
The well-studied eccentric connectivity index directly consider the contribution of all edges in a graph. By considering the total eccentricity sum of all non-adjacent vertex, Hua et al. proposed a new topological index, namely, eccentric connectivity coindex of a connected graph. The eccentric connectivity coindex of a connected graph G is defined as
¯ξc(G)=∑uv∉E(G)(εG(u)+εG(v)).
Where εG(u) (resp. εG(v)) is the eccentricity of the vertex u (resp. v). In this paper, some extremal problems on the ¯ξc of graphs with given parameters are considered. We present the sharp lower bounds on ¯ξc for general connecteds graphs. We determine the smallest eccentric connectivity coindex of cacti of given order and cycles. Also, we characterize the graph with minimum and maximum eccentric connectivity coindex among all the trees with given order and diameter. Additionally, we determine the smallest eccentric connectivity coindex of unicyclic graphs with given order and diameter and the corresponding extremal graph is characterized as well.
Citation: Hongzhuan Wang, Xianhao Shi, Ber-Lin Yu. On the eccentric connectivity coindex in graphs[J]. AIMS Mathematics, 2022, 7(1): 651-666. doi: 10.3934/math.2022041
[1] | Jianping Li, Leshi Qiu, Jianbin Zhang . Proof of a conjecture on the ϵ-spectral radius of trees. AIMS Mathematics, 2023, 8(2): 4363-4371. doi: 10.3934/math.2023217 |
[2] | Muhammad Kamran Jamil, Muhammad Imran, Aisha Javed, Roslan Hasni . On the first general Zagreb eccentricity index. AIMS Mathematics, 2021, 6(1): 532-542. doi: 10.3934/math.2021032 |
[3] | Ali N. A. Koam, Ali Ahmad, Azeem Haider, Moin A. Ansari . Computation of eccentric topological indices of zero-divisor graphs based on their edges. AIMS Mathematics, 2022, 7(7): 11509-11518. doi: 10.3934/math.2022641 |
[4] | Fan Wu, Xinhui An, Baoyindureng Wu . Sombor indices of cacti. AIMS Mathematics, 2023, 8(1): 1550-1565. doi: 10.3934/math.2023078 |
[5] | Zenan Du, Lihua You, Hechao Liu, Yufei Huang . The Sombor index and coindex of two-trees. AIMS Mathematics, 2023, 8(8): 18982-18994. doi: 10.3934/math.2023967 |
[6] | Zhen Wang, Kai Zhou . On the maximum atom-bond sum-connectivity index of unicyclic graphs with given diameter. AIMS Mathematics, 2024, 9(8): 22239-22250. doi: 10.3934/math.20241082 |
[7] | Dijian Wang, Dongdong Gao . Laplacian integral signed graphs with few cycles. AIMS Mathematics, 2023, 8(3): 7021-7031. doi: 10.3934/math.2023354 |
[8] | Dalal Awadh Alrowaili, Uzma Ahmad, Saira Hameeed, Muhammad Javaid . Graphs with mixed metric dimension three and related algorithms. AIMS Mathematics, 2023, 8(7): 16708-16723. doi: 10.3934/math.2023854 |
[9] | Chenxu Yang, Meng Ji, Kinkar Chandra Das, Yaping Mao . Extreme graphs on the Sombor indices. AIMS Mathematics, 2022, 7(10): 19126-19146. doi: 10.3934/math.20221050 |
[10] | Yuni Listiana, Liliek Susilowati, Slamin Slamin, Fadekemi Janet Osaye . A central local metric dimension on acyclic and grid graph. AIMS Mathematics, 2023, 8(9): 21298-21311. doi: 10.3934/math.20231085 |
The well-studied eccentric connectivity index directly consider the contribution of all edges in a graph. By considering the total eccentricity sum of all non-adjacent vertex, Hua et al. proposed a new topological index, namely, eccentric connectivity coindex of a connected graph. The eccentric connectivity coindex of a connected graph G is defined as
¯ξc(G)=∑uv∉E(G)(εG(u)+εG(v)).
Where εG(u) (resp. εG(v)) is the eccentricity of the vertex u (resp. v). In this paper, some extremal problems on the ¯ξc of graphs with given parameters are considered. We present the sharp lower bounds on ¯ξc for general connecteds graphs. We determine the smallest eccentric connectivity coindex of cacti of given order and cycles. Also, we characterize the graph with minimum and maximum eccentric connectivity coindex among all the trees with given order and diameter. Additionally, we determine the smallest eccentric connectivity coindex of unicyclic graphs with given order and diameter and the corresponding extremal graph is characterized as well.
