Available online at www.sciencedirect.

com

Electronic Notes in Discrete Mathematics 63 (2017) 333–342 www.elsevier.com/locate/endm

On self-centeredness of tensor product of some

graphs

Priyanka Singh 1
Department of Mathematics
Indian Institute of Technology Kharagpur
Kharagpur, India

Pratima Panigrahi 2
Department of Mathematics
Indian Institute of Technology Kharagpur
Kharagpur, India

Abstract
A simple connected graph G is said to be a self-centered graph if every vertex has
the same eccentricity. Tensor product of self-centered graphs may not be a self-
centered graph. In this paper we study self-centeredness property of tensor product
of cycles with themselves and other graphs, complete graphs with themselvesand
other self-centered graphs, wheel graphs with themselves and other self-centered
graphs, and square of cycles with themselves.
Keywords: Eccentricity, Radius, Diameter, Self-centered graph, Tensor Product.

1
Email: priyankaiit22@gmail.com
2
Email: pratima@maths.iitkgp.ernet.in

https://doi.org/10.1016/j.endm.2017.11.030
1571-0653/© 2017 Elsevier B.V. All rights reserved.
334 P. Singh, P. Panigrahi / Electronic Notes in Discrete Mathematics 63 (2017) 333–342

1 Introduction
The concept of self-centered graphs and its applications have been widely
discussed in [2]. One can find the application of self-centered graphs in facil-
ity location problems where the resources (facility) must be located at cen-
tral node for its efficient use. Tensor product of graphs was introduced by
Whitehead and Russell [9]. Modeling of internet graphs is one of the ap-
plications of tensor product of graphs [4]. In this paper we consider sim-
ple and connected graphs only. For any two vertices u and v in a graph
G, the length of a shortest u − v path is known as the distance between u
and v and is denoted by d(u, v). The eccentricity of a vertex v in G, de-
noted by e(v), is defined as the distance between v and a vertex farthest
from v in G, i.e., e(v) = max{d(v, u) : u ∈ V (G)}. The radius rad(G)
and diameter diam(G) of graph G are respectively the minimum and maxi-
mum eccentricity of the vertices, i.e., rad(G) = min{e(v) : v ∈ V (G)} and
diam(G) = max{e(v) : v ∈ V (G)}. A graph G is called a self-centered graph
if eccentricity of every vertex is the same. Further, G is called a d-self-centered
graph if the eccentricity of every vertex is d. The kth power Gk of a graph G
is the graph on the same set of vertices as G and any two distinct vertices are
adjacent if and only if the distance between them is at most k in G.
T ensor product of n graphs G1 , . . . , Gn , denoted by G1 ⊗ G2 ⊗ . . . ⊗ Gn , is
the graph G with vertex set V (G) = {(x1 , x2 , . . . , xn ) : xi ∈ V (Gi )} and two
vertices x = (x1 , . . . , xn ) and y = (y1 , . . . , yn ) are adjacent in G if and only if
xi ∼ yi in Gi for every 1 ≤ i ≤ n. The distance between any two vertices x
and y in G is given by [4],

d(x, y) = min{k|each Gi has an xi − yi walk of length k for 1 ≤ i ≤ n}. (1)

Any two walks in a graph are said to have the same parity if the dif-
ference of their lengths is even; otherwise they have opposite parity. For
any two vertices x and y (not necessarily distinct) in a graph G, the upper
distance [1] between x and y, denoted by D(x, y), is the minimum length
of an x − y walk whose parity differs from a shortest x − y path. If no
such walk exists, then we take D(x, y) = ∞. The upper eccentricity [1]
E(x) of a vertex x ∈ V (G) is defined as E(x) = max{D(x, y) : y ∈
V (G), and y need not be different from x}.
For any two vertices x and y in a bipartite graph G there is no x − y walk
in G which is opposite parity with a shortest x − y path in G. So we have the
following result.
 P. Singh, P. Panigrahi / Electronic Notes in Discrete Mathematics 63 (2017) 333–342 335

Proposition 1.1 For any vertex x in a bipartite graph G the upper eccentric-
ity E(x) of x is ∞.
The eccentricity of a vertex x = (x1 , . . . , xn ) in G1 ⊗ . . . ⊗ Gn is given by [1],
e(x) = μ{e(x1 ), . . . , e(xn ), E(x1 ), . . . , E(xn )}, (2)
where μ is a function defined on a finite set X of real numbers as:

second largest element of X, if max(X) has multiplicity 1,
μ(X) = (3)
max(X) − 1, otherwise.
In [2], [3] and [6], authors have broadly studied and surveyed self-centered
graphs. Several algorithms are described to construct self-centered graphs in [7].
Stanic [8] worked on the self-centeredness of Cartesian and complete product of
graphs, and also discussed on minimal self-centered graphs. In [5], authors have
studied the self-centeredness of extensions of cycles by adding an edge. Also, self-
centeredness property of tensor product of graphs were not studied before.
It may be noted that tensor product of self-centered graphs is not always a self-
centered graph, as given in the example below.
Example 1.1 Here we consider tensor product of two 2-self-centered graphs G and
H as shown in Figure 1. Now consider the vertices (a, 1) and (a, 2) in G ⊗ H. From
equations 2 and 3, we get e(a, 1) = 4 and e(a, 2) = 5. Thus, eccentricity of vertices
(a, 1) and (a, 2) are not equal and hence G ⊗ H is not a self-centered graph.

Fig. 1. Two 2-self-centered graphs

In this paper, we discuss self-centeredness property of tensor product of some

self-centered graphs with themselves and with some other self-centered graphs as
well as non-self-centered graphs. More specifically, we discuss self-centeredness
property of tensor product of cycles with themselves, cycles with self-centered bipar-
tite graphs, cycles with paths, complete graphs with arbitrary self-centered graphs,
wheel graphs with themselves and arbitrary self-centered graphs, and square of
cycles with themselves.
In the following theorem, we find upper eccentricities of vertices belonging to
some graphs like wheel graphs Wn , paths Pn , complete graphs Kn , and cycles Cn .
Recall that a wheel graph Wn on n vertices, n ≥ 4, consists of a cycle C on n − 1
vertices and a vertex v such that every vertex of C is adjacent to v. Moreover, the
vertex v is called the central vertex and C is called the circumf erence of Wn .
Theorem 1.1 For an arbitrary vertex x in Wn , Pn , Kn or Cn , the upper eccen-
tricity of x is given by
336 P. Singh, P. Panigrahi / Electronic Notes in Discrete Mathematics 63 (2017) 333–342

⎧
⎪
⎪ 3, if x ∈ V (Wn ),
⎪
⎪
⎪
⎪
⎨∞, if x ∈ V (Pn ),
E(x) = 3, if x ∈ V (Kn ) and n ≥ 3,
⎪
⎪
⎪
⎪ n, if x ∈ V (Cn ) and n is odd,
⎪
⎪
⎩
∞, if x ∈ V (Cn ) and n is even.

Proof. Since paths and even cycles are bipartite graphs, the result is true for them
(by Proposition 1.1). Next, let x ∈ V (Wn ). If x is the central vertex then for any
vertex y in Wn − {x}, D(x, y) = 2 and D(x, x) = 3 because d(x, y) = 1, d(x, x) = 0
and there is vertex z on the circumference of Wn different from y such that xzy
and xzyx are walks in Wn . Therefore, E(x) = 3 in this case. If x is not the central
vertex v then for any vertex y in Wn − {x}, d(x, y) = 1 or d(x, y) = 2. If y = v then
there is a vertex z on the circumference of Wn such that xzw is a walk of length
two. If d(x, y) = 1 and y = v then xvy is a shortest x − y walk having different
parity with d(x, y). Therefore D(x, y) = 2 whenever d(x, y) = 1. If d(x, y) = 2,
then xvzy is the smallest x − y walk where z is a vertex on the circumference of
Wn and adjacent to y. So D(x, y) = 3 in this case. Finally, D(x, x) = 3 because
xvzx is a walk where z is a neighbor of x in the circumference of Wn and therefore
E(x) = 3.
Now let x be a vertex of the complete graph Kn . Then d(x, y) = 1 for x = y
and for any vertex z other than x and y, we get that xzy is a walk of length two,
and thus D(x, y) = 2. Again, d(x, x) = 0 and xzyx is a walk of length three and
thus D(x, x) = 3. This shows that E(x) = 3 for any vertex x in Kn . Finally, we
consider the case of an odd cycle Cn . Let x ∈ V (Cn ). Obviously, D(x, x) = n.
For any vertex y ∈ V (Cn ) − {x}, D(x, y) = n − d(x, y) which is less than n. Thus
E(x) = n. 2

2 Tensor product of cycles with themselves and other

graphs
We first study the self-centeredness property of tensor product of odd cycles with
themselves and with an arbitrary graph, and then tensor product of odd cycles with
at least one even cycle.
Theorem 2.1 Let T = Cn1 ⊗ Cn2 ⊗ . . . ⊗ Cnm be the tensor product of m odd
cycles and nk = max{ni : 1 ≤ i ≤ m}. Then T is d-self-centered graph where
d = μ{n1 , . . . , nm }.

Proof. This can be easily proved by using equations 2, 3, and Theorem 1.1. 2

Theorem 2.2 Let T = Cn1 ⊗ . . . ⊗ Cnm−1 ⊗ H be the tensor product of odd cycles
Cni , 1 ≤ i ≤ m − 1, and a bipartite k-self-centered graph H. Then T is a d-self-
 P. Singh, P. Panigrahi / Electronic Notes in Discrete Mathematics 63 (2017) 333–342 337

centered graph where d = max{k, ni : 1 ≤ i ≤ m − 1}.

Proof. Consider an arbitrary vertex x = (x1 , . . . , xm ) in T where xi ∈ V (Cni ),

i = 1, . . . , m − 1 and xm ∈ V (H). Since H is a bipartite graph, E(xm ) = ∞
(by Proposition 1.1) and e(xm ) = k as H is a k-self-centered graph. Now us-
ing Theorem 1.1, we have e(x) = μ{ n1 /2
, . . . , nm−1 /2
, k, n1 , . . . , nm−1 , ∞} =
max{ n1 /2
, . . . , nm−1 /2
, k, n1 , . . . , nm−1 } = max{k, ni : 1 ≤ i ≤ m − 1}. 2
Corollary 2.1 Let T = Cn1 ⊗ . . . ⊗ Cnm−1 ⊗ H be the tensor product of m − 1
odd cycles Cni , 1 ≤ i ≤ m − 1, and a bipartite non-self-centered graph H with
diameter not more than nk , where nk = max{ni : 1 ≤ i ≤ m − 1}. Then T is a
nk -self-centered graph.

Proof. The result follows because eccentricity of a vertex is at the most diameter
of the graph. 2

In the lemma below we discuss about all possible walks of different lengths
between pair of vertices in an even or an odd cycle.
Lemma 2.1 (a) Let Cn be an odd cycle. Then for any x, y ∈ V (Cn ), there are x−y
walks of even lengths as well as x − y walks of odd lengths in Cn . Moreover,
length of these walks are given as below:

d(x, y) + 2t, t ≥ 0, if d(x,y) is even
Length of an even x − y walk in Cn =
n − d(x, y) + 2t, t ≥ 0, if d(x,y) is odd


n − d(x, y) + 2t, t ≥ 0, if d(x,y) is even
Length of an odd x − y walk in Cn =
d(x, y) + 2t, t ≥ 0, if d(x,y) is odd
where t is a non-negative integer.

(b) Let Cn be an even cycle. For x, y ∈ V (Cn ) we have the following.

(i) If d(x,y) is even then x − y walk of every even length ≥ d(x, y) + 2k
is possible and no x − y walk of odd length is there in Cn where k ≥ 0
is a non-negative integer.
(ii) If d(x,y) is odd then x − y walk of every odd length ≥ d(x, y) + 2k,
is possible and there is no x − y walk of even length in Cn where k
is a non-negative integer.
Proof.

(a) Let Cn be an odd cycle. For x, y ∈ V (Cn ), let P be the shortest x − y path in
Cn . So length of P is d(x, y). If d(x, y) is even (odd) then Cn − (P − {x, y}) is
an odd (even) x − y walk of length n − d(x, y). If d(x, y) is even (odd) then by
traversing P or Cn − (P − {x, y}) and then walking back and forth along the
338 P. Singh, P. Panigrahi / Electronic Notes in Discrete Mathematics 63 (2017) 333–342

edge zy, where z is the neighbor of y in P or Cn − (P − {x, y}), we get walks of

any even (odd) length greater than d(x, y) or a walk of any odd (even) length
greater than n − d(x, y).

(b) Let Cn be an even cycle. If for any two vertices x and y, d(x, y) is even (odd)
then only even (odd) length walks are possible between x and y because n is
even. In the similar way as in (a), if P is a shortest x − y path in Cn then
we can achieve x − y walks of every even (odd) length greater than d(x, y) by
walking back and forth along the edge yz, where z is the neighbor of y in P .
2

In our next theorem we shall use a result by Hammack et al. [4] as given below.

Proposition 2.1 [4] If T is the tensor product of n connected non-trivial graphs

and k of them are bipartite then T has 2k−1 components.
In the following theorem we discuss self-centeredness property of tensor product
of both even and odd cycles. And as a particular case of this, we get self-centeredness
property of tensor product of even cycles only.
Theorem 2.3 Let T = Cn1 ⊗ Cn2 ⊗. . .⊗Cnk ⊗. . .⊗Cnm be the tensor product of m
cycles with ni ≥ 4, 1 ≤ i ≤ m, Cn1 , Cn2 , . . . , Cnk be even cycles and Cnk+1 , . . . , Cnm
be odd cycles for some 1 ≤ k ≤ m. Suppose nl = max{ni : 1 ≤ i ≤ k}. Then total
number of components in T is 2k−1 and each component is a d-self-centered graph,
where d is given by

max{ n2l , nk+1 , . . . , nm }, if n2l is odd
d=
max{ n2l , nk+1 − 1, . . . , nm − 1}, if n2l is even

Proof. Let V (Ci ) = {1, 2, . . . , ni }, 1 ≤ i ≤ m, be the vertex sets of given cycles.

Since Cn1 , Cn2 , . . . , Cnk are even cycles, by Proposition 2.1 T is a disconnected graph
with 2k−1 components. We shall find all these 2k−1 components and then show that
each one of them is a self-centered graph.
Consider the collection F of induced subgraphs of T , where F = {G0 } ∪ {Gt1 :
1 ≤ t1 ≤ k} ∪ {Gt1 t2 : 1 ≤ t1 < t2 ≤ k} ∪ . . . ∪ {Gt1 t2 ...tk/2 : 1 ≤ t1 < t2 . . . ≤
tk/2 ≤ k} where the vertex set of the subgraphs in F are as given below:
V (G0 ) = {(x1 , . . . , xm ) | each xi is an odd integer or each xi is an even integer, for
1 ≤ i ≤ k}.
V (Gt1 ) = {(x1 , . . . , xm ) | only xt1 is even (odd) and other coordinates are odd
(even)}, for 1 ≤ t1 ≤ k.
V (Gt1 t2 ) = {(x1 , . . . , xm ) | both xt1 and xt2 are even (odd) and other coordinates
are odd (even)}, for 1 ≤ t1 < t2 ≤ k.
..
.
 P. Singh, P. Panigrahi / Electronic Notes in Discrete Mathematics 63 (2017) 333–342 339

V (Gt1 t2 ...tk/2 ) = {(x1 , . . . , xm ) | xt1 , xt2 . . . , xtk/2 are all even (odd) and others are
odd (even)}, for 1 ≤ t1 < t2 . . . < tk/2 ≤ k}.
Obviously F is a partition of V (T ). Further, if any two vertices x = (x1 , . . . , xm )
and y = (y1 , . . . , ym ) belong to the same element of F then up to k th position all xi
and yi are of the same parity or different parity, and if x and y are not in the same
element of F then for some i parity of xi and yi differ, but not for all, 1 ≤ i ≤ k.
If N is the total number of elements in F then we have N = 2k−1 , because
k k k 
0 + 1 + . . . + k/2 , when n is odd
N =  k  k   
1 k
0 + 1 + . . . + 2 k/2 , when n is even
   k 
and as ki = k−i , 2N = 2k .
Next we shall show that every element in F induces a connected subgraph in
T and no vertex in any element in F has a neighbour in any other element in F.
In other words, we show that for any two vertices x and y in T there is a path or
no path between x and y depending on x and y belong to the same element or not,
and this will imply that every element in F is a connected component of T .
Let x = (x1 , . . . , xm ) and y = (y1 , . . . , ym ) be two arbitrary vertices in T . To
begin with, consider that vertices x and y belong to the same element S in T . In
this case, parity of corresponding co-ordinates of these two vertices are either same
or different up to k th position. First consider the case where parity of x and y up to
k th positions are the same. Then by Lemma 2.2 (b) we get walks of even length only
between xi and yi in Ci for every 1 ≤ i ≤ k. Further, if d(xi , yi ) in Ci is even for
all i > k then by Lemma 2.1(a), xi − yi walks of every even length greater than or
equal to d(xi , yi ) is possible in Ci . Thus, if p = max{d(xi , yi ) : 1 ≤ i ≤ m}, then p is
the smallest possible integer such that there is an xi − yi walk of length p for every
1 ≤ i ≤ m. So d(x, y) = p. Since p is finite, there is a x − y walk in S. If for some
j1 , j2 , . . . , jr ∈ {k +1, . . . , m}, d(xjt , yjt ) is odd, for 1 ≤ t ≤ r, then by Lemma 2.1(a)
xjt − yjt walk of every even length greater than or equal to njt − d(xjt − yjt ) exists
in Cnjt . Thus, in this case d(x, y) = max{l, d(xi , yi ) : 1 ≤ t ≤ r, 1 ≤ i = jt ≤ m},
where l = max{njt − d(xjt , yjt ) : 1 ≤ t ≤ r}. So d(x, y) is a finite number in this
case and we get a x − y walk in S.
Next, let parity of xi and yi differ for 1 ≤ i ≤ k. Then d(xi , yi ) is odd in
the respective cycle Ci and by Lemma 2.1(b)(ii) the only possible walks between
xi and yi are all of odd length for 1 ≤ i ≤ k. If for all i > k, d(xi , yi ) is odd,
then d(x, y) = max{d(xi , yi ) : 1 ≤ i ≤ m}, and is a finite number. Further, if for
some j1 , . . . , jr ∈ {k + 1, . . . , m}, d(xjt , yjt ) in Cjt is even, then as before d(x, y) =
max{l , d(xi , yi ) : 1 ≤ i ≤ m, i = jk }} where l = max{jk − d(xjk , yjk ) | j1 , . . . , jr ∈
{m + 1, . . . , n}}. So in this case also d(x, y) is a finite number and there is x − y
walk in S. Hence, every element in F induces a connected subgraph.
Now we consider that vertices x and y are in different elements of F. Then
xi and yi are of different parity for some i, 1 ≤ i ≤ k, but not for all. So, there
340 P. Singh, P. Panigrahi / Electronic Notes in Discrete Mathematics 63 (2017) 333–342

exist i and j, 1 ≤ i, j ≤ k, such that xi and yi have the same parity where as xj
and yj are of different parity. So d(xi , yi ) is even and d(xj , yj ) is odd. In this case
we will not get any xi − yi walk of odd length and xj − yj walk of even length by
Lemma 2.1(b). Hence, d(x, y) will not be defined i.e., there is no x − y path in T .
Thus, every element of F is a component of T . Next we shall show that each of
these components are self-centered subgraphs.
Let x = (x1 , . . . , xk , . . . , xm ) be a vertex in an arbitrary component S of T . We
know that e(x) = max{d(x, z) : z ∈ S}. Let e(x) = d(x, y) and y = (y1 , . . . , yk , . . . , ym ).
We shall determine y such that d(x, y) is maximum. Because of (1), y is such a ver-
tex in S such that for all i, 1 ≤ i ≤ m, xi − yi walk in Cni is of maximum length as
much as possible among all xi − zi walk in Cni for zi ∈ V (Ci ). We have two cases
according as n2l is odd or even, where nl = max{ni : 1 ≤ i ≤ k}.
Case 1. n2l is odd. We will have d(xl , yl ) = n2l because diam(Cnl ) = n2l . Since
nl
2 is odd, xl and yl have different parity. Then all xi and yi will have different
parity for 1 ≤ i ≤ k. Since x and y are in the same component then d(xi , yi ) is odd
and by Lemma 2.1(b) we can obtain xi − yi walk of length n2l for all i, 1 ≤ i ≤ k.
Now for i ≥ k + 1, whenever xi = yi we get xi − yi walk of maximum odd length,
i.e., ni . So, by Lemma 2.2 for every 1 ≤ i ≤ m there is a xi − yi walk of length
equal to max{ n2l , nk+1 , . . . , nm }, and is the minimum possible integer. Therefore,
d(x, y) = max{ n2l , nk+1 , . . . , nm }.
Case 2. n2l is even. We will have d(xl , yl ) = n2l . Since n2l is even and Cnl is an
even cycle, xl and yl have the same parity. Then xi and yi also have the same parity
and we will get xi − yi walk of length n2l for 1 ≤ i ≤ k. For i ≥ k + 1, whenever
d(xi , yi ) = 1, xi −yi walk of maximum even length, i.e., ni −1, is possible. Therefore,
e(x) = max{ n21 , n22 , . . . , n2k , nk+1 − 1, . . . , nm − 1} = max{ n2l , nk+1 − 1, . . . , nm−1 }.2

3 Tensor product of complete graphs, wheels and square

of cycles
In the following we study self-centeredness property of tensor product of complete
graphs with other d-self-centered graphs.
Theorem 3.1 Let T = Kn1 ⊗ . . . ⊗ Knm−1 ⊗ H be the tensor product of m − 1
complete graphs Kni , ni ≥ 3, 1 ≤ i ≤ m − 1, and a d-self-centered graph H with
|V (H)| ≥ 3. Then T is a k-self-centered graph, where
⎧
⎪
⎪ 2, if d = 1
⎪
⎪
⎪
⎪
⎨3, if d = 2 and H is a bipartite graph
k = 2, if d = 2, H is not a bipartite graph but E(z) = 3 for every vertex z in H
⎪
⎪
⎪
⎪ 3, if d = 2, H is not a bipartite graph but E(z) > 3 for every vertex z in H
⎪
⎪
⎩
d, if d ≥ 3.
 P. Singh, P. Panigrahi / Electronic Notes in Discrete Mathematics 63 (2017) 333–342 341

We know that wheel graphs are not self-centered graphs, but their tensor product
is a self-centered graph which is the content of the next theorem and can be proved
using equations 2, 3, and Theorem 1.1.
Theorem 3.2 Let T = Wn1 ⊗ . . . ⊗ Wnm be the tensor product of m wheel graphs
where ni ≥ 5. Then T is a 2-self-centered graph.
Theorem 3.3 Let T = Wn1 ⊗ . . . ⊗ Wnm−1 ⊗ H be the tensor product of m − 1
wheel graphs Wni , ni ≥ 5, and a d-self-centered graph H with d ≥ 3. Then T is a
d-self-centered graph.
In the next result we prove that k th power of a self-centered graph is also a
self-centered graph.
Theorem 3.4 Let G be a d-self-centered graph. Then Gk is a d/k -self-centered
graph, where 1 ≤ k ≤ d.
Proof. Since G is a d-self-centered graph, for every vertex u there exists a ver-
tex v such that d(u, v) = d, and let a path of length d between u and v be
P : u = v1 , v2 , . . . , vd+1 = v. Now in Gk , there exists a path P  which skips
every vertex at a distance k in P starting from the vertex u which is P  : u =
v1 , v1+k , . . . , v1+mk , vd+1 = v, where m =  kd − 1 when k divides d, and m =  kd
otherwise. It may be noted that the length of this path P  is d/k and hence the
result. 2
Corollary 3.1 For any cycle Cn , the graph Cn2 is a d-self-centered graph, where
n
d =  22 .
In the next result given below we calculate upper eccentricity of vertices in
square of cycles.
Lemma 3.1 Let x be an arbitrary vertex in Cn2 . If eccentricity of every vertex in
Cn2 is d then

3 if n = 4, 5,
E(x) =
d + 1 otherwise.

Proof. Since the diameter of the cycles C4 and C5 is two, C42 and C52 are complete
graphs with upper eccentricity of each vertex is three (by Theorem 1.1). Next, let
n ≥ 6 and V (Cn2 ) = {u0 , . . . , un−1 }. Without loss of generality, consider the vertex
u0 . The farthest vertex from u0 (in Cn2 ) is uk where k = n2
, and d(u0 , uk ) = d =
 k2 . If k is even then P1 : u0 , u2 , . . . , uk−2 , uk is an u0 − uk path of length d in Cn2 .
Then P2 : u0 , u2 , u4 , . . . , uk−2 , uk−1 , uk is an u0 − uk path of length d + 1 in Cn2 ,
and P2 is of opposite parity with P1 . If k is odd then P3 : u0 , u2 , . . . , uk−1 , uk is an
u0 − uk path of length d and P4 : u0 , u2 , . . . , uk−3 , uk−2 , uk−1 , uk is a path of length
d + 1 in Cn2 . Paths P3 and P4 are of opposite parity. We have D(u0 , u0 ) = 3 and
since d + 1 ≥ 3, E(u0 ) = d + 1. The result is true for all vertices in Cn2 because of
342 P. Singh, P. Panigrahi / Electronic Notes in Discrete Mathematics 63 (2017) 333–342

symmetric structure of it. 2

Theorem 3.5 Let T = Cn21 ⊗ Cn22 ⊗ . . . ⊗ Cn2m be the tensor product of m square
ni
 
of cycles, where each Cn2i is a di -self-centered graph, ni ≥ 6, and di =  2
2 ,
1 ≤ i ≤ m. Then

(i) T is d-self-centered graph when each di is distinct, where d = dl + 1, where dl

is the second largest element of {d1 , . . . , di }, for 1 ≤ i ≤ m.

(ii) T is dk -self-centered graph when there is at least one more index j = k such
that dj = dk , where dk = max{di : 1 ≤ i ≤ m}.

Proof. Follows from equations 2, 3, and Lemma 3.1. 2

References

[1] Bendall, S., and R. Hammack, Centres Of n-Fold Tensor Products of Graphs,
Discussiones Mathematicae Graph Theory, 24 (2004), 491-591.

[2] Buckley, F., Self-Centered Graphs, Annals of the New York Academy of
Sciences, 576 (1989), 71-78.

[3] Buckley, F., Z. Miller and P. J. Slater, On graphs conataining a given graph as
center, Journal of Graph Theory, 5 (1981), 427-434.

[4] Hammack, R., W. Imrich and S. Klavžar, “Handbook of Product Graphs”,

Second edition, CRC Press (2011).

[5] Huilgol, M. I., and C. Ramprakash, Cyclic Edge Extensions- Self-centered

graphs, Journal of Mathematics and Computer Science, 10 (2014), 131-137.

[6] Janakiraman, T. N., On Special Classes of Self-centered Graphs, Discrete

Mathematics, 126 (1994), 411-414.

[7] Janakiraman, T. N., M. Bhanumathi and S. Muthammai, Self-centered super

graph of a graph and center number of a graph, Ars Combinatoria, 87 (2008),
271-290.

[8] Stanic, Z., Some Notes on Minimal Self-Centered Graphs, AKCE International
Journal of Graphs and Combinatorics, 7 (2010), 97-102.

[9] Whitehead, A. N., and R. Bertrand, P rincipia M athematica, 2, University

Press (1912).