MATCH MATCH Commun. Math. Comput. Chem.

58 (2007) 183-194
Communications in Mathematical
and in Computer Chemistry ISSN 0340 - 6253

A Unified Approach to Extremal Cacti for

Different Indices

Huiqing LIU1 ∗ Mei LU2 †

1 School of Mathematics and Computer Science, Hubei University, Wuhan 430062, China
2 Department of Mathematical Sciences, Tsinghua University, Beijing 100084, China.

(Received January 31, 2007)

Abstract
Many chemical indices have been invented in theoretical chemistry, such
as Wiener index, Merrifield-Simmons index, Hosoya index, spectral radius and
Randić index, etc. The extremal trees and unicyclic graphs for these chemical
indices are interested in existing literature. Let G be a molecular graph (called
a cacti), which all of blocks of G are either edges or cycles. Denote G (n, r)
the set of cacti of order n and with r cycles. Obviously, G (n, 0) is the set
of all trees and G (n, 1) is the set of all unicyclic graphs. In this paper, we
present a unified approach to the extremal cactus, which have the same or
very similar structures, for Wiener index, Merrifield-Simmons index, Hosoya
index and spectral radius. From our results, we can derive some known results.

1. Introduction
Mathematical descriptors of molecular structure, such as various topological in-
dices, have been widely used in structure-property-activity studies (see [10, 11, 16]).
Among the numerous topological indices considered in chemical graph theory, only
a few have been found noteworthy in practical application (see [15]). The Wiener
index is the first chemical index introduced in 1947 by Harold Wiener. It was shown
∗
email: hql [email protected]; Partially supported by NNSFC (No. 10571105).
†
email: [email protected]; Partially supported by NNSFC (No. 10571105).
 - 184 -

that there are excellent correlations between the Wiener index of the molecular graph
of an organic compound and a variety of physical and chemical properties of the or-
ganic compound (see [20], [21]). M. Randić [18] showed that if alkanes are ordered so
that their Randić-index decrease then the extent of their branching should increase.
The Hosoya index of a graph was introduced by Hosoya in 1971 [9] and was applied
to correlations with boiling points, entropies, calculated bond orders, as well as for
coding of chemical structures (see [14, 17]). Merrifield and Simmons [14] developed
a topological approach to structural chemistry. The cardinality of the topological
space in their theory turns out to be equal to Merrifield-Simmons index of the re-
spective molecular graph G. There have been many publications on these chemical
indices (see [4]-[7], [12], [13], [20]-[24]). In [12], Li and Zheng put forward a problem,
which asked for a more unified approach that can cover extremal result for as many
as chemical indices as possible. Here, we present a unified and simple approach to
extremal cactus for the Wiener index, Merrifield-Simmons index, Hosoya index and
spectral radius.
In order to discuss our results, we first introduced some terminologies and nota-
tions of graphs. Other undefined notations may refer to [1, 2]. Let G = (V, E) be
a simple undirected graph of order n. For a vertex u of G, we denote the neigh-
borhood and the degree of u by NG (u) and dG (u), respectively. For two vertices
u and v (u = v) of G, the distance between u and v, denoted by dG (u, v), is the
number of edges in a shortest path joining u and v in G. For H ⊆ V (G), we let
NH (u) = NG (u) ∩ H. Denote NH [u] = NH (u) ∪ {u}. We will use G − x or G − xy
to denote the graph that arises from G by deleting the vertex x ∈ V (G) or the edge
xy ∈ E(G). Similarly, G + xy is a graph that arises from G by adding an edge
xy ∈/ E(G), where x, y ∈ V (G).
We list the definitions of some topological indices as follows.
(i) The Wiener index of G, is defined as

W (G) = dG (u, v),
u,v

where dG (u, v) is the distance between u and v in G and the sum goes over all the
pairs of vertices.
(ii) The Merrifield-Simmons index, is defined as

σ(G) = i(G; k),
k≥0

where i(G; k) is the number of k-independent vertex sets of G. Note that i(G; 0) = 1.
 - 185 -