Throughout this paper, all graphs considered are finite, simple, undirected and connected. For a graph G=(V,E) with vertex set V=V(G) and edge set E=E(G). The degree of a vertex v∈V(G), denoted by dG(v), is the number of edges incident with v. For vertices u,v∈V(G), the distance d(u,v) is defined as the length of a shortest path between u and v in G. The eccentricity εG(v) or ε(v) of a vertex v is the maximum distance from v to any other vertex in a graph G. The diameter of a connectd graph is the maximum eccentricity of any vertex in the graph. A pendent vertex is a vertex of degree 1. Let Pn, Sn and Cn denote the path, the star and the cycle on n vertices, respectively. By G−v or G∖v we denote the graph obtained from G by deleting a vertex v∈V(G). By G−uv we denote the graph obtained from G by deleting an edge uv∈E(G) (This notation is naturally extended if more than one edge are deleted). Similarly, G+uv is obtained from G by adding an edge uv∉E(G). A path in a connected graph is said to be a diametrical path, if this path is of length equal to the diameter. A connected graph is said to be a tree if it contains no cycles. Connected graphs in which the number of edges equals the number of vertices are called unicyclic graphs. A cactus is a connected graph in which any two simple cycles have at most one vertex in common. The set of cacti with n vertices and k cycles is denoted by C(n,k). If G∈C(n,k), then |E(G)|=n+k−1. Other notation and terminology not defined here will conform to those in [4]. In organic chemistry, a molecular graph represents the topology of a molecule. A topoloical index is a function defined on a molecular graph regardless of the labeling of its vertices. Till now, a number of topological indices are introduced and widely used in QSAR/QSPR studies. One of them is the eccenric connectivity index (ECI) of graph G, denoted by ξc(G), was introduced by Gupta et al. [12], which is defined as
ξc(G)=∑u∈V(G)dG(u)εG(u). |
The eccentric connectivity index has been shown to give a high degree of predictability properties and may provide leads for the development of safe and potent anti-HIV compounds [5,13]. Furthermore, the eccentric connectivity index also has a lot of applications in neural science and entropy, see [14,18]. For the mathematical properties of this index see [1,9,11,17] and the references cited therein.
The eccentric connectivity index of a connected graph G can be rewritten as
ξc(G)=∑uv∈E(G)(εG(u)+εG(v)). |
As is know that the eccentric connectivity index has been used extensively in physical and biological properties. They are defined as sums of contributions dependent on the eccentricity of adjacent vertices over all edges of a graph. By considering analogous contributions from pairs of non-adjacent vertices capturing and quantifying a possible influence of remote pairs of vertices to the molecule's properties, and motivated from [2,3], Hua and Miao [8] considered the total eccentricity sum of non-adjacent vertex pairs which is defined for a connected graph G as
¯ξc(G)=∑uv∉E(G)(εG(u)+εG(v)), | (1) |
and call this eccentricity-based graph invariant the eccentric connectivity coindex ¯ξc(G). By (1), eccentric connectivity coindex can be rewritten as
¯ξc(G)=∑u∈V(G)εG(u)(n−1−dG(u)). | (2) |
The cactus graph has many applications in real life and much works has been done to study the extremal graph according to different index. For more results on the cactus one may be referred to [6,7,10,15,16]. In this paper, we continue the above direction of research by considering the extremal problems on the eccentric connectivity coindex.
This paper is complied as follows. In Section 2, we present the sharp lower bounds on ¯ξc for general connecteds graphs. In Section 3, we characterize the extremal graphs with the minimum ¯ξc among cacti of given order and cycles. In Section 4, we characterize the minimal and maximal ¯ξc of trees with given order and diameter. In Section 5, we study the minimal ¯ξc among unicyclic graphs on n vertices with diameter and characterize the extremal graphs.
Hua (2019) characterize all extremal graphs with the maximum and minimum eccentric connectivity coindex among all connected graphs of given order and establish various lower bounds for this index in terms of several other graph parameters. In this section, we continue the investigation along the lines of [8] and present the sharp lower bounds on ¯ξc for general connected graphs with minimum degree.
Theorem 2.1. Let G (≠Kn) be a connected graph of order n with minimum degree δ. Then
¯ξc(G)≥4(n−1−δ), |
with equality if and only if G≅Kδn (δ<n−1). Where Kδn is a connected graph of order n obtained by joining a vertex to the δ vertices in Kn−1.
