On the quotients between the (revised) Szeged index and Wiener index of graphs∗

arXiv:1606.05950v1 [math.CO] 20 Jun 2016

Jing Chena , Shuchao Lib , Huihui Zhangb,† a b

Department of Mathematics, Hunan First Normal University, Changsha, Hunan 410205, China

Faculty of Mathematics and Statistics, Central China Normal University, Wuhan 430079, P.R. China

Abstract: Let Sz(G), Sz ∗ (G) and W (G) be the Szeged index, revised Szeged index and Wiener index of a graph G. In this paper, the fourth, fifth, sixth and seventh largest Wiener indices among all unicyclic graphs of order n ≥ 10 are determined, respectively; as well the first, second, third, and fourth largest Wiener indices among all bicyclic graphs of order n ≥ 7 are determined, respectively. All the corresponding extremal graphs are identified. Based on these results, further relation on the quotients between the (revised) Szeged index and the Wiener index are studied. Sharp lower bound on Sz(G)/W (G) is determined for all connected graphs each of which contains at least one non-complete block. As well the connected graph with the second smallest value on Sz ∗ (G)/W (G) is identified for G containing at least one cycle. Keywords: Szeged index; Revised Szeged index; Wiener index AMS subject classification: 05C69; 05C05 1. Introduction We consider that all graphs in this paper are finite, undirected and simple. We follow the notation and terminologies in [3] except otherwise stated. Let G = (VG , EG ) be a connected graph with vertex set VG and edge set EG . We call a connected graph G a unicyclic graph (resp. bicyclic graph) if |EG | = |VG | (resp. |EG | = |VG |+1). For convenience, let |G| := |VG |. In the subsequent sections, we use G − v, or G − uv to denote the graph obtained from G by deleting vertex v ∈ VG , or edge uv ∈ EG , respectively (it is naturally extended if at least two vertices or edges are deleted). Let G + uv be the graph obtained from G by adding an edge uv 6∈ EG . For a subset S of VG , denoted by G[S] the subgraph induced by S. For v ∈ VG , denoted by NG (v) (or N (v) for short) the set of all neighbors of v in G and let d(v) = |NG (v)| be the degree of v in G. In particular, let △(G) = max{d(x)|x ∈ VG } and δ(G) = min{d(x)|x ∈ VG }. Call u a pendant vertex in G, if d(u) = 1. Denoted by Pn , Cn , Sn and Kn the path, cycle, star and complete graph of order n, respectively. We call Ln,r a lollipop if it is obtained by identifying some vertex of Cr with an end-vertex of Pn−r+1 . A graph is cyclic if it is connected having at least one cycle. Recall that G is called k-connected if |G| > k and G − X is connected for every set X ⊆ VG with |X| < k. The greatest integer k such that G is k-connected is the connectivity κ(G) of G. In particular, let κ(Kn ) = n − 1 for all n ≥ 1. Clear, κ(G) = 0 if and only if G is disconnected or K1 . For u, v ∈ VG , the distance between u and v in G, denoted by dG (u, v), is the length of a shortest path connecting u and v. The diameter, d(G), of G is equal to maxu,v∈VG dG (u, v). The Wiener index of G is defined as the sum of all distances between pairs of unordered vertices in G, i.e., W (G) =

X

dG (u, v) =

1 X DG (u), 2

(1.1)

u∈VG

{u,v}⊆VG

∗ Financially supported by the National Natural Science Foundation of China (Grant Nos. 11271149, 11371162), the Program for New Century Excellent Talents in University (Grant No. NCET-13-0817) and the Special Fund for Basic Scientific Research of Central Colleges (Grant No. CCNU13F020)). † E-mail: [email protected] (J. Chen), [email protected] (S.C. Li), [email protected] (H.H. Zhang)

1

P where DG (u) = x∈VG dG (u, x). It was first proposed by Wiener in 1947, while its equivalent definition as in (1.1) was given by Hosoya in 1971; see [18]. This graph invariant was extensively studied, in which most of the researchers focused on the lower bounds of Wiener index among a class of graphs, one may be referred to [9, 14, 24, 26, 27, 32] and the references therein. There are only a few papers concerning the upper bounds on the Wiener index of graphs; see [11, 39, 37, 12]. Recently, Mukwembi and Vetr´ık [28] obtained sharp upper bounds on the Wiener index of trees with diameter at most 6, which shows that it is very difficult to obtain the upper bound on the Wiener index of some given class of graphs. Given an edge e = uv in G, define three sets with respect to e as follows: Nu (e) = {w ∈ VG : dG (u, w) < dG (v, w)},

Nv (e) = {w ∈ VG : dG (v, w) < dG (u, w)},

N0 (e) = {w ∈ VG : dG (u, w) = dG (v, w)}. Clearly, VG = Nu (e) ∪ Nv (e) ∪ N0 (e). For convenience, let nu (e) = |Nu (e)|, nv (e) = |Nv (e)| and n0 (e) = |N0 (e)|. It is easy to see nu (e) + nv (e) + n0 (e) = |VG |. If G is bipartite, then N0 (e) = ∅ holds for all e ∈ EG . Consequently, for any bipartite graph G with e ∈ EG , nu (e) + nv (e) = |VG |. From [10, 16, 34], we know that, for a tree T , its Wiener index can be defined alternatively as W (T ) =

X

nu (e)nv (e).

(1.2)

e=uv∈ET

Motivated from (1.2), Gutman [15] introduced the Szeged index of graph G, which is defined by Sz(G) =

X

nu (e)nv (e).

e=uv∈EG

Randi´c [32] observed that the vertices at equal distances from the end-vertices of an edge does not contribute to the Szeged index, and so he proposed a modified version of the Szeged index as the revised Szeged index, which is defined by ∗

Sz (G) =

X

e=uv∈EG

   n0 (e) n0 (e) nu (e) + nv (e) + . 2 2

For more recent results on (revised) Szeged index, one may be referred to these in [2, 19, 22, 23, 31, 33, 35]. It is known that W (G) ≤ Sz(G) ≤ Sz ∗ (G) holds for a connected graph G. Dobrynin and Gutman [8] studied the structure of G with the property of Sz(G) ≥ W (G) and conjectured that Sz(G) = W (G) holds if and only if every block of G is complete. This conjecture was confirmed by the same authors in [9]. Later, Khodashenas, Nadjafi-Arani, Ashrafi and Gutman [20] shortened the confirmed proof for this result. Pisanski and Randi´c [31] showed that, if G is connected, then Sz ∗ (G) ≥ Sz(G) with equality if and only if G is bipartite. Thus, Sz ∗ (G) = Sz(G) = W (G) holds if and only if G is a tree. It is interesting to study the relation between the (revised) Szeged index and Wiener index. Nadjafi-Arani, Khodashenas and Ashrafi [29] investigated the structure of a graph G satisfying Sz(G) − W (G) = n, where n is an integer other than the order of G. As well they [30] characterized the graphs whose Szeged index and Wiener index differ by 4 and 5. Klavzar and Nadjafi-Arani [22] obtained some bounds for Wiener index and the Szeged index for connected molecular graphs, and they also studied Sz(G) − W (G) of networks; see [21]. Several graph theorists use the AutoGraphiX to study the relationship involving graph invariants. One may be referred to [1, 4, 13] for more detailed information. Hansen et al. [17] used the computer program AutoGraphiX to generate eight conjectures on the difference (resp. quotient) between the (revised) Szeged index and Wiener index. Chen, Li and Liu [5, 6] confirmed three of the above conjectures, which can be summarized as follows.

2

Theorem 1.1 ([5, 6]). Let G be a connected bipartite graph with n ≥ 4 vertices and |EG | ≥ n edges. Then Sz(G) − W (G) ≥ 4n − 8. The equality holds if and only if G is composed of a cycle C4 on 4 vertices and a tree T on n − 3 vertices sharing a single vertex. Theorem 1.2 ([6]). Let G be a connected graph with n ≥ 5 vertices with an odd cycle and girth g ≥ 5. Then Sz(G) − W (G) ≥ 2n − 5. The equality holds if and only if G is composed of a cycle C5 on 5 vertices, and one tree rooted at a vertex of the C5 or two trees, respectively, rooted at two adjacent vertices of the C5 . Theorem 1.3 ([6]). Let G be a connected graph with n ≥ 4 vertices and |EG | ≥ n edges and with an odd cycle. Then

n2 + 4n − 6 . 4 The equality holds if and only if G is composed of a cycle C3 on 3 vertices and a tree T on n − 2 vertices sharing Sz ∗ (G) − W (G) ≥

a single vertex. Recently, Zhao and two authors of the current paper studied the further relation between the Wiener index and the (revised) Szeged index; see [38]. Along this line, Li and Zhang [25] also proved three of the above conjectures on the quotients between the (revised) Szeged index and Wiener index, which are described as the following theorems. Theorem 1.4 ([25]). Let G be a connected graph of n ≥ 4 vertices containing at least one cycle. (i) If G is a bipartite graph, then

24(n − 2) Sz ∗ (G) ≥1+ 3 W (G) n − 13n + 36 with equality if and only if G is the lollipop Ln,4 .

(ii) If G is a non-bipartite graph, then 3(n2 + 4n − 6) Sz ∗ (G) ≥1+ W (G) 2(n3 − 7n + 12) with equality if and only if G is the lollipop Ln,3 . Theorem 1.5 ([25]). Let G be a unicyclic graph on n ≥ 4 vertices. Then  Sz(G) 2 − n28+7 , if n is odd, ≤ 2, if n is even W (G) with equality if and only if G is the lollipop Ln,n−1 if n is odd and the cycle Cn if n is even. Theorem 1.6 ([25]). Let G be a unicyclic graph on n ≥ 4 vertices. Then  Sz ∗ (G) 2 + n22−1 , if n is odd, ≤ 2, if n is even W (G) with equality if and only if G is the cycle Cn . Further more, Li and Zhang [25] as well determined sharp lower bounds on Sz(G)/W (G) for cyclic graph G with girth g ≥ 4. Theorem 1.7 ([25]). Let G be a connected graph of order n ≥ 5 with at least one cycle and girth g ≥ 4. (i) If G is a bipartite graph, then 24(n − 2) Sz(G) ≥1+ 3 W (G) n − 13n + 36 with equality if and only if G is the lollipop Ln,4 . 3

H1

H2 ... G′′

G′ Figure 1: Graphs G′ and G′′ .

(ii) If G is a non-bipartite graph, then Sz(G) 6(2n − 5) ≥1+ 3 W (G) n − 25n + 90 with equality if and only if G is the lollipop Ln,5 . It is difficult to determine sharp bounds on quotients between the (revised) Szeged index and Wiener index. For example, for unicyclic graph G, Hansen et al. [17] proposed conjectures on sharp upper bounds of Sz(G)/W (G) and Sz ∗ (G)/W (G). Fortunately, we give confirmed proofs for them (see Theorems 1.5 and 1.6). Given a connected graph G, Dobrynin and Gutman [9] showed that Sz(G)/W (G) ≥ 1 with equality if and only if every block of G is complete. Note that Sz ∗ (G)/W (G) ≥ 1 with equality if and only if G is a tree. By Theorem 1.4, we know that Ln,4 is the unique graph with the smallest value on Sz ∗ (G)/W (G) among n-vertex cyclic graphs. Theorem 1.7 only considers the case for the graph of girth at least 4, whereas the case for cyclic graph of girth 3 is still open. Motivated from these facts, it is natural and interesting for us to study the further relation on the quotients between the (revised) Szeged index and Wiener index of connected graphs. 2. Preliminaries In this section, we give some preliminary results which are used to prove our main results in the subsequent sections. (0) (1) For convenience, let K4− be the graph obtained by deleting an edge from K4 . Let Bn (resp. Bn ) be the bicyclic graph of order n obtained by identifying a pendant vertex of Pn−3 with a vertex in K4− of degree 2 (resp. (0,s) degree 3). Denote by Bn a graph of order n obtained by attaching paths Pn−s−4 and Ps to two 2-degree (0)

(0,0)

vertices of K4− , respectively, 0 ≤ s ≤ ⌊ n−4 2 ⌋. In particular, Bn = Bn

. (0,s)

Lemma 2.1 ([36]). Let G be an n-vertex bicyclic graph of diameter n − 2. Then G ∈ {Bn

: 0 ≤ s ≤ ⌊ n−4 2 ⌋}.

Lemma 2.2 ([38]). Let G be an n-vertex connected graph of girth g = 3 and at least one block is non-complete. Then Sz(G) − W (G) ≥ 2, the equality holds if and only if G ∼ = G′ , where G′ is depicted in Fig. 1 satisfying each block of H1 (resp. H2 ) being a complete graph. Tang and Deng [37] showed that if G is an n-vertex unicyclic graph with n ≥ 6, then W (G) ≤ W (Ln,3 ). In fact, if n = 4, 5, it is routine to check that this result also holds. Hence, Lemma 2.3. Let G be a unicyclic graph on n ≥ 4 vertices, then W (G) ≤ W (Ln,3 ). Given a cycle Cr whose vertices are labeled consecutively by v1 , v2 , . . . , vr , then let Cr (T1 , T2 , . . . , Tr ) be the graph obtained from Cr and rooted trees Ti ’s by identifying the root, say ri , of Ti with vi on Cr , i = 1, 2, . . . , r. We Pr may assume that |Ti | = ni + 1, ni ≥ 0, i = 1, 2, . . . , r. Thus, |Cr (T1 , T2 , . . . , Tr )| = i=1 |Ti |. In particular, if every rooted tree is a star whose root is just its center, then we denote Cr (T1 , T2 , . . . , Tr ) by Cr (Sn1 +1 , Sn2 +1 , . . . , Snr +1 ); whereas if the non-trivial rooted trees are just Ti , Tj , . . . , Tk (i, j, . . . , k ∈ {1, 2, . . . , r}), then we denote it by Cri,j,...,k (Ti , Tj , . . . , Tk ). 4

ˆ = Cri,j,...,k (Sni +1 , Snj +1 , . . . , Sn +1 ) and G ˜ = Cri,j,...,k (Pni +1 , Pnj +1 , . . . , Pn +1 ). Then Lemma 2.4 ([37]). Let G k k ˆ ≤ W (G) ≤ W (G) ˜ W (G) for any graph G = Cri,j,...,k (Ti , Tj , . . . , Tk ) with |Tt | = nt + 1, t = i, j, . . . , k. The equality on the left (resp. right) ˜ ˆ (resp. G ∼ holds if and only if G ∼ = G). =G Lemma 2.5 ([25]). Let G be an n-vertex unicyclic graph with girth r, r ≥ 5. If r is odd, then W (G) ≤ (n3 − 25n + 90)/6 with equality if and only if G ∼ = Ln,5 . Let Tn (n1 , n2 , . . . , nl ) be an n-vertex tree obtained from a star Sl+1 by replacing its l edges by l paths Pl Pn1 , Pn2 , . . . , Pnl with i=1 ni = n − 1. Furthermore, if the number of nk in Tn (n1 , n2 , . . . , nl ) is mk , we denote

k it nm for short. k

Lemma 2.6 ([11]). Let T be a tree of order n ≥ 9. If T ≇ Pn , Tn (n − 3, 1, 1), Tn(n − 4, 2, 1), Tn(n − 5, 3, 1), G′′ , where G′′ is depicted in Fig. 1. Then W (G) < W (Tn (n − 5, 3, 1)) < W (G′′ ) < W (Tn (n − 4, 2, 1)) < W (Tn (n − 3, 1, 1)) < W (Pn ). By direct calculation, we may obtain the Wiener index of each n-vertex trees for 5 ≤ n ≤ 8 (based on Table 2 of the Appendix in [7]). Together with Lemma 2.6, one may obtain Lemma 2.7. Let T be a tree of order n ≥ 5. If T ≇ Pn , Tn (n− 3, 1, 1), then W (T ) < W (Tn (n− 3, 1, 1)) < W (Pn ). Further more, if n = 8 and T ≇ P8 , T8 (5, 1, 1), T8(4, 2, 1), T8 (3, 3, 1), G′′ , then W (T ) < W (G′′ ) < W (T8 (3, 3, 1)) < W (T8 (4, 2, 1)) < W (T8 (5, 1, 1)) < W (Pn ), where G′′ is the tree on 8 vertices as depicted in Fig. 1. Lemma 2.8. Let G be a graph obtained from vertex disjoint connected graphs G1 and G2 by identifying a vertex of G1 with a vertex of G2 and denote the common vertex by v. Then W (G) = W (G1 ) + W (G2 ) + (|G2 | − 1)DG1 (v) + (|G1 | − 1)DG2 (v). Proof. Note that dG (x, y) = dGi (x, y) for {x, y} ⊆ VGi , i = 1, 2. By (1.1), we have W (G)

=

X

dG (x, y) +

X

dG1 (x, y) +

x,y∈VG1

=

x,y∈VG1

=

X

x,y∈VG2

X

dG2 (x, y) +

X

(dG1 (x, v) + dG2 (v, y))

x∈VG1 \{v},y∈VG2 \{v}

X

dG1 (x, v) +

W (G1 ) + W (G2 ) + (|G2 | − 1)

X

dG1 (x, v) + (|G1 | − 1)

x∈VG1 \{v}

=

X

dG2 (y, v)

x∈VG1 \{v},y∈VG2 \{v}

x∈VG1 \{v},y∈VG2 \{v}

=

dG (x, y)

x∈VG1 \{v},y∈VG2 \{v}

x,y∈VG2

W (G1 ) + W (G2 ) +

X

dG (x, y) +

W (G1 ) + W (G2 ) + (|G2 | − 1)DG1 (v) + (|G1 | − 1)DG2 (v),

X

dG2 (y, v)

y∈VG2 \{v}

as desired. The following result is the direct consequence of the above lemma. Corollary 2.9. Let u be a pendant vertex of an n-vertex connected graph G and v be the unique neighbor of u. Then W (G) = W (G − u) + DG−u (v) + n − 1.

5

Proof. Clearly v is a cut vertex of G. Let G1 = G − u and G2 be an edge, say uv. Then our result follows directly by Lemma 2.8. Lemma 2.10. Let Ck be a cycle and H be a connected graph containing at least one edge with v ∈ VH . Suppose that G1 is a graph obtained by identifying the vertex v of H with a vertex in Ck , whereas G2 is a graph obtained by identifying the vertex v of H with a minimal degree vertex of Lk,3 . Then W (G1 ) ≤ W (G2 ) with equality if and only if k = 3. Proof. Clearly, if k = 3, then G1 ∼ = G2 . So we consider k ≥ 4 in that follows. By Lemma 2.8, we have W (G1 ) =

W (H) + W (Ck ) + (k − 1)DH (v) + (|H| − 1)DCk (v),

W (G2 ) =

W (H) + W (Lk,3 ) + (k − 1)DH (v) + (|H| − 1)DLk,3 (v).

Thus, W (G1 ) − W (G2 ) = W (Ck ) − W (Lk,3 ) + (|H| − 1)(DCk (v) − DLk,3 (v)).

(2.1)

It is easy to see that |H| ≥ 2. On the one hand, by Lemma 2.3, we have W (Ck ) ≤ W (Lk,3 ). On the other hand, 2

2

= DCk (v) − DLk,3 (v) ≤ k4 − k −k−2 2 W (G1 ) < W (G2 ), as desired.

−k2 +2k+4 4

< 0 for k ≥ 4. In view of (2.1), we have W (G1 ) − W (G2 ) < 0, i.e.,

For simplicity, let Bn,k := Tn (1k , (n − k)1 ). A leaf in Bn,k is called a unit leaf if it is adjacent to the unique i maximum degree vertex. Let Bn,k be a graph obtained by adding i edges to connect unit leaves in Bn,k .

Lemma 2.11. Let G be an n-vertex connected graph of diameter d ≤ n − 3. For all v in VG , we have DG (v) ≤ i (n2 − n − 6)/2 with equality if and only if G ∈ {Bn,4 : i = 0, 1, 2, 3} and v is the farthest pendant vertex from the i maximum degree of Bn,4 , 0 ≤ i ≤ 3.

Proof. For all v ∈ VG , by the definition of DG (v) we have X DG (v) = dG (x, v) x∈VG

≤

= ≤

(1 + 2 + · · · + d) + d(n − d − 1) " # 2 1 2n − 1 (2n − 1)2 − d− − 2 2 4 n2 − n − 6 . 2

(Since d ≤ n − 3)

(2.2)

(2.3)

Equality in (2.2) holds if and only if there are exactly n − d vertices each of which is of distance d from v, 2

whereas equality in (2.3) holds if and only if d = n − 3. Hence, DG (v) = n −n−6 if and only if d = n − 3, v is 2 an end-point of a diameter of G and there are exactly 3 vertices in G such that each is of distance d from v, i.e., i G ∈ {Bn,4 : i = 0, 1, 2, 3}.

3. Upper bounds on W (G) Tang and Deng [37] identified the first to third largest Wiener indices in the class of all unicyclic graphs of order n ≥ 6. In this section, we not only determine the fourth, fifth, sixth, seventh largest values on the Wiener index of graphs among all unicyclic graphs of order n ≥ 10, but also obtain the first, second, third and fourth greatest values on the Wiener index of graphs among all bicyclic graphs of order n ≥ 7. The corresponding extremal graphs are also identified, respectively. All the obtained results here are used to prove our main results in Section 4. 6

3.1. Unicyclic graphs By the definition of Cr (T1 , T2 , . . . , Tr ) in Section 2, every unicyclic graph G can be viewed as the form Cr (T1 , T2 , . . . , Tr ), where r is the girth of G and Ti is the rooted tree of G. Lemma 3.1. Let G = Cr (Pn1 +1 , Pn2 +1 , . . . , Pnr +1 ) be an n-vertex unicyclic graph with nk , nt ≥ 1 for k < t. (i) If

(ii) If

P

i6=k,t (ni

P

+ 1)dG (vi , vk ) ≤

P

+ 1)dG (vi , vt ), then

P

+ 1)dG (vi , vt ), then

i6=k,t (ni

W (G) < W (Cr (Pn1 +1 , . . . , Pnk−1 +1 , P1 , Pnk+1 +1 . . . , Pnt−1 +1 , Pnk +nt +1 , Pnt+1 +1 . . . , Pnr +1 )).

i6=k,t (ni

+ 1)dG (vi , vk ) >

i6=k,t (ni

W (G) < W (Cr (Pn1 +1 , . . . , Pnk−1 +1 , Pnk +nt +1 , Pnk+1 +1 . . . , Pnt−1 +1 , P1 , Pnt+1 +1 . . . , Pnr +1 )).

Proof. For convenience, denote by Pnk +1 = u1 u2 . . . unk +1 and Pnt +1 = w1 w2 . . . wnt +1 , where u1 = vk , w1 = vt . (i) Let G′ = G − u1 u2 + u2 wnt +1 , i.e., G′ = Cr (Pn1 +1 , . . . , Pnk−1 +1 , P1 , Pnk+1 +1 . . . , Pnt−1 +1 , Pnk +nt +1 , Pnt+1 +1 . . . , Pnr +1 ). In what follows, we show that W (G) < W (G′ ). Let A = (VG \VPnk +1 ) ∪ {u1 }. By the definition of W (G), we have W (G)

=

W (Pnk +1 ) +

X

dG (x, y) +

X

dG (x, y) +

X

dG (x, y) + (n − nk − 1)

x,y∈A

=

W (Pnk +1 ) +

x∈VPn

x,y∈A

=

W (Pnk +1 ) +

x∈VPn

k +1

k

X

dG (x, y)

X

(dG (x, vk ) + dG (vk , y))

\{u1 },y∈A\{u1 }

\{u1 },y∈A\{u1 } +1

x,y∈A

X

x∈VPn

k

dG (x, vk ) + nk

X

dG (vk , y)

y∈A

\{u1 } +1

Similarly, we have W (G′ ) = W (Pnk +1 ) +

X

X

dG′ (x, y) + (n − nk − 1)

x,y∈A

x∈VPn

k

dG′ (x, wnt +1 ) + nk

X

dG′ (wnt +1 , y).

y∈A

\{u1 } +1

Note that dG (x, y) = dG′ (x, y), dG (wnt +1 , y) = dG′ (wnt +1 , y) and dG (z, vk ) = dG′ (z, wnt +1 ) for all x, y ∈ A, z ∈ VPnk +1 \{u1 }. Then 

W (G) − W (G′ ) = nk 

X

dG (vk , y) −

y∈A

X

y∈A

Thus, in order to prove W (G) < W (G′ ), it suffices to show that direct calculation, we have X

dG (vk , y) =

X

X

i6=k y∈VPn

y∈A

=



dG (wnt +1 , y) .

P

y∈A dG (vk , y)

−

(3.1) P

y∈A dG (wnt +1 , y)

(dG (vk , vi ) + dG (vi , y))

i +1

(nt + 1)dG (vk , vt ) +

X

i6=k,t

7

(ni + 1)dG (vk , vi ) +

X i6=k

DPni +1 (vi )

< 0. By

and X

dG (wnt +1 , y) =

nt + dG (vk , vt ) +

X

X

i6=k,t y∈VPn

y∈A

=

(nt + dG (vt , vi ) + dG (vi , y)) + DPnt +1 (vt )

i +1

(n − nk − nt − 1)nt + dG (vk , vt ) +

X

(ni + 1)dG (vt , vi ) +

i6=k,t

X

DPni +1 (vi ).

i6=k

P P Bearing in mind that i6=k,t (ni + 1)dG (vi , vk ) ≤ i6=k,t (ni + 1)dG (vi , vt ), we have X X dG (vk , y) − dG (wnt +1 , y) ≤ nt (dG (vk , vt ) − n + nk + nt + 1). y∈A

(3.2)

y∈A

P If dG (vk , vt ) = 1, then dG (vk , vt ) − n + nk + nt + 1 = nk + nt + 2 − n = − i6=k,l (ni + 1) ≤ −1. Together with (3.1) and (3.2), we have W (G) < W (G′ ). P If dG (vk , vt ) ≥ 2, then one may observe that n−nk −nt −2 = i6=k,l (ni +1) ≥ 2dG (vk , vt )−2. Thus, dG (vk , vt )− P P n + nk + nt + 1 ≤ 1 − dG (vk , vt ) ≤ −1. Thus, in view of (3.2) we have y∈A dG (vk , y) < y∈A dG (wnt +1 , y). Together with (3.1), we obtain W (G) < W (G′ ), as desired. (ii) By the same discussion as in the proof of (i), we may show that (ii) holds, which is omitted here.

Lemma 3.2. Let Cr (T1 , T2 , . . . , Tr ) be an n-vertex unicyclic graph with r ≥ 6. If r is even, then W (Cr (T1 , T2 , . . . , Tr )) ≤

n3 − 37n + 168 6

with equality if and only if Cr (T1 , T2 , . . . , Tr ) ∼ = Ln,6 if n 6= 8 and Cr (T1 , T2 , . . . , Tr ) ∼ = L8,6 or C8 otherwise. n3 −37n+168 . 6

Repeated using Lemmas 2.4 and 3.1 yields     1 3 3 5 3 W (Cr (T1 , T2 , . . . , Tr )) ≤ W (Ln,r ) = n + − r2 + 3r − 1 n + r − 3r2 + r 6 2 4

Proof. By Lemma 2.8, one has W (Ln,6 ) =



(3.3)

with equality if and only if G ∼ = Ln,r . If r ≥ 8, by direct calculation we have     3 2 5 3 1 r − 3r − 36 n − r − 3r2 + r − 168 W (Ln,6 ) − W (Ln,r ) = 6 2 4     1 3 2 3 5 3 ≥ r − 3r − 36 r − r − 3r2 + r − 168 (As n ≥ r, r2 − 3r − 36 > 0) (3.4) 6 2 4 2   1 1 3 r − 37r + 168 (As r ≥ 8) = 6 4 ≥ 0,

(3.5)

where the equality in (3.4) holds if and only if n = r; while the equality in (3.5) holds if and only if r = 8. Hence, W (Ln,6 ) = W (Ln,r ) if and only if n = r = 8, i.e., Ln,r ∼ = C8 . 3

So, together with (3.3) and (3.5), it follows that W (Cr (T1 , T2 , . . . , Tr )) ≤ n −37n+168 . The equality holds if 6 ∼ ∼ and only if Cr (T1 , T2 , . . . , Tr ) = Ln,6 if n 6= 8, and Cr (T1 , T2 , . . . , Tr ) = L8,6 or C8 if n = 8. This completes the proof. 3 As depicted in Fig. 2, Hn0 , Hn1 and Hn2 are n-vertex unicyclic graphs, whereas H10 is a 10-vertex unicyclic 0 1 graph. Let An = {Ln,3 , Ln,4 , Hn , Hn , C3 (Pn−3 , P2 , P1 ), C3 (Pn−4 , P3 , P1 ), C4 (Pn−4 , P1 , P2 , P1 )}.

Theorem 3.3. Let G = Cr (T1 , T2 , . . . , Tr ) be a connected unicyclic graph with n ≥ 10 vertices. Suppose that 2 G 6∈ An , then W (G) ≤ (n3 − 19n + 54)/6. The equality holds if and only if G ∼ , or = Hn2 if n ≥ 12, G ∼ = H11 2 3 H or H if n = 10. C3 (P6 , P4 , P1 ) if n = 11 and G ∼ = 10

10

8

... ...

Hn0

...

Hn2

Hn1

3 H10

3 Figure 2: Graphs Hn0 , Hn1 , Hn2 and H10 .

3

. If r ≥ 5 is odd, by Lemma 2.5, we have W (G) ≤ Proof. By direct calculation, we have W (Hn2 ) = n −19n+54 6 3 3 n3 −25n+90 n3 −19n+54 < , whereas if r ≥ 6 is even, by Lemma 3.2, we have W (G) ≤ n −37n+168 < n −19n+54 . 6 6 6 6 Hence, in order to complete the proof, it suffices to consider r = 3 and r = 4. Then choose an n-vertex unicyclic graph G 6∈ An of girth r ≤ 4 such that W (G) is as large as possible. Recall that for each rooted tree Ti in G, one has |Ti | = ni + 1. Hence, by Lemma 2.4, we have G = Cr (Pn1 +1 , Pn2 +1 , . . . , Pnr +1 ). If there exist at least three non-trivial rooted trees in G, then without loss of generality, we may assume that T1 , Tk are the non-trivial rooted trees with the first two smallest sizes. By Lemma 3.1, we have W (G) < W (G′ ) P P if i6=1,k (ni + 1)dG (vi , v1 ) ≤ i6=1,k (ni + 1)dG (vi , vk ); and W (G) < W (G′′ ) otherwise, where G′ =Cr (P1 , . . . , Pnk−1 +1 , Pn1 +nk +1 , Pnk+1 +1 , . . . , Pnr +1 ),

G′′ =Cr (Pn1 +nk +1 , . . . , Pnk−1 +1 , P1 , Pnk+1 +1 , . . . , Pnr +1 ). If both G′ and G′′ are not in An , then we obtain a contradiction to the maximality of W (G); otherwise, it is 3 3 < n −19n+54 . routine to check that G ∼ = C3 (Pn−4 , P2 , P2 ). By simple computing, we have W (G) = n −25n+96 6 6 Therefore, there are at most two non-trivial rooted trees contained in G. Bearing in mind that r = 3 or r = 4. Hence, we proceed by considering the following two possible cases. Case 1. r = 3. In this case, if G contains just one non-trivial rooted tree. Assume, without loss of generality, that T1 is the non-trivial rooted tree. By Lemma 2.8, we have W (G) = W (T1 ) + 2DT1 (v1 ) + 2n − 3.

(3.6)

Note that G ≇ Ln,3 , Hn0 , Hn1 . Hence, we have T1 ≇ Pn−2 , Tn−2 (n − 5, 1, 1), Tn−2 (n − 6, 2, 1). If d(T1 ) = n − 4, then it is routine to check that DT1 (v1 ) ≤ 1 + 2 + · · · + (n − 4) + n − 6 =

n2 − 5n . 2

Equality holds if and only if T1 ∼ = Tn−2 (n − 7, 3, 1) and v1 is the farthest pendant vertex from the maximum degree vertex of Tn−2 (n − 7, 3, 1). 2

2

= n −5n with equality if and only if If d(T1 ) ≤ n − 5, by Lemma 2.11, we have DT1 (v1 ) ≤ (n−2) −(n−2)−6 2 2 T1 ∼ = Bn−2,4 and v1 is the farthest pendant vertex from the maximum degree vertex of Bn−2,4 . Therefore, we obtain 2DT1 (v1 ) ≤ n2 − 5n with equality if and only if T1 ∼ = Tn−2 (n − 7, 3, 1), or Bn−2,4 and v1 is the farthest pendant vertex from the maximum degree vertex of Tn−2 (n − 7, 3, 1), or Bn−2,4 . In view of (3.6), we determine the sharp upper bound on W (T1 ) in what follows. If |G| = 10, then |T1 | = 8. By Lemma 2.7(ii), we have W (T1 ) ≤ W (T8 (3, 3, 1)) = 75. The equality holds if and only if T1 ∼ = T8 (3, 3, 1). By (3.6), together with 2DT1 (v1 ) ≤ 102 − 50 = 50, one has W (G) ≤ 75 + 50 + 19 = 144. Equality holds if and only if T1 ∼ = T8 (3, 3, 1), i.e., G ∼ = H3 . 10

Next we consider |G| ≥ 11. If T1 ∼ = G′′ (G′′ is depicted in Fig. 1), then 2DT1 (v1 ) ≤ (n − 5)(n − 4) + 2(n − 3) = 3 3 2 < n −19n+54 for n ≥ 11. If n − 7n + 14. By direct calculation (based on (3.6)), we have W (G) ≤ n −31n+120 6 6

9

T1 ≇ G′′ , by Lemma 2.6 we have W (T1 ) ≤ W (Tn−2 (n − 7, 3, 1)) = ≤

W (G)

=
Sz(Bn ). Thus, we have (0,⌊ n−4 2 ⌋)

W (Bn

by (3.13) one has

(0,⌊ n−4 2 ⌋−1)

) < W (Bn

(0,s+1) Sz(Bn )

−

(0,s)

) − W (Bn

(0,s) Sz(Bn )

(3.12) (3.13)

) = 2s + 5 − n < 0,

= n − (2s + 5) > 0, i.e.,

) < · · · < W (Bn(0,1) ) < W (Bn(0,0) ) = W (Bn(0) ),

Sz(Bn(0,0) ) = Sz(Bn(0) ) < Sz(Bn(0,1) ) < · · ·
2 ∼ only if G = Hn .

24(n−2) n3 −19n+54

for n ≥ 10. Thus, we have

Sz ∗ (G) W (G)

Sz ∗ (G) W (G)

= 1+

≥ 1+

24(n−2) n3 −19n+54

24(n−2) n3 −19n+54

with equality if

with equality if and

This completes the proof. References [1] M. Aouchiche, G. Caporossi, P. Hansen, M. Laffay, AutoGraphiX: a survey, Electron. Notes Discrete Math. 22 (2005) 515-520. [2] M. Aouchiche, P. Hansen, On a conjecture about the Szeged index, European J. Combin. 31 (2010) 1662-1666. [3] J.A. Bondy, U.S.R. Murty, Graph Theory, in: GTM, vol. 224, Springer, 2008. [4] G. Caporossi, P. Hansen, Variable neighborhood search for extremal graphs 1. The AutoGraphiX system, Discrete Math. 212 (2000) 29-44. [5] L.L. Chen, X.L. Li, M.M. Liu, I. Gutman, On a relation between the Szeged and the Wiener indices of bipartite graphs, Trans. Comb. 1 (4) (2012) 43-49. [6] L.L. Chen, X.L. Li, M.M. Liu, The (revised) Szeged index and the Wiener index of a non-bipartite graph, European J. Combin. 36 (2014) 237-246.

16

[7] D.M. Cvetkovi´c, M. Boob, H. Sachs, Spctral of Graphs. Theory and Application, New York, London, 1979. [8] A. Dobrynin, I. Gutman, On a graph invariant related to the sum of all distances in a graph, Publ. Inst. Math.(Beograd) (N.S.) 56 (1994) 18-22. [9] A. Dobrynin, I. Gutman, Solving a problem connected with distances in graphs, Graph Theory Notes N. Y. 28 (1995) 21-23. [10] A. Dobrynin, R. Entringer and I. Gutman, Wiener index of trees: theory and applications, Acta Appl. Math., 66 (2001) 211-249. [11] H. Deng, The trees on n ≥ 9 vertices with the first to seventeenth greatest Wiener indices are chemical trees, MATCH Commun. Math. Comput. Chem. 57 (2007) 393-402. [12] H. Dong, B. Zhou, Maximum Wiener index of unicyclic graphs with fixed maximum degree, Ars Combin. 103 (2012) 407-416. [13] Z. Du, A. Ilic, On AGC conjectures regarding average eccentricity, MATCH Commun. Math. Comput. Chem. 69 (2013) 579-587. [14] W. Gao, W.F. Wang, The vertex version of weighted Wiener number for bicyclic molecular structures, Comput. Math. Methods Med. 2015, Art. ID 418106, 10 pp. [15] I. Gutman, A formula for the Wiener number of trees and its extension to graphs containing cycles, Graph Theory Notes N. Y. 27 (1994) 9-15. [16] I. Gutman, O.E. Polansky, Mathematical Concepts in Organic Chemistry, Springer, Berlin, 1986. [17] P. Hansen et al. Computers and conjectures in chemical graph theory, in: Plenary Talk in the International Conference on Mathematical Chemistry, August 4-7, 2010, Xiamen, China. [18] H. Hosoya, Topological index. A newly proposed quantity characterizing the topological nature of structural isomers of saturated hydrocarbons, Bull. Chem. Soc. Jpn. 4 (1971) 2332-2339. [19] A. Ilic, Note on PI and Szeged indices, Math. Comput. Modelling 52 (2010) 1570-1576. [20] H. Khodashenas, M.J. Nadjafi-Arani, A.R. Ashrafi, I. Gutman, A new proof of the Szeged-Wiener theorem, Kragujevac J. Math. 35 (2011) 165-172. [21] S. Klavˇzar, M.J. Nadjafi-Arani, Wiener index versus Szeged index in networks, Discrete Appl. Math. 161 (2013) 1150-1153. [22] S. Klavˇzar, M.J. Nadjafi-Arani, Improved bounds on the difference between the Szeged index and the Wiener index of graphs, European J. Combin. 39 (2014) 148-156. [23] X.L. Li, M.M. Liu, Bicyclic graphs with maximal revised Szeged index, Discrete Appl. Math. 161 (2013) 2527-2531. [24] S.C. Li, Y.B. Song, On the sum of all distances in bipartite graphs, Discrete Appl. Math. 169 (2014) 176-185. [25] S.C. Li, H.H. Zhang, Proofs of three conjectures on the quotients of the (revised) Szeged index and the Wiener index and beyond, submitted. [26] M. Liu, B. Liu, On the variable Wiener indices of trees with given maximum degree, Math. Comput. Modelling 52 (2010) 1651-1659. [27] W. Luo, B. Zhou, On ordinary and reverse Wiener indices of non-caterpillars, Math. Comput. Modelling 50 (2009) 188-193. [28] S. Mukwembi, T. Vetr´ık, Wiener index of trees of given order and diameter at most 6, Bull. Austra. Math. Soc. 89 (2014) 379-39. [29] M.J. Nadjafi-Arani, H. Khodashenas, A.R. Ashrafi, On the differences between Szeged and Wiener indices of graphs, Discrete Math. 311 (2011) 2233-2237. [30] M.J. Nadjafi-Arani, H. Khodashenas, A.R. Ashrafi, Graphs whose Szeged and Wiener numbers differ by 4 and 5, Math. Comput. Modelling 55 (2012) 1644-1648. [31] T. Pisanski, M. Randi´c, Use of the Szeged index and the revised Szeged index for measuring network bipartivity, Discrete Appl. Math. 158 (2010) 1936-1944. [32] M. Randi´c, On generalization of Wiener index for cyclic structures, Acta Chim. Slov. 49 (2002) 483-496. [33] S. Simi´c, I. Gutman, V. Balti´c, Some graphs with extremal Szeged index, Math. Slovaca 50 (2000) 1-15. [34] H. Wiener, Structural determination of paraffin boiling points, J. Am. Chem. Soc. 69 (1947) 17-20. [35] R. Xing, B. Zhou, On the revised Szeged index, Discrete Appl. Math. 159 (2011) 69-78. [36] K.X. Xu, K.C. Das, Extremal unicyclic and bicyclic graphs with respect to Harary index, Bull. Malays. Math. Sci. Soc. 36 (2013), 373-383. [37] Z.K. Tang, H.Y. Deng, The (n, n)-graphs with the first three extremal Wiener indices, J. Math. Chem. 43 (2008) 60-74. [38] H.H. Zhang, S.C. Li, L.F. Zhao, On the further relation between the (revised) Szeged index and the Wiener index of graphs, Discrete Appl. Math. 206 (2016) 152-164.

17

[39] X.D. Zhang, Y. Liu, M.X. Han, Maximum Wiener index of trees with given degree sequence, MATCH Commun. Math. Comput. Chem. 64 (2010) 661-682.

18