(iii) The Hosoya index, is defined as


z(G) = m(G; k),
k≥0

where m(G; k) is the number of k-independent edge sets of G. Note that m(G; 0) = 1.
(iv) The Randić index of G is defined (see [18]) as
 1
R(G) = (d(u)d(v))− 2 ,
u,v

where d(u) denotes the degree of the vertex u of the molecular graph G, the sum-
mation goes over all pairs of adjacent vertices of G.
(v) The spectral radius, ρ(G), of G is the largest eigenvalue of A(G), where A(G)
be the adjacency matrix of a graph G. When G is connected, A(G) is irreducible and
by the Perron-Frobenius Theorem, the spectral radius is simple and has a unique
positive eigenvector. We will refer to such an eigenvector as the Perron vector of G.
Let G be a connected graph. We call G a cactus if all of blocks of G are ei-
ther edges or cycles. Denote G (n, r) the set of cacti of order n and with r cycles.
Obviously, G (n, 0) is the set of all trees and G (n, 1) is the set of all unicyclic graphs.
We use G0 (n, r) to denote the cactus obtained from the n-vertex star by adding
r mutually independent edges (see Fig. 1).
2r 2r 2r
        

   G0 (n, r)
n−1−2r
(n = 2r + 1) G0 (n, r)
0 (n = 2r + 2)
G (n, r)
Fig. 1

2. Lemmas
Denote the characteristic polynomial of a graph G by φ(G; λ).

Lemma 2.1 (see [19]). Let v be a vertex of a graph G, and let C (v) be the set
of all cycles containing v. Then
 
φ(G; λ) = λφ(G − v; λ) − φ(G − v − w; λ) − 2 φ(G − V (Z); λ).
vw∈E(G) Z∈C (v)
 - 186 -

Lemma 2.2 (see [22]). Let G be a connected graph, and let u, v ∈ V (G).
Suppose v1 , v2 , . . . , vs ∈ N (v) \ N (u) (1 ≤ s ≤ dG (v)) and x = (x1 , x2 , . . . , xn )t
is the Perron vector of A(G), where xi corresponds to the vertex vi (1 ≤ i ≤ n).
Let G∗ be the graph obtained from G by deleting the edges vvi and adding the edges
uvi , 1 ≤ i ≤ s. If xu ≥ xv , then

ρ(G) < ρ(G∗ ).

Lemma 2.3 (see [8]). Let G be a graph and v ∈ V (G). Then

(i) σ(G) = σ(G − v) + σ(G − NG [v]);

(ii) z(G) = z(G − v) + z(G − {u, v});
u∈NG (v)
(iii) Moreover, if G1 , G2 , · · · , Gω are the components of a graph G, then σ(G) =
)ω )ω
j=1 σ(Gj ) and z(G) = j=1 z(Gj ).

From Lemma 2.3, if v is a vertex of G, then σ(G) > σ(G − v). Moreover, if G is
a graph with at least one edge, then z(G) > z(G − v).

X Y

v H u

X Y

v H u G v H u
Y X

G∗1 G∗2
Fig. 2

Lemma 2.4. Let H, X, Y be three connected graphs disjoint in pair. Suppose

that u, v are two vertices of H, v  is a vertex of X, u is a vertex of Y . Let G be the
graph obtained from H, X, Y by identifying v with v  and u with u , respectively. Let
G∗1 be the graph obtained from H, X, Y by identifying vertices v, v  , u , and let G∗2 be
the graph obtained from H, X, Y by identifying vertices u, v  , u (see Fig. 2). Then
(i) σ(G∗1 ) > σ(G) or σ(G∗2 ) > σ(G);
∗
(ii) z(G1 ) < z(G) or z(G∗2 ) < z(G);
∗
(iii) W (G1 ) < W (G) or W (G∗2 ) < W (G);
∗
(iv) ρ(G1 ) > ρ(G) or ρ(G∗2 ) > ρ(G).
 - 187 -

Proof. (i) Denote a = σ(X − v), a = σ(X − NX [v]), b = σ(Y − u) and

b = σ(Y − NY [u]). Then a > a > 0 and b > b > 0. Let iu,v be the number
of independent vertex subsets in H containing both u and v. Then iu,v = 0 if
uv ∈ E(G), iu,v = σ(H − NH [u] − NH [v]) if uv ∈ E(G). By Lemma 2.3, we have

σ(G) = σ(G − v) + σ(G − NG [v])

= abσ(H − v − u) + ab σ(H − v − NH [u]) + a bσ(H − u − NH [v]) + a b iu,v .

Similarly, we have

σ(G∗1 ) = ab[σ(H − v − u) + σ(H − v − NH [u])] + a b [σ(H − u − NH [v]) + iu,v ],

σ(G∗2 ) = ab[σ(H − v − u) + σ(H − u − NH [v])] + a b [σ(H − v − NH [u]) + iu,v ].

Therefore

σ(G) − σ(G∗1 ) = a (b − b )σ(H − u − NH [v]) − a(b − b )σ(H − v − NH [u]),

σ(G) − σ(G∗2 ) = b (a − a )σ(H − v − NH [u]) − b(a − a )σ(H − u − NH [v]).

If σ(G) − σ(G∗1 ) ≥ 0, then (b − b )[a σ(H − u − NH [v]) − aσ(H − v − NH [u])] ≥ 0.

Since a > a and b > b , we have σ(H − u − NH [v]) > σ(H − v − NH [u]). So

σ(G) − σ(G∗2 ) = (a − a )[b σ(H − v − NH [u]) − bσ(H − u − NH [v])]

< (a − a )[b σ(H − v − NH [u]) − bσ(H − v − NH [u])]
= (a − a )(b − b)σ(H − v − NH [u]) < 0.

(ii) Let δ = 0 if uv ∈ / E(G) and δ = 1 if uv ∈ E(G). Let e0 = 1 if uv ∈ E(G);

 z(X−v−x)
and e0 = 2 if uv ∈ E(G). Denote p = z(X − v), q = z(Y − u), p = z(X−v)
,
x∈NX (v)
 z(Y −u−y)  
q = z(Y −u)
, ru = z(H −v−u−u ), rv = z(H −v−u−v  ),
y∈NY (u) u ∈NH−v (u) v  ∈NH (v)−u
 
r0 = z(H − v − u − v  − u ).
v  ∈NH (v)−u u ∈NH−v−v (u)
By Lemma 2.3, we have

z(G) = z(G − v) + z(G − v − v  ) + δz(G − v − u)
v  ∈N G (v)−u

= e0 z(G − v − u) + z(G − v − u − u )
u ∈NG−v (u)
  
+ z(G − v − v  − u) + z(G − v − v  − u − u )
v  ∈NG (v)−u v  ∈NG (v)−u u ∈NG−v−v (u)
 - 188 -

 
= e0 z(G − v − u) + z(G − v − u − u ) + z(G − v − u − u )
u ∈N H−v (u) u ∈N Y (u)
  
+ z(G − v − u − v  − u ) + z(G − v − u − v  )
u ∈NY (u) v  ∈NX (v) v  ∈NH (v)−u
  

+ z(G − v − u − v ) + z(G − v − u − v  − u )
v  ∈NX (v) v  ∈NX (v) u ∈NH−v (u)
 
+ z(G − v − u − v  − u )
v  ∈N H (v)−u u ∈N Y (u)
 
+ z(G − v − u − v  − u )
v  ∈NH (v)−u u ∈NH−v−v (u)

= pq · [e0 z(H − v − u) + q  z(H − v − u) + ru + p q  z(H − v − u) + rv

+p z(H − v − u) + ru p + rv q  + r0 ]
= pq[z(H − v − u)(e0 + q  + p + p q  ) + rv (1 + q  ) + ru (1 + p ) + r0 ].

Similarly, we get

z(G∗1 ) = pq[z(H − v − u)(e0 + q  + p ) + ru (1 + p + q  ) + rv + r0 ],

z(G∗2 ) = pq[z(H − v − u)(e0 + q  + p ) + rv (1 + p + q  ) + ru + r0 ].

Thus

z(G) − z(G∗1 ) = pqq  [z(H − v − u)p + rv − ru ],

z(G) − z(G∗2 ) = pqp [z(H − v − u)q  + ru − rv ].

If z(G) − z(G∗1 ) ≤ 0, then pqq  [z(H − v − u)p + rv − ru ] ≤ 0, that is, ru − rv ≥

z(H − v − u)p . So

z(H − v − u)q  + ru − rv ≥ z(H − v − u)q  + z(H − v − u)p

= z(H − v − u)(q  + p ) > 0.

Note that pqp > 0, and hence z(G∗2 ) < z(G).

(iii) We have
  
W (G) = dG (x, y) + dG (x, y) + dG (x, y)
x,y∈V (X) x,y∈V (Y ) x∈V (X−v),y∈V (Y −u)
  
+ dG (x, y) + dG (x, y) + dG (x, y)
x,y∈V (H) x∈V (H),y∈V (Y −u) x∈V (H),y∈V (X−v)
  
= dX (x, y) + dY (x, y) + dG (x, y)
x,y∈V (X) x,y∈V (Y −u) x∈V (X−v),y∈V (Y −u)
 - 189 -

  
+ dH (x, y) + dG (x, y) + dG (x, y).
x,y∈V (H) x∈V (H),y∈V (Y −u) x∈V (H),y∈V (X−v)

Thus

W (G) − W (G∗1 )
 
= [dG (x, y) − dG∗1 (x, y)] + [dG (x, y) − dG∗1 (x, y)],
x∈V (X−v),y∈V (Y −u) x∈V (H),y∈V (Y −u)
 
> [dG (x, y) − dG∗1 (x, y)] = [dH (x, u) − dH (x, v)], (1)
x∈V (H),y∈V (Y ) x∈V (H−u−v)

W (G) − W (G∗2 )
 
= [dG (x, y) − dG∗2 (x, y)] + [dG (x, y) − dG∗2 (x, y)]
x∈V (X),y∈V (Y ) x∈V (X−v),y∈V (H)

> [dH (x, v) − dH (x, u)]. (2)
x∈V (H−u−v)

If W (G) − W (G∗1 ) ≤ 0, then by (1), x∈V (H−u−v) [dH (x, u) − dH (x, v)] < 0. Thus by
(2), W (G) − W (G∗2 ) > 0.
(iv) Let x = (xv1 , xv2 , . . . , xvn ) is the Perron vector of A(G), where xvi corresponds
to the vertex vi (1 ≤ i ≤ n). If xv ≥ xu , then ρ(G∗1 ) > ρ(G), and if xv < xu , then
ρ(G∗2 ) > ρ(G) by Lemma 2.2.

u2 uq−1 u
1
u1
H u Cq H u Cq−1
uq−1
uq−2
G G
Fig. 3
Let Fn be the nth Fibonacci number, i.e., F0 = F1 = 1, Fn = Fn−1 + Fn−2 . Note
that σ(Pn ) = Fn+1 , z(Pn ) = Fn .

Lemma 2.5. Suppose that G be a graph of order n ≥ 7 obtained from a connected

graph H ∼
= P1 and a cycle Cq = u0 u1 · · · uq−1 u0 (q ≥ 4) by identifying u0 with a vertex
u of the graph H (see Fig. 3). Let G = G − uq−1 uq−2 + uuq−2 . Then
(i) σ(G ) > σ(G);
(ii) z(G ) < z(G);
(iii) W (G ) < W (G);
(iv) ρ(G ) > ρ(G).
 - 190 -

Proof. (i) By Lemma 2.3, we have

σ(G) = σ(G − u) + σ(G − NG [u]) = Fq σ(H − u) + Fq−2 σ(H − NH [u]),

σ(G ) = σ(G − u) + σ(G − NG [u]) = 2Fq−1 σ(H − u) + Fq−3 σ(H − NH [u]).

Therefore σ(G ) − σ(G) = Fq−3 σ(H − u) − Fq−4 σ(H − NH [u]) > 0.

(ii) By Lemma 2.3, we have

z(G) = z(G − u) + z(G − u − u )
u ∈NG (u)

= (Fq−1 + 2Fq−2 )z(H − u) + z(H − u − u ),
u ∈NH (u)

z(G ) = z(G − u) + z(G − u − u )
u ∈NG (u)

= (2Fq−2 + 2Fq−3 )z(H − u) + z(H − u − u ).
u ∈NH (u)


Therefore z(G ) − z(G) = −Fq−4 z(H − u) < 0.
 3
q /8 if q is even,
(iii) By the definition of Wiener index, W (Cq ) =
(q 3 − q)/8 if q is odd.

Note that if q ≥ 4 is even, then q−2 j=0 dG (uq−1 , uj ) = 1 + 2(2 + 3 + · · · + q/2) =

(q 2 + 2q − 4)/4, thus

W (G) − W (G )
  q−1

= [dG (x, ui ) − dG (x, ui )] + [dG (ui , uj ) − dG (ui , uj )]
x∈V (H)−u0 i=0 0≤i<j≤q−1

q−2  q 3 (q − 1)3 − (q − 1) q 2 + 2q − 4
= dH (x, u0 ) + − −
2 8 8 4
x∈V (H)−u0

q − 2 q 3 (q − 1)3 − (q − 1) q 2 + 2q − 4
≥ + − −
2 8 8 4
2
q − 2q
= > 0.
8

Note that if q ≥ 5 is odd, then q−2j=0 dG (uq−1 , uj ) = 1 + 2(2 + 3 + · · · + (q − 1)/2) +
(q + 1)/2 = (q 2 + 2q − 3)/4, thus

W (G) − W (G )
  q−1

= [dG (x, ui ) − dG (x, ui )] + [dG (ui , uj ) − dG (ui , uj )]
x∈V (H)−u0 i=0 0≤i<j≤q−1
 - 191 -

q−3  q 3 − q (q − 1)3 q 2 + 2q − 3
= dH (x, u0 ) + − −
2 8 8 4
x∈V (H)−u0

q − 3 q 3 − q (q − 1)3 q 2 + 2q − 3
≥ + − − (3)
2 8 8 4
q 2 − 4q − 5
= . (4)
8
If q ≥ 6, then W (G) − W (G ) > 0 by (4); and if q = 5, then |V (H)| ≥ 3 by n ≥ 7,
and hence the inequality in (3) should be strictly. Therefore W (G) − W (G ) > 0.
(iv) Let x = (xu0 , xu1 , . . . , xun−1 ) is the Perron vector of A(G), where xui cor-
responds to the vertex ui (0 ≤ i ≤ n − 1, u0 = u). If xu ≥ xuq−1 , then let
G∗ = G − uq−1 uq−2 + uuq−2 . If xu < xuq−1 , then let

G∗ = G − uu1 − uw1 − · · · − uws + uq−1 u1 + uq−1 w1 + · · · + uq−1 ws ,

where NH (u) = {w1 , . . . , ws }. Then in either case, G∗ ∼

= G . Thus, by Lemma 2.2,

ρ(G ) > ρ(G).

Lemma 2.6. (i) σ(G0 (n, r)) = 3r 2n−2r−1 + 1;

(ii) z(G0 (n, r)) = 2r (n − r);
(iii) W (G0 (n, r)) = (n − 1)2 − r;
(iv) ρ(G0 (n, r)) is the root of λ4 − nλ2 − 2rλ + (n − 2r − 1) = 0.

Proof. (i) By Lemma 2.3, we get

σ(G0 (n, r)) = σ(G0 (n, r) − u) + σ(G0 (n, r) − NG0 (n,r) [u]) = 3r 2n−2r−1 + 1.
(ii) By Lemma 2.3, we get

z(G0 (n, r)) = z(G0 (n, r) − u) + z(G0 (n, r) − u − u )
uu ∈E
= 2r + (n − 2r − 1)2r + 2r2r−1 = 2r (n − r).

(iii) Note that W (K1,n ) = (n − 1)2 , and hence W (G0 (n, r)) = (n − 1)2 − r.
(iv) By Lemma 2.1, we have

φ(G0 (n, r); λ) = λn−2r (λ2 − 1)r − (n − 2r − 1)λn−2r−2 (λ2 − 1)r

−2rλn−2r (λ2 − 1)r−1 − 2rλn−2r−1 (λ2 − 1)r−1
= λn−2r−2 (λ2 − 1)r−1 (λ4 − nλ2 − 2rλ + (n − 2r − 1)).

Since ρ(G0 (n, r)) > 1, ρ(G0 (n, r)) is the root of λ4 − nλ2 − 2rλ + (n − 2r − 1) = 0.
 - 192 -

Let C(a1 , a2 , . . . , ar ; k) be a graph obtained from r cycles Cai , 1 ≤ i ≤ r and k

edges by taking one vertex of each cycle and each edge, and combining them as one

vertex. Denote C 0 (n, r) = {C(a1 , a2 , . . . , ar ; k) : ai ≥ 3, 1 ≤ i ≤ r, ri=1 (ai − 1) +
0 0
k + 1 = n}. Then C (n, r) ⊆ C (n, r) and G (n, r) = C(3, . . . , 3; n − 2r − 1).
  
r

3. Results
In this section, we derive the extremal cacti for the Wiener index, Merrifield-
Simmons index, Hosoya index and spectral radius by a unified approach.
In [3], Borovicanin and Petrovic show that G0 (n, r) is the maximal spectral radius
in the set G (n, r). Here, in order to discover the unification of our approach, we still
consider the spectral radius.
Denote f (G) ∈ {σ(G), ρ(G), −z(G), −W (G)}.

Theorem 3.1. Let G ∈ G (n, r), n ≥ 7. Then

f (G) ≤ f (G0 (n, r))

with equality holds if and only if G ∼

= G0 (n, r).

Proof. We have to prove that if G ∈ G (n, r), then f (G) ≤ f (G0 (n, r)) with
equality only if G ∼ = G0 (n, r).
Let Vc = {v ∈ V (G) : v is a cutvertex of G}.
Choose G ∈ G (n, r) such that f (G) is as large as possible. In the following, we
will show some facts.
Fact 1. G ∈ G 0 (n, r), i.e., |Vc | = 1.
Proof of Fact 1. Suppose that |Vc | > 1 . Let u, v ∈ Vc and H be a component
containing u, v with NG (u) \ NH (u), NG (v) \ NH (v) = ∅. Denote NG (u) \ NH (u) =
{w1 , w2 , . . . , ws } and NG (v) \ NH (v) = {v1 , v2 , . . . , vt }. Then s, t ≥ 1. Let G∗1 =
G − {uw1 , . . . , uws } + {vw1 , . . . , vws } and G∗2 = G − {vv1 , . . . , vvt } + {uv1 , . . . , uvt }.
Then G∗1 , G∗2 ∈ G (n, r). But, by Lemma 2.4, either f (G∗1 ) > f (G) or f (G∗2 ) > f (G),
a contradiction. Therefore |Vc | = 1.
By Fact 1, we let u denote the only cut-vertex of G.
Fact 2. G ∼ = G0 (n, r).
Proof of Fact 2. Assume that G ∼ = G0 (n, r). Then there exists a cycle Cq =
uu1 · · · uq−1 u with q ≥ 4. Let G = G − u1 u2 + uu2 . Then G ∈ G (n, r). By Lemma
2.5, f (G ) > f (G), a contradiction.
Therefore the proof of Theorem 3.1 is complete.
 - 193 -

In [13], Lu, Zhang and Tian prove that G0 (n, r) is the minimal Randić index in
the set G (n, r). Combining to Theorem 3.1, we have the following result.