Proof. Let vn be one vertex of degree δ. Denote by NG(vn)={v1,v2,⋯,vδ}, also let S={vδ+1,⋯,vn−1}. Since G≠Kn, we have δ≤n−2 and |s|≥1. Thus V(G)=NG(vn)∪Sn∪S. For vi∈NG(vn), we have εG(vi)≥1 and dG(vi)≤n−1. For vi∈S, we have dG(vi)≤n−2 and εG(vi)≥2. Moreover, εG(vn)≥2 and dG(vn)=δ. By the definition of ¯ξc(G), we have
¯ξc(G)=∑v∈NG(vn)εG(v)(n−1−dG(v))+∑v∈SεG(v)(n−1−dG(v))+εG(vn)(n−1−δ)≥0+2(n−1−(n−2))(n−δ−1)+2(n−1−δ)=4(n−1−δ). |
Suppose the equality holds in above equation, then all the inequalities in the above must be equalities. Thus, we have dG(vi)=n−2, εG(vi)=2 for each vi∈S, dG(vi)=n−1, εG(vi)=2 for each vi∈NG(vn) and dG(vn)=δ, εG(vn)=2. Hence, we find that vertex vn is adjacent to the vertices of degree n−1 and each vertex in S is degree n−2. So G≅Kδn (δ<n−1). This completes the proof.
In this section, we turn our attention to eccentric connectivity coindex for cacti and in particular on extremal cacti regarding ¯ξc.
We start with a useful lemma.
Lemma 3.1 ([8]). Let G be a connected graph of order n, size m and diameter d. Then
¯ξc(G)≥2n(n−1)−4m, |
with equality if and only if d≤2.
Let Ckn be the cactus by adding k indepent edges among pendent vertices of Sn (see Figure 1).
Theorem 3.2. Let G be a cactus on n≥5 vertices with k cycles. Then
¯ξc(G)≥2n2−6n−4k+4. |
The equality holds if and only if T≅Ckn.
Proof. Suppose that N is the set of vertices of degree n−1. n0 is the number of elements in N. Assume that n0≥2, let u, v be two vertices in G such that dG(u)=dG(v)=n−1. Then εG(u)=εG(v)=1. It follows that G is not a cactus. Since there exists cycles sharing common edges in G, then n0=0 or n0=1. If n0=1, then there is a unique vertex v in G such that dG(v)=n−1, thus εG(v)=1, hence each vertex in G∖v is adjacent to v. Therefore the cacutus G is obtained by introducing k indepedent edges among pendent vertices of Sn, then G≅Ckn and ¯ξc(G)=2n2−6n−4k+4.
Now, we assume that n0=0. Let d be the diameter of G. Then d≥3. Otherwise, if d≤2, let u be the vertex of maximal degree in G. Then any other vertex of G∖u must be adjacent to u, otherwise d≥3, then n0≥1, a contradiction. By Lemma 3.1,
¯ξc(G)>2n(n−1)−4m=2n(n−1)−4(n+k−1)=2n2−6n+4−4k. |
Note that, there are exactly n+k−1 edges in cacutus on n vertices and k cycles. This completes the proof.
In this section, we shall determine the tree of diameter d with the minimum and maximum ¯ξc respectively.
The volcano graph Vn,d is the graph obtained from a path Pd+1 and a set S of n−d−1 vertices by joining vertex in S to the central vertex of Pd+1. Obviously, if d is even, there is only one center of Pd+1. If d is odd, there are two central vertices of Pd+1 (See Figure 2). The caterpillar tree with respect to Pd+1=u0u1⋯ud, denoted by CP(S1,⋯,Sd−1), is the tree obtained from Pd by attaching Si new vertices to ui, for 1≤i≤d−1.
Let
f1(n,d)={∑d2−1i=1(d−i)(2n−6)+n2d−nd2−nd+2n2−6n−4d+3d2+42if n is even,∑d−12−1i=1(d−i)(2n−6)+n2d−nd2+3n2+3nd−8n+3d2−4d+12if n is odd, |
Theorem 4.1. Let T be a tree on n (n≥5) vertices with diameter d≥2. Then
¯ξc(T)≥f1(n,d). |
The equality holds if and only if T≅Vn,d.
Proof. Let T0 be a graph chosen among all trees of order n with diameter d such that T0 has the smallest ¯ξc. First, we have the following claim.
Claim 1. Among all trees T with order n and diameter d, min(¯ξc(T)) is achived on caterpillars.
Proof of Claim 1. Let T be any tree that is not a caterpillar with order n and diameter d. