Graph Theory 32 (2012) 749–769 doi:10.7151/dmgt.1647
WIENER AND VERTEX PI INDICES OF THE STRONG PRODUCT OF GRAPHS
K. Pattabiraman and P. Paulraja Department of Mathematics, Annamalai University
Annamalainagar 608 002, India e-mail: pramank@gmail.com
pprajaau@sify.com
Abstract
The Wiener index of a connected graph G, denoted by W (G), is defined as
12P
u,v∈V (G)
d
G(u, v). Similarly, the hyper-Wiener index of a connected graph G, denoted by W W (G), is defined as
12W (G) +
14P
u,v∈V (G)
d
2G(u, v). The vertex Padmakar-Ivan (vertex PI) index of a graph G is the sum over all edges uv of G of the number of vertices which are not equidis- tant from u and v. In this paper, the exact formulae for Wiener, hyper- Wiener and vertex PI indices of the strong product G ⊠ K
m0,m1,...,mr−1, where K
m0,m1,...,mr−1is the complete multipartite graph with partite sets of sizes m
0, m
1, . . . , m
r−1, are obtained. Also lower bounds for Wiener and hyper-Wiener indices of strong product of graphs are established.
Keywords: strong product, Wiener index, hyper-Wiener index, vertex PI index.
2010 Mathematics Subject Classification: 05C12, 05C76.
1. Introduction
All the graphs considered in this paper are connected and simple. For vertices u, v ∈ V (G), the distance between u and v in G, denoted by d
G(u, v), is the length of a shortest (u, v)-path in G. The strong product of graphs G and H, denoted by G⊠H, is the graph with vertex set V (G)×V (H) = {(u, v) : u ∈ V (G), v ∈ V (H)}
and (u, x)(v, y) is an edge whenever (i) u = v and xy ∈ E(H), or (ii) uv ∈ E(G) and x = y, or (iii) uv ∈ E(G) and xy ∈ E(H), see Figure 1.
A topological index of a graph is a parameter related to the graph, it does
not depend on labeling or pictorial representation of the graph. In theoreti-
cal chemistry, molecular structure descriptors (also called topological indices) are
used for modeling physicochemical, pharmacological, toxicological, biological and other properties of chemical compounds [7]. Several types of such indices exist, especially those based on vertex and edge distances. One of the most inten- sively studied topological indices is the Wiener index. The Wiener index [26] is one of the oldest molecular-graph-based structure-descriptors [25]. Its chemical applications and mathematical properties are well studied in [6, 20].
b b
P
4b b b b
bbb b b b b
b b b b
b
P
5X
1X
2X
3Y
1Y
2Y
3Y
4Figure 1.
P
5⊠ P
4.bb
bb bb b b
b b b b
X
4X
5(x
1, y
4) x
1x
2x
3x
4x
5y
1y
2y
3y
4(x
2, y
4) (x
3, y
4) (x
4, y
4) (x
5, y
4) (x
5, y
2)(x
5, y
3)
Let G be a connected graph. Then W iener index of G, denoted by W (G), is defined as W (G) =
12P
u,v∈V (G)
d
G(u, v) with the summation going over all pairs of vertices of G. The hyper-W iener index of a connected graph G, de- noted by W W (G), is defined as W W (G) =
12W (G)+
14P
u,v∈V (G)
d
2G(u, v), where d
2G(u, v) = (d
G(u, v))
2.
The hyper-Wiener index of an acyclic graph was first introduced by Randi´c [24]. Then Klein et al. [16] studied hyper-Wiener index for all connected graphs.
Applications of the hyper-Wiener index as well as its calculation are well studied in [15, 17, 18, 19, 21].
Let e = uv be an edge of the graph G. The number of vertices of G whose distance to the vertex u is smaller than the distance to the vertex v is denoted by n
u(e). Analogously, n
v(e) is the number of vertices of G whose distance to the vertex v is smaller than the distance to the vertex u; here the vertices equidistant from both the ends of the edge e = uv are not counted. Another topological index, namely, the vertex Padmakar-Ivan (vertex PI) of G, denoted by P I(G), is defined as follows, P I(G) = P
e=uv∈E(G)
(n
u(e) + n
v(e)). For e = uv in G, the number of equidistant vertices of e is denoted by N
G(e). Then the above definition is equivalent to P I(G) = P
e∈E(G)
(|V (G)| − N
G(e)).
The vertex PI index is the topological index related to equidistant vertices;
Khadikar et al. [11] investigated the chemical applications of the vertex PI in- dex. The mathematical properties of the vertex PI index and its applications in chemistry and nanoscience are well studied in [1, 2, 3, 5, 12, 13, 23]. In [22]
we have studied the Wiener, hyper-Wiener and vertex PI indices of the tensor product of graphs. In this paper, we obtain the Wiener, hyper-Wiener and vertex PI indices of the graph G ⊠ K
m0,m1,...,mr−1, where K
m0,m1,...,mr−1is the complete multipartite graph with partite sets of sizes m
0, m
1, . . . , m
r−1. Also we have ob- tained lower bounds for Wiener and hyper-Wiener indices of the strong product of graphs.
If m
0= m
1= · · · = m
r−1= s in K
m0,m1,...,mr−1, then we denote K
m0,m1,...,mr−1by K
r(s). For S ⊆ V (G), hSi denotes the subgraph of G induced by S. A path and cycle on n vertices are denoted by P
nand C
n, respectively. We call C
3a triangle. For disjoint subsets S, T ⊂ V (G), by d
G(S, T ), we mean the sum of the distances in G from each vertex of S to every vertex of T, that is, d
G(S, T ) = P
s∈S,t∈T
d
G(s, t). For disjoint subsets S, T ⊂ V (G), E(S, T ) denotes the set of edges of G having one end in S and the other end in T. Notations and definitions which are not given here can be found in [4] or [9].
In this paper, besides some other results, we prove the following results.
Theorem 1. Let G be a connected graph with n vertices. Then
W (G ⊠ K
m0,m1,...,mr−1) = n
20W (G) + n(n
20− q − n
0) + (n
20− 2q − n
0)|E(G)|, where n
0= P
r−1i=0
m
iand q is the number of edges of K
m0,m1,...,mr−1.
Theorem 2. Let G be a connected graph with n vertices. Then W W (G ⊠ K
m0,m1,...,mr−1) = n
20W W (G) +
n2(3n
20− 4q − 3n
0) + 2(n
20− 2q − n
0)|E(G)|, where n
0= P
r−1i=0
m
iand q is the number of edges of K
m0,m1,...,mr−1.
Let (V
1, V
2, . . . , V
χ) be a proper χ(G)-colouring of G, where χ(G) is the chromatic number of G, such that no V
ican be augmented by adding a vertex of V
j, j ≥ i+1, that is, no vertex of V
jis nonadjacent to all the vertices of V
i, i < j, in G. Without loss of generality we assume that |V
1| ≥ |V
2| ≥ · · · ≥ |V
r|. We call such a χ(G)- colouring a decreasing χ(G)-colouring of G.
Based on the above results we have obtained the following lower bounds for the Wiener and hyper-Wiener indices of the graph G ⊠ G
′, where G and G
′are connected graphs.
Theorem 3. Let G be a connected graph with n vertices and m edges; let G
′be a graph with χ(G
′) = r ≥ 2. If the decreasing color classes of G
′have m
0, m
1, . . . , m
r−1vertices, then W (G⊠G
′) ≥ W (G⊠K
m0,m1,...,mr−1) = n
20W (G) +n(n
20−q −n
0)+(n
20−2q −n
0)m and W W (G⊠G
′) ≥ W W (G⊠K
m0,m1,...,mr−1) = n
20W W (G) +
n2(3n
20− 4q − 3n
0) + 2(n
20− 2q − n
0)m, where P
r−1i=0
m
i= n
0, n
0is
the number of vertices of G
′and q is the number of edges of K
m0,m1,...,mr−1.
2. Wiener Index of G ⊠ K
m0,m1,...,mr−1Let G be a simple connected graph with V (G) = {v
0, v
1, . . . , v
n−1} and let K
m0,m1,...,mr−1, r ≥ 2, be the complete multiparite graph with partite sets V
0, V
1, . . . , V
r−1and let |V
i| = m
i, 0 ≤ i ≤ r − 1. In the graph G ⊠ K
m0,m1,...,mr−1, let B
ij= v
i× V
j, v
i∈ V (G) and 0 ≤ j ≤ r − 1. For our convenience, we write
V (G) × V (K
m0,m1,...,mr−1) = S
n−1i=0
{v
i× S
r−1 j=0V
j}
= S
n−1i=0
{(v
i× V
0) ∪ (v
i× V
1) ∪ · · · ∪ (v
i× V
r−1)}
= S
n−1i=0
{B
i0∪ B
i1∪ · · · ∪ B
i(r−1)} = S
r−1 n−1 j=0i=0
B
ij, where Bij = v
i× V
j.
Let B = {B
ij}
i=0,1,...,n−1 j=0,1,...,r−1. Let X
i= S
r−1j=0
B
ijand Y
j= S
n−1i=0
B
ij; we call X
iand Y
jas layer and column of G ⊠ K
m0,m1,...,mr−1. Further, if v
iv
k∈ E(G), then the induced subgraph hB
ij∪ B
kpi of G ⊠ K
m0,m1,...,mr−1is isomorphic to K
|Vj||Vp|or, m
pindependent edges joining the corresponding vertices of B
ijand B
kjaccording to j 6= p or j = p. It is used in the proof of the next lemma.
The proof of the following lemma follows easily from the properties and structure of G ⊠ K
m0,m1,...,mr−1, see Figure 2 and Figure 3. The lemma is used in the proof of the main theorems of this paper.
Lemma 4. Let G be a connected graph and let B
ij, B
kp∈ B of the graph H = G ⊠ K
m0,m1,...,mr−1, where r ≥ 2.
(i) If v
iv
k∈ E(G), then d
H(B
ij, B
kp) =
( m
jm
p, if j 6= p, m
j(2m
j− 1), if j = p.
(ii) If v
iv
k∈ E(G), then d /
H(B
ij, B
kp) = m
jm
pd
G(v
i, v
k).
(iii) d
H(B
ij, B
ip) =
( m
jm
p, if j 6= p, 2m
j(m
j− 1), if j = p.
Proof of Theorem 1. Let H = G ⊠ K
m0,m1,...,mr−1. Clearly, W (H) = 12P
Bij,Bkp∈B
d
H(B
ij, B
kp)
=
12( P
n−1 i=0P
r−1 j,p=0j6=p
d
H(B
ij, B
ip) + P
n−1 i,k=0i6=k
P
r−1j=0
d
H(B
ij, B
kj) + P
n−1i,k=0 i6=k
P
r−1 j,p=0j6=p
d
H(B
ij, B
kp) + P
n−1 i=0P
r−1j=0
d
H(B
ij, B
ij)).
W (H) = 1
2 (A
1+ A
2+ A
3+ A
4), (1)
where A
1, A
2, A
3and A
4are the sums of the terms of the above expression, in
order.
v0
VerticesofG
Vertices ofKm0, m1, ..., mr−1
vi
v1
vk+1
vn−1
V0 Vj Vp Vr−1
Bkj
vℓ
vk
V1
bb b b
b
B00 B01 B0(r−1)
Bi0 Bi1 Bij Bi(r−1)
B(n−1)0 B(n−1)1
bb bbb
Figure 2 Ifvivk∈ E(G), then shortest paths of length 1 and 2 from BijtoBkjis shown in solid edges.
The broken edges give a shortest path of length 2 from a vertex ofBijto another vertex ofBij.
v0
VerticesofG
Vertices ofKm0, m1, ..., mr−1
vi+1
v1
vk
vn−1
V0 Vj Vp Vr−1
Bkj
Bij
vi
vk−1
V1
vi+2
bb b
b b
b b
B01
B00 B0(r−1)
Bi0 Bi1 Bi(r−1)
B(n−1)0 B(n−1)1
bBkp
Figure 3. If(vi, vk) is a shortest path of length ℓ in G, then a shortest path from any vertex of Bij
to any vertex ofBkj(resp. any vertex ofBijto any vertex ofBkp, p 6= j) of length ℓ is shown in solid edges (resp. broken edges).
We shall obtain A
1to A
4of (1), separately.
A
1= P
n−1 i=0P
r−1 j,p=0j6=p
d
H(B
ij, B
ip)
= P
n−1 i=0( P
r−1p=0 p6=0
d
H(B
i0, B
ip)+ P
r−1 p=0 p6=1d
H(B
i1, B
ip)+ · · · + P
r−1 p=0 p6=r−1d
H(B
i(r−1), B
ip))
= P
n−1 i=0( P
r−1p=0 p6=0
m
0m
p+ P
r−1 p=0 p6=1m
1m
p+ · · · + P
r−1 p=0 p6=r−1m
r−1m
p),
since |B
ip| = m
pand also see Lemma 4, therefore A1= n( X
r−1
a,p=0 a6=p
m
am
p).
(2)
A
2= P
r−1 j=0P
n−1 i,k=0i6=k
d
H(B
ij, B
kj)
= P
r−1 j=0( P
n−1i,k=0 i6=k vivk∈E(G)
d
H(B
ij, B
kj) + P
n−1 i,k=0i6=k vivk∈E(G)/
d
H(B
ij, B
kj))
= P
r−1 j=0( P
n−1i,k=0 i6=k vivk∈E(G)
m
j(2m
j− 1) + P
n−1 i,k=0i6=k vivk∈E(G)/
m
2jd
G(v
i, v
k)),
by Lemma 4, A2= P
r−1
j=0
( P
n−1 i,k=0i6=k vivk∈E(G)
m
j((1 + d
G(v
i, v
k))m
j− 1) + P
n−1 i,k=0i6=k vivk∈E(G)/
m
2jd
G(v
i, v
k)) , where 1 + dG(v
i, v
k) = 2 if v
iv
k∈ E(G).
A
2= ( P
r−1j=0
m
2j)( P
n−1 i,k=0i6=k vivk∈E(G)
d
G(v
i, v
k)) + P
r−1 j=0P
n−1 i,k=0i6=k vivk∈E(G)
m
j(m
j− 1)
+( P
r−1j=0
m
2j)( P
n−1 i,k=0i6=k vivk∈E(G)/
d
G(v
i, v
k)) = ( P
r−1j=0
m
2j)( P
n−1 i,k=0i6=k
d
G(v
i, v
k)) +( P
r−1j=0
m
j(m
j− 1))(2|E(G)|),
by combining the first and last sums of the above line. By the definition of Wiener index,
A
2= ( X
r−1j=0
m
2j)(2W (G)) + ( X
r−1j=0
m
j(m
j− 1))(2|E(G)|).
(3)
A
3= P
n−1 i,k=0i6=k
P
r−1 j,p=0,j6=p
d
H(B
ij, B
kp) = P
n−1 i,k=0i6=k
( P
r−1 p=0 p6=0d
H(B
i0, B
kp) + P
r−1p=0 p6=1
d
H(B
i1, B
kp) + · · · + P
r−1 p=0 p6=r−1d
H(B
i(r−1), B
kp))
= P
n−1 i,k=0i6=k
( P
r−1 p=0 p6=0m
0m
pd
G(v
i, v
k) + P
r−1 p=0 p6=1m
1m
pd
G(v
i, v
k) + · · · + P
r−1p=0 p6=r−1
m
r−1m
pd
G(v
i, v
k)), by Lemma 4,
A
3= P
n−1 i,k=0i6=k
( P
r−1 a,p=0a6=p
m
am
p)d
G(v
i, v
k).
By the definition of Wiener index, A3= ( X
r−1
a,p=0 a6=p
m
am
p)(2W (G)).
(4)
A
4= P
n−1 i=0( P
r−1j=0
d
H(B
ij, B
ij)) = P
n−1 i=0( P
r−1j=0
2m
j(m
j− 1)),
by Lemma 4, and hence
A
4= n( X
r−1j=0
2m
j(m
j− 1)).
(5)
Using (2), (3), (4) and (5) in (1), we have W (H) =12(n( P
r−1
a, p = 0 a 6= p
m
am
p)+( P
r−1j = 0
m
2j)(2W (G))+(
r−1
P
j = 0
m
j(m
j− 1))(2|E(G)|) +( P
r−1a, p = 0 a 6= p
m
am
p)(2W (G)) + n( P
r−1j = 0
2m
j(m
j− 1)))
= ( P
r−1j = 0
m
2j+ P
r−1 a, p = 0a 6= p
m
am
p)W (G) +
n2( P
r−1 a, p = 0a 6= p
m
am
p+ P
r−1j = 0
2m
j(m
j− 1)) +( P
r−1j = 0
m
j(m
j− 1)) |E(G)|,
by combining terms involving W (G) and the first and last terms,
W (H) = (n
20− 2q) + 2q)W (G) +
n2(2q + 2(n
20− 2q − n
0) + (n
20− 2q − n
0)|E(G)|, since Pr−1
j=0
m
2j= n
20−2q, P
r−1 a,p=0a6=p
m
am
p= 2q and P
r−1j=0
m
j(m
j−1) = n
20−2q−n
0, W (H) = n
20W (G) + n(n
20− q − n
0) + (n
20− 2q − n
0)|E(G)|.
If m
i= s, 0 ≤ i ≤ r − 1, in Theorem 1, we have the following
Corollary 5. Let G be a connected graph with n vertices and m edges. Then W (G ⊠ K
r(s)) = r
2s
2W (G) +
nrs2(rs + s − 2) + rs(s − 1)m.
As K
r= K
r(1), the above corollary gives the following
Corollary 6. Let G be a connected graph with n vertices and m edges. Then W (G ⊠ K
r) = r
2W (G) +
nr2(r − 1).
It can be easily verified that W (K
n) =
n(n−1)2, W (P
n) =
n(n26−1)and W (C
n) =
(
n38
, n is even,
n(n2−1)
8
, n is odd.
By [10], W (Q
n) = n2
2(n−1).
Now using Theorem 1 and Corollaries 5, 6 and the Wiener indices of K
n, P
n, C
nand Q
nwe obtain the exact Wiener indices of the following graphs.
1. W (K
n⊠ K
m0,m1,...,mr−1) =
n2(2nn
20− 2nq − nn
0− n
0), where n
0= P
r−1 i=0m
iand q is the number of edges of K
m0,m1,...,mr−1.
2. W (K
n⊠ K
r(s)) =
nrs2(nrs + ns − n − 1).
3. W (P
n⊠ K
m0,m1,...,mr−1) =
n6(n
2n
20+ 11n
20− 18q − 12n
0) − (n
20− 2q − n
0).
4. W (P
n⊠ K
r(s)) =
rs6(n
3rs + 2nrs + 9ns − 12n − 6s + 6).
5. W (P
n⊠ K
r) =
nr6(n
2r + 3r − 4).
6. W (C
n⊠ K
m0,m1,...,mr−1) = (
n8
(n
2n
20+ 16n
20− 24q − 16n
0), if n is even,
n
8
(n
2n
20+ 15n
20− 24q − 16n
0), if n is odd.
7. W (C
n⊠ K
r(s)) = (
nrs8
(n
2rs + 4rs + 12s − 16), if n is even,
nrs
8
(n
2rs + 3rs + 12s − 16), if n is odd.
8. W (C
n⊠ K
r) = (
nr8
(n
2r + 4r − 4), if n is even,
nr
8
(n
2r + 3r − 4), if n is odd.
9. W (Q
n⊠ K
m0,m1,...,mr−1) = 2
n−1(nn
202
n−1+ (n + 2)n
0(n
0− 1) − 2(n + 1)q).
10. W (Q
n⊠ K
r(s)) = 2
n−1((n2
n−1+ 1)r
2s
2+ (n + 1)rs
2− (n + 2)rs).
11. W (Q
n⊠ K
r) = 2
n−1r(n2
n−1+ r − 1).
3. Hyper-Wiener Index of G ⊠ K
m0,m1,...,mr−1In this section, we obtain the hyper-Wiener index of the graph G⊠K
m0,m1,...,mr−1. First we give a notation used in the proof of Theorem 2.
For two subsets S, T ⊂ V (G), we define d
2G(S, T ) = P
s∈S,t∈T
d
2G(s, t), where d
2G(s, t) = (d
G(s, t))
2.
The proof of the following lemma follows easily from the structure of G ⊠ K
m0,m1,...,mr−1. The lemma is used in the proof of the main theorem of this section.
Lemma 7. Let G be a connected graph and let B
ij, B
kp∈ B of the graph H = G ⊠ K
m0,m1,...,mr−1, where r ≥ 2.
(i) If v
iv
k∈ E(G), then d
2H(B
ij, B
kp) =
( m
jm
p, if j 6= p, m
j(4m
j− 3), if j = p.
(ii) If v
iv
k∈ E(G), then d /
2H(B
ij, B
kp) = m
jm
pd
2G(v
i, v
k).
(iii) d
2H(B
ij, B
ip) =
( m
jm
p, if j 6= p, 4m
j(m
j− 1), if j = p.
Proof of Theorem 2. Let H = G ⊠ K
m0,m1,...,mr−1. By the definition of hyper- Wiener index,
W W (H) =
12W (G) +
14P
Bij,Bkp∈B
d
2H(B
ij, B
kp)
=
12W (G) +
14( P
n−1 i=0P
r−1 j,p=0j6=p
d
2H(B
ij, B
ip) + P
n−1 i,k=0i6=k
P
r−1j=0
d
2H(B
ij, B
kj) + P
n−1i,k=0 i6=k
P
r−1 j,p=0j6=p
d
2H(B
ij, B
kp) + P
n−1 i=0P
r−1j=0
d
2H(B
ij, B
ij)).
Denote by A
5, A
6, A
7and A
8the sums of the terms of the above expression, in order, and hence
W W (H) = 1
2 W (G) + 1
4 (A
5+ A
6+ A
7+ A
8).
(6)
We shall obtain A
5to A
8, in (6), separately.
A
5= P
n−1 i=0P
r−1 j,p=0j6=p
d
2H(B
ij, B
ip) = P
n−1 i=0( P
r−1p=0 p6=0
d
2H(B
i0, B
ip) + P
r−1p=0 p6=1
d
2H(B
i1, B
ip) + · · · + P
r−1 p=0 p6=r−1d
2H(B
i(r−1), B
ip))
= P
n−1 i=0( P
r−1p=0 p6=0
m
0m
p+ P
r−1 p=0 p6=1m
1m
p+ . . . + P
r−1 p=0 p6=r−1m
r−1m
p), by Lemma 7, therefore
A
5= n( X
r−1a,p=0 a6=p
m
am
p).
(7)
A
6= P
r−1 j=0P
n−1 i,k=0i6=k
d
2H(B
ij, B
kj) = P
r−1 j=0( P
n−1i,k=0 i6=k vivk∈E(G)
d
2H(B
ij, B
kj) + P
n−1i,k=0 i6=k,vivk∈E(G)/
d
2H(B
ij, B
kj)) = P
r−1 j=0( P
n−1i,k=0 i6=k vivk∈E(G)
m
j(4m
j− 3)
+ P
n−1 i,k=0i6=k vivk∈E(G)/
m
2jd
2G(v
i, v
k)), by Lemma 7,
A
6= P
r−1 j=0( P
n−1i,k=0 i6=k vivk∈E(G)
m
j((3 + d
2G(v
i, v
k))m
j− 3) + P
n−1 i,k=0i6=k vivk∈E(G)/
m
2jd
2G(v
i, v
k)),
where 3 + d
2G(v
i, v
k) = 4 if v
iv
k∈ E(G).
A
6= P
r−1 j=0( P
n−1i,k=0 i6=k vivk∈E(G)
(m
2jd
2G(v
i, v
k) + 3m
j(m
j− 1)) + P
n−1 i,k=0i6=k vivk∈E(G)/
m
2jd
2G(v
i, v
k))
= P
r−1 j=0( P
n−1i,k=0 i6=k
m
2jd
2G(v
i, v
k) + P
n−1 i,k=0i6=k vivk∈E(G)
3m
j(m
j− 1)),
by combining the first and last sum of the previous line, A6= ( X
r−1
j=0
m
2j)( X
n−1i,k=0 i6=k
d
2G(v
i, v
k)) + ( X
r−1j=0
3m
j(m
j− 1))(2|E(G)|).
(8)
A
7= P
n−1 i,k=0i6=k
P
r−1 j,p=0,j6=p
d
2H(B
ij, B
kp) = P
n−1 i,k=0i6=k
( P
r−1 p=0 p6=0d
2H(B
i0, B
kp) + P
r−1p=0 p6=1
d
2H(B
i1, B
kp) + · · · + P
r−1 p=0 p6=r−1d
2H(B
i(r−1), B
kp))
= P
n−1 i,k=0i6=k
( P
r−1 p=0 p6=0m
0m
pd
2G(v
i, v
k) + P
r−1 p=0 p6=1m
1m
pd
2G(v
i, v
k) + · · · + P
r−1p=0 p6=r−1
m
r−1m
pd
2G(v
i, v
k)), by Lemma 7, and hence
A
7= P
n−1 i,k=0i6=k
( P
r−1 a,p=0a6=p
m
am
p)d
2G(v
i, v
k).
A
7= ( X
r−1a,p=0 a6=p
m
am
p)( X
n−1i,k=0 i6=k
d
2G(v
i, v
k)).
(9)
A
8= X
n−1i=0
( X
r−1j=0
d
2H(B
ij, B
ij)) = X
n−1i=0
( X
r−1j=0
4m
j(m
j− 1)) (10)
= n( X
r−1j=0
4m
j(m
j− 1)).
Using Theorem 1, and the equations (7), (8), (9) and (10) in (6), we have W W (H) = 12(( P
r−1
j=0
m
2j+ P
r−1 a,p=0a6=p
m
am
p)W (G) +
n2( P
r−1 a,p=0a6=p
m
am
p+ P
r−1j=0
2m
j(m
j− 1)) + ( P
r−1j=0
m
j(m
j− 1))|E(G)|) +
14(( P
r−1j=0
m
2j)( P
n−1 i,k=0i6=k
d
2G(v
i, v
k)) + n( P
r−1 a,p=0a6=p
m
am
p) +( P
r−1j=0
3m
j(m
j− 1))(2|E(G)|) + ( P
r−1 a,p=0a6=p
m
am
p)( P
n−1 i,k=0i6=k
d
2G(v
i, v
k)) +n( P
r−1j=0
4m
j(m
j− 1))) = ( P
r−1j=0
m
2j+ P
r−1 a,p=0a6=p
m
am
p)(
12W (G) +
14P
n−1i,k=0 i6=k
d
2G(v
i, v
k)) +
n4( P
r−1 a,p=0a6=p
2m
am
p+ P
r−1j=0
6m
j(m
j− 1)) +2( P
r−1j=0
m
j(m
j− 1))|E(G)| = ( P
r−1j=0
m
2j+ P
r−1 a,p=0a6=p
m
am
p)W W (G) +
n2( P
r−1a,p=0 a6=p
m
am
p+ P
r−1j=0
3m
j(m
j− 1)) + 2( P
r−1j=0
m
j(m
j− 1))|E(G)|
= ((n
20− 2q) + 2q)W W (G) +
n2(2q + 3(n
20− 2q − n
0)) + 2(n
20− 2q − n
0)|E(G)|
= n
20W W (G) +
n2(3n
20− 4q − 3n
0) + 2(n
20− 2q − n
0)|E(G)|.
If m
i= s, 0 ≤ i ≤ r − 1, in Theorem 2, we have the following
Corollary 8. If G is a connected graph with n vertices and m edges, then W W (G ⊠ K
r(s)) = r
2s
2W W (G) +
nrs2(rs + 2s − 3) + 2rs(s − 1)m.
As K
r= K
r(1), the above corollary gives the following
Corollary 9. If G is a connected graph with n vertices and m edges, then W W (G ⊠ K
r) = r
2W W (G) +
nr2(r − 1).
It can be easily verified, see also [8], that W W (K
n) =
n(n−1)2, W W (P
n) =
n4+2n3−n2−2n
24
and
W W (C
n) =
(
n2(n+1)(n+2)48
, n is even,
n(n2−1)(n+3)
48
, n is odd.
By [10], W (Q
n) = n2
2(n−1)and hence W W (Q
n) = n(n + 3)2
2n−4.
Now using Theorem 2 and Corollaries 8, 9 and the hyper-Wiener indices of K
n, P
n, C
nand Q
n, we obtain the exact hyper-Wiener indices of the following graphs.
1. W W (K
n⊠ K
m0,m1,...,mr−1) =
n2(3nn
20− 4nq − 2nn
0− n
0), where n
0= P
r−1i=0
m
iand q is the number of edges of K
m0,m1,...,mr−1.
2. W W (K
n⊠ K
r(s)) =
nrs2(nrs + 2ns − 2n − 1).
3. W W (P
n⊠ K
m0,m1,...,mr−1) =
24n(n
3n
20+ 2n
2n
20− nn
20+ 82n
20− 84n
0− 144q)
−2(n
20− 2q − n
0).
4. W W (P
n⊠ K
r(s)) =
nrs24(n
3rs + 2n
2rs − nrs + 10rs + 72s − 84) − 2rs(s − 1).
5. W W (P
n⊠ K
r) =
nr24(n
3r + 2n
2r − nr + 10r − 12).
6. W W (C
n⊠ K
m0,m1,...,mr−1)
= (
n48
(n(n + 1)(n + 2)n
20+ 168n
20− 288q − 168n
0), if n is even,
n
48
((n
2− 1)(n + 3)n
20+ 168n
20− 288q − 168n
0), if n is odd.
7. W W (C
n⊠K
r(s)) = (
nrs48
(nrs(n + 1)(n + 2) + 24rs +144s −168), if n is even,
nrs
48
((n
2− 1)(n + 3)rs + 24rs + 144s − 168), if n is odd.
8. W W (C
n⊠ K
r) = (
nr48
(n(n + 1)(n + 2)r + 24r − 24), if n is even,
nr
48
((n
2− 1)(n + 3)r + 24r − 24), if n is odd.
9. W W (Q
n⊠ K
m0,m1,...,mr−1)
= n
20n(n + 3)2
2n−4+ 2
n−1(3n
20− 4q − 3n
0+ 2nn
20− 4nq − 2nn
0).
10. W W (Q
n⊠ K
r(s)) = n(n + 3)2
2n−4r
2s
2+ 2
n−1(rs(rs + 2s − 3) + 2nrs(s − 1)).
11. W W (Q
n⊠ K
r) = n(n + 3)2
2n−4r
2+ 2
n−1r(r − 1).
4. Lower Bounds for Wiener and Hyper-Wiener Indices of the Strong Product of Graphs
In this section, we establish lower bounds for Wiener and hyper-Wiener indices of G ⊠ G
′.
As the proof of the following lemma is trivial, we just quote the statement.
Lemma 10. W (K
m0,m1,...,mr−1) = n
20− n
0− q and W W (K
m0,m1,...,mr−1) = 2n
20− 2n
0− 3q, where P
r−1i=0
m
i= n
0and q is the number of edges of K
m0,m1,...,mr−1. The following lemma follows as G ⊆ K
m0,m1,...,mr−1.
Lemma 11. Let G be a connected graph on n vertices with chromatic number χ(G) = r ≥ 2. If C is the decreasing χ(G)-coloring of G with sizes of the colour classes m
0, m
1, . . . , m
r−1, then W (G) ≥ W (K
m0,m1,...,mr−1) = n
20− n
0− q and W W (G) ≥ W W (K
m0,m1,...,mr−1) = 2n
20− 2n
0− 3q, where P
r−1i=0
m
i= n
0, where
n
0is the number of vertices of G and q is the number of edges of K
m0,m1,...,mr−1.
Proof of Theorem 3. By Lemma 11, W (G
′) ≥ W (K
m0,m1,...,mr−1). As G
′is a subgraph of K
m0,m1,...,mr−1, we have W (G ⊠ G
′) ≥ W (G ⊠ K
m0,m1,...,mr−1), since d
G⊠G′((x
1, y
1), (x
2, y
2)) ≥ d
G⊠Km0,m1,...,mr−1((x
1, y
1), (x
2, y
2)) for any pair of ver- tices (x
1, y
1) and (x
2, y
2) of G⊠G
′. Thus, W (G⊠G
′) ≥ W (G⊠K
m0,m1,...,mr−1) = n
20W (G) + n(n
20− q − n
0) + (n
20− 2q − n
0)m, by Theorem 1.
Similarly, d
2G⊠G′((x
1, y
1), (x
2, y
2)) ≥ d
2G⊠Km0,m1,...,mr−1
((x
1, y
1), (x
2, y
2)) for any pair of vertices (x
1, y
1) and (x
2, y
2) of G ⊠ G
′. Consequently, W W (G ⊠ G
′) ≥ W W (G ⊠ K
m0,m1,...,mr−1) = n
20W W (G) +
n2(3n
20− 4q − 3n
0) + 2(n
20− 2q − n
0)m, by Theorem 2.
5. Vertex PI Index of G ⊠ K
m0,m1,...,mr−1In this section, we compute the vertex PI index of H = G ⊠ K
m0,m1,...,mr−1. First we introduce some notations for our convenience. Let G be a con- nected graph and E
′⊆ E(G); let N
G(E
′) = P
e∈E′
N
G(e), where N
G(e) is the number of vertices equidistant from the edge e in G. For e = v
iv
k∈ E(G) let E
ikjp= E(B
ij, B
kp), where B
ij′s
are as defined above. We denote N
H(E
ikjp) = P
e′∈Eikjp
N
H(e
′).
For e = v
iv
k∈ E(G), we define four sets, namely, S
1(e), S
2(e), S
3(v
i) and S
3(v
k) as follows: let S
1(e) = {x ∈ V (G) | d(x, v
i) = 1 = d(x, v
k)}, that is, the set of vertices which are lying on a triangle containing the edge e; let
|S
1(e)| = s
1(e). Let S
2(e) = {x ∈ V (G) | d(x, v
i) = d(x, v
k) = k > 1}, that is, the set of vertices which are at distance k > 1 from both the ends v
iand v
kof e; let |S
2(e)| = s
2(e). Clearly, N
G(e) = s
1(e) + s
2(e). Let S
3(v
i) = {x ∈ N (v
i) | x is not an isolated vertex in hN (v
i)i
G}; let |S
3(v
i)| = s
3(v
i). Similarly, S
3(v
k) = {x ∈ N (v
k) | x is not an isolated vertex in hN (v
k)i
G}; let |S
3(v
k)| = s
3(v
k). Let S(v
i) = S
3(v
i) − S
1(e) − {v
k} and S(v
k) = S
3(v
k) − S
1(e) − {v
i}, where e = v
iv
k.
Let T (e) ⊂ V (G) be set of vertices which are equidistant from the edge e ∈ E(G). For e = v
iv
k∈ E(G) and a ∈ T (e) we define N
Ha(e
′) as the number of equidistant vertices of e
′∈ E(X
i, X
k) lying in X
a(⊆ V (H)). Consequently, N
Ha(E(X
i, X
k)) = P
e′∈E(Xi,Xk)
N
Ha(e
′).
Now we define N
HT(e)(E(X
i, X
k)) = P
a∈T (e)
N
Ha(E(X
i, X
k)). For e ∈ E
′, N
GE′(e) denotes the number of equidistant vertices of the edge e in G.
In G⊠K
m0,m1,...,mr−1, define, E
1= {(u, v)(x, y) ∈ E(G⊠K
m0,m1,...,mr−1)|ux ∈ E(G) and v = y}, E
2= {(u, v)(x, y) ∈ E(G ⊠ K
m0,m1,...,mr−1)|ux ∈ E(G)and vy ∈ E(K
m0,m1,...,mr−1)} and E
3= {(u, v)(x, y) ∈ E(G⊠K
m0,m1,...,mr−1)|u = x and vy ∈ E(K
m0,m1,...,mr−1)}.
Clearly, E
1∪ E
2∪ E
3= E(G ⊠ K
m0,m1,...,mr−1).
Also, |E
1| = |E(G)||V (K
m0,m1,...,mr−1)|, |E
2| = 2|E(G)||E(K
m0,m1,...,mr−1)| and
|E
3| = |V (G)||E(K
m0,m1,...,mr−1)|.
Theorem 12. Let G be a connected graph on n vertices, then
P I(G⊠K
m0,m1,...,mr−1) = n
0(n
0+2q)P I(G)−(n
30+n
20−n
0−2n
0q−4q− P
r−1 h=0m
3h) ( P
e∈E(G)
(s
3(v
i)+s
3(v
k)−2s
1(e)−2))−2(2q−n
30+2n
0q+ P
r−1h=0
m
3h)( P
e∈E(G)
s
1(e))
− (2n
20− 2n
0+ 16n
0q − 6n
30+ 8q + 6 P
r−1h=0
m
3h)|E(G)| + n(n
30− 2n
0q − P
r−1 h=0m
3h).
Proof. An edge e = v
iv
k∈ E(G) contributes edges for E
1, E
2and E
3⊆ E(hX
i∪ X
ki) in H. We compute N
HE1(e
′), for an e
′∈ E
1∩ E(hX
i∪ X
ki
H), N
HE2(e
′), for an e
′∈ E
2∩ E(hX
i∪ X
ki
H) and N
HE1(e
′), for an e
′∈ E
3∩ E(hX
i∪ X
ki
H), separately, in the items (1), (2) and (3) below.
Case 1. If e
′∈ E
1∩ E(hX
i∪ X
ki
H), then e
′∈ E(B
ij, B
kj) for some j.
Case 1(i). If v
ℓ∈ S
1(e), then every vertex in X
ℓis an equidistant ver- tex (of distance one or two) from e
′∈ E(hX
i∪ X
ki
H). Hence, N
HS1(e)(e
′) = ( P
r−1h=0
m
h)|S
1(e)| = ( P
r−1h=0
m
h)s
1(e).
As there are m
jedges of E
1between B
ijand B
kj, NHS1(e)(E
ikjj) = X
e′∈Ejjik
N
HS1(e)(e
′) = m
j( X
r−1h=0
m
h)s
1(e).
(11)
Case 1(ii). If v
ℓ∈ S
2(e), then every vertex of X
ℓis an equidistant vertex, of distance d
G(v
i, v
ℓ) = d
G(v
k, v
ℓ), from e
′∈ E(hX
i∪ X
ki
H). Hence,
N
HS2(e)(e
′) = ( X
r−1h=0
m
h)|S
2(e)| = ( X
r−1h=0
m
h)s
2(e).
As there are m
jedges of E
1in E(B
ij, B
kj), we have NHS2(e)(E
ikjj) = X
e′∈Ejjik
N
HS2(e)(e
′) = m
j( X
r−1h=0
m
h)s
2(e).
(12)
Case 1(iii). Let v
ℓ∈ S(v
i) ∪ S(v
k). Without loss of generality, we assume that v
ℓ∈ S(v
i), the case v
ℓ∈ S(v
k) is similar. Let e
′= x
tijx
tkj∈ E(B
ij, B
kj) for some j, where x
tijand x
tkjare the t
thvertices in B
ijand B
kj, respectively; the ends of e
′are equidistant (of distance two) from all the vertices of B
ℓj− {x
tℓj} but not all other vertex of X
ℓ. Consequently, we have N
H{S(vi)∪S(vk)}(e)(e
′) = (m
j− 1)|S(v
i) ∪ S(v
k)| = (m
j− 1)(|S
3(v
i)| + |S
3(v
k)| − 2|S
1(e)| − 2) = (m
j− 1)(s
3(v
i) + s
3(v
k) − 2s
1(e) − 2).
As there are m
jedges in E(B
ij, B
kj), we have NH{S(vi)∪S(vk)}(e)(E
ikjj) = X
e′∈Ejjik