Theorem 3.2. The maximal spectral radius [3], the maximal Merrifield-Simmons
index, the minimal Hosoya index, the minimal Wiener index and the minimal Randić
index [13] in the set G (n, r) (n ≥ 7) are obtained uniquely at G0 (n, r).

Acknowledgments. The authors are thankful to anonymous referee for his/her

useful comments.

References
[1] B. Bollobás, Modern Graph Theory, Springer-Verlag, New York, Inc., 1998.

[2] J.A. Bondy and U.S.R. Murty, Graph Theory with its Applications, Macmillan,
New York, 1976.

[3] B. Borovicanin and M. Petrovic, On the index of cactuses with n vertices, Publ.
Inst. Math. Beograd 79 (2006) 13-18.

[4] O. Chan, I. Gutman, T.K. Lam and R. Merris, Algebraic connections between
topological indices, J. Chem. Inform. Comput. Sci. 38(1998) 62-65.

[5] S.J. Cyvin, I. Gutman and N. Kolakovic, Hosoya index of some polymers,
MATCH Commun. Math. Comput. Chem. 24(1989) 105-117.

[6] I. Gutman, On the Hosoya index of very large molecules, MATCH Commun.
Math. Comput. Chem. 23(1988) 95-103.

[7] I. Gutman, Extremal hexagonal chains, J. Math. Chem. 12(1993) 197-210.

[8] I. Gutman and O.E. Polansky, Mathematical Concepts in Organic Chemistry,

Springer, Berlin, 1986.

[9] H. Hosoya, Topological index, a newly proposed quantity characterizing the

topological nature of structural isomers of saturated hydrocarbons, Bull. Chem.
Soc. Jpn. 44(1971) 2332-2339.

[10] L.B. Kier and L.H. Hall, Molecular Connectivity in Chemistry and Drug Re-
search, Academic Press, San Francisco, 1976.
 - 194 -

[11] L.B. Kier and L.H. Hall, Molecular Connectivity in Structure-Activity Analysis,
Wiley, 1986.

[12] X. Li and J. Zheng, A unified approach to the extremal trees for different indices,
MATCH Commun. Math. Comput. Chem. 54(2005) 195-208.

[13] M. Lu, L.Z. Zhang and F. Tian, On the Randić index of cacti, MATCH Com-
mun. Math. Comput. Chem. 56(2006) 551-556.

[14] R.E. Merrifield and H.E. Simmons, Topological Methods in Chemistry, Wiley,
New York, 1989.

[15] Z. Mihatić and N. Trinajstić, A graph theoretical approach to structure-property

relationship, J. Chem. Educ. 69(1992) 701-702.

[16] R. Todeschini and V. Consonni, Handbook of Molecular Descriptors, Wiley-

VCH, Weinheim, 2000.

[17] L. Türker, Contemplation on the Hosoya indices, Journal of Molecular Structure

(Theochem) 623(2003) 75-77.

[18] M. Randić, On characterization of molecular branching, J. Amer. Chem. Soc.

97(1975) 6609-6615.

[19] A.J. Schwenk and R.J. Wilson, On the eigenvalues of a graph, In: Selected
Topics in Graph Theory, Academic Press, New York, 1978, 307-336.

[20] H. Wiener, Structural determination of paraffin boiling point, J. Amer. Chem.

Soc. 69(1947) 17-20.

[21] H. Wiener, Vapor pressure-temperature relationships among the branched paraf-

fin hydroarbons, J. Phys. Chem. 52(1948) 425-430.

[22] B.F. Wu, E.L. Xiao and Y. Hong, A sharp bound on the largest eigenvalue of
the Laplacian matrix of a graph, Linear Algebra Appl. 395(2005) 343-395.

[23] A.M. Yu and F. Tian, A Kind of graphs with Minimal Hosoya Indices and
Maximal Merrifield-Simmons Indices, MATCH Commun. Math. Comput. Chem.
55(2006) 103-118.

[24] L.Z. Zhang and F. Tian, Extremal catacondensed benzenoids, J. Math. Chem.
34(2003) 111-122.