1 , . . . , l t of nPsuch that l1 | · · · | l t ,
(d0 , . . . , dm ) ∈ S such that f (D) = max f (S ) and t
we can find a clique of size i=1 ϕ(l i ) in P (Zn ).
di−1 /di > di /di+1 for some i. Further, we assume Now suppose that X is a maximal clique in P (Zn ).
that i is a maximum with respect to this property Then V (X ) = X d0 ∪X d1 ∪· · ·∪X dm (d0 = n), where X di
for all C ∈ S with f (C) = max f (S ), i.e., possesses of all elements of order di , |X di | = ϕ(di ),
and dm | · · · | P d0 . Since X is a maximal clique,
c j−1 cj m
§
i = max j such that > , for some C ∈ S then |V (X )| = i=0 ϕ(di ) = ω(P (Zn )). Hence by
cj c j+1 Lemma 2, m = λ1 + · · · + λ r and di = max{h : h |
ª di−1 , h < di−1 }. 2
such that f (C) = max f (S ) . The following corollary gives a shortened for-
mula for ω(P (Zn )).
for all C = (c0 , . . . , cm ). Let p = di−1 /di and q =
di /di+1 . If di0 = pdi+1 and d 0j = d j whenever j 6= i, λ λ
Corollary 2 Let n = p1 1 p2 2 . . . pλr r such that p1 <
then p2 < . . . < p r . Then
D0 = (d00 , d10 , . . . , dm
0
)∈S
r+1
‚ Œ
and one can see that f (D ) ¾ f (D) and the equality
0 X n λ −1
ω(P (Zn )) = ϕ λ λ λi−1
i−1
pi−1 +ϕ(n).
holds if and only if p = 3, q = 2, q | di+1 and p - di+1 . i=2 p1 1 p2 2 . . . pi−1
By assumption f (D) = f (D0 ) and hence p = 3, q = 2,
q | di+1 and p - di+1 . But there exists j > i such that Proof :
d 0j−1 /d 0j > d 0j /d 0j+1 = 2, which contradicts the choice
of i. The uniqueness is a direct consequence of the ω(P (Zn )) = ϕ(1) + ϕ(p r ) + · · · + ϕ(pλr r −1 )
inequalities di−1 /di ¶ di /di+1 (0 < i < m), which λ
+ ϕ(pλr r )(ϕ(1) + ϕ(p r−1 ) + · · · + ϕ(p r−1
−1
hold for all D ∈ S admitting f (D) = max f (S ). The
r−1
))
λ r−1 λ r−2 −1
converse is clear. 2 + ϕ(pλr r p r−1 )(ϕ(1) + ϕ(p r−2 ) · · · + ϕ(p r−2 ))
λ λ +
Theorem 7 Let n = p1 1 p2 2 . . . pλr r with p1 < p2 <
· · · < p r . Then ..
.
λ λ −1
+ ϕ(pλr r . . . p2 2 )(ϕ(1) + ϕ(p1 ) + · · · + ϕ(p1 1 ))
‹  ‚ Œ
n n
ω(P (Zn )) = ϕ(n) + ϕ +···+ϕ λ + ϕ(n).
p1 p 1 1

www.scienceasia.org
ScienceAsia 41 (2015) 77

Thus Theorem 8 Q Aut(P (Zn )) has a subgroup isomorphic

λ
to Sϕ(n)+1 × d|n,d6=1,n Sϕ(d) .
ω(P (Zn )) = pλr r −1 + ϕ(pλr r )p r−1
r−1 −1

λ λ Corollary 4 Let n be a natural number such that for

+ ϕ(pλr r p r−1
−1
r−1
)p r−2
r−2
every x, y ∈ Zn , deg(x) 6= deg( y) whenever |x| 6= | y|.
λ λ λ −1
+ · · · + ϕ(pλr r p r−1
r−1
. . . p2 2 )p1 1 + ϕ(n). Then
Y
λ λ
Let ni = pi i pi+1
i+1
. . . pλr r and n r+1 = 1. Then Aut(P (Zn )) ∼= Sϕ(n)+1 × Sϕ(d) .
d|n,d6=1,n
r+1
λi−1 −1
X
ω(P (Zn )) = ϕ(ni )pi−1 + ϕ(n) Proof : If α ∈ Aut(P (Zn )) and |x| 6= | y|, then α(x) 6=
i=2 y. Hence α(Ai ) = Ai , for i = 0, 1, 2, . . . , r, where
r+1
Ai ’s are defined as above. Thus α ∈ SA0 × SA1 × · · · ×
‚ Œ
X n λ −1
= ϕ λ λ λi−1
i−1
pi−1 SA r . 2
i=2 p1 1 p2 2 . . . pi−1
As an example, if G = Z12 and x = 2, y = 6
+ ϕ(n). are elements of G. Then |x| = 6 and | y| = 2,
2 and deg(x) = deg( y) = 9. However, there is no
λ λ automorphism of P (Zn ) which maps Q x to y. Hence
Corollary 3 Let n = p1 1 p2 2 . . . pλr r such that p1 < we have Aut(P (Z12 )) ∼= Sϕ(12)+1 × d|12,d6=1,12 Sϕ(d) .
p2 < · · · < p r . Then The same happens for Z24 .
r+1
‚ Œ We state the following conjecture for
X n λ −1 Aut(P (Zn )). In spite the fact that we have
χ(P (Zn )) = ϕ λ λ λi−1
i−1
pi−1
i=2 p1 1 p2 2 . . . pi−1 much evidence for some small values of n and
+ ϕ(n). strongly believe that it is true for all n having at
least two distinct prime divisors, we have not able
Proof : By Theorem 5 and Corollary 2, the result is to prove it completely as yet.
clear. 2
Conjecture 1 For every natural number n,
AUTOMORPHISM GROUP OF POWER GRAPH Y
OF P (Zn ) Aut(P (Zn )) ∼
= Sϕ(n)+1 × Sϕ(d) .
d|n,d6=1,n
In Ref. 10, Cameron and Ghosh showed that the
only finite group whose automorphism group is the Finally, we compute the degree of all vertices in
same as that of its power graph is the Klein four P (Zn ).
group. In this section, we discuss the automorphism
group of P (Zn ). For any graph Γ and v ∈ V (Γ ), the Theorem 9 The degree of an arbitrary vertex x in
set of neighbours of the vertex v in Γ is denoted by P (Zn ) is
NΓ (v), or briefly by N (v). Furthermore, we define X
N [v] = N (v) ∪ {v}. deg(x) = d − 1 + ϕ(r d),
r d|n,r¾2
Lemma 3 If x, y ∈ V (P (Zn )) and |x| = | y|, then
N [x] = N [ y]. where d is the order of x in Zn .
Proof : As Zn has a unique subgroup of order |x|, we Proof : By Lemma 1, the number of elements of
have 〈x〉 = 〈 y〉 and the proof is straightforward. 2 order r d in Zn is ϕ(r d). The result follows immedi-
Now let {d1 , d2 , . . . , d r } be the set of all divisors of n ately. 2
other than 1 and n. Let A0 = {g ∈ Zn : |g| = n} ∪ {0} We may give a similar formula for the vertex
and Ai = {g ∈ Zn : |g| = di }, for i = 1, 2, . . . , r. Then degrees of P (Zn ) as follows.
by Lemma 1, the induced subgraphs of P (Zn ) on
Ai ’s are complete of order ϕ(di ), for i = 1, 2 . . . , r Corollary 5 Let x be a vertex in P (Zn ). Then
and the induced subgraph of P (Zn ) on A0 is a n
complete graph of order ϕ(n) + 1. Hence, by deg(x) = d − 1 − ϕ(d) + ϕ(d)σ(m) ,
dm
Lemma 3, every bijection α : Zn −→ Zn such that
where d is the order of x in Zn , m = p|d,p k kn/d p k
Q
α|Ai : Ai −→ Ai is a permutation for i = 0, 1, 2, . . . , r,
is an automorphism of P (Zn ). Thus we have the and σ(m) is the sum of divisors m. Also, p k k n/d
following theorem. means that dn /p k does not have divisor p.

www.scienceasia.org
78 ScienceAsia 41 (2015)

Proof : By Theorem 9, we have 8. Lytkina DV (2007) Structure of a group with ele-

X ments of order at most 4. Siberian Math J 48, 283–7.
deg(x) = d − 1 + ϕ(r d). 9. Chudnovsky M, Robertson N, Seymour P, Thomas R
r>1,r|n/d (2006) The strong perfect graph theorem. Ann Math
164, 51–229.
a a b b
Assume n = p11 Q p22 . . . p rar and d = p1 1 p2 2 . . . p rbr . 10. Cameron PJ, Ghosh S (2011) The power graph of a
ci finite group. Discrete Math 311, 1220–2.
Also, let r x = bi 6=0 pi for each divisor r =
c c
p11 p22 . . . p rcr of n. Then
b +c −1
Y Y
c −1
ϕ(r d) = pi i i (pi − 1) pi i (pi − 1)
bi 6=0 bi =0,ci 6=0
Y Y Y
c b −1 c −1
= pi i pi i (pi − 1) pi i (pi − 1)
bi 6=0 bi 6=0 bi =0,ci 6=0
 ‹
r
= r x ϕ(d)ϕ .
rx
Hence
‹ 
X X r
ϕ(r d) = r x ϕ(d)ϕ
r|n/d r|n/d
rx
 ‹
X r
= ϕ(d) rx ϕ
r|n/d
rx
 ‹
X r
= ϕ(d) rx ϕ
0 rx
r x |n x ,
(r/r x )|n0 /n0x
X   ‹
X r
= ϕ(d) rx ϕ
r x |n0x (r/r x )|n0 /n0
r x
x

n0
= ϕ(d)σ(n0x ) 0 ,
nx

where n0 = n/d. 2
REFERENCES
1. Bosak J (1963) The graphs of semigroups. In: Fiedler
M (ed) Theory of Graphs and its Applications: Proceed-
ings of the Symposium held in Smolenice in June 1963,
pp 119–25.
2. Kelarev AV, Quinn SJ (2000) A combinatorial prop-
erty and power graphs of groups. In: Contributions
to General Algebra, vol 12, pp 229–35.
3. Kelarev AV, Quinn SJ (2002) Directed graph and
combinatorial properties of semigroups. J Algebra
251, 16–26.
4. Chakrabarty I, Ghosh Sen MK (2009) Undirected
power graphs of semigroups. Semigroup Forum 78,
410–26.
5. Cameron PJ (2010) The power graph of a finite
group II. J Group Theor 13, 779–83.
6. Roland S (1994) Subgroup Lattices of Groups, 2nd
edn, Walter de Gruyter, Berlin, pp 12–3.
7. Kuratowski K (1930) Sur le problème des courbes
gauches en topologie. Fund Math 15, 271–83.

www.scienceasia.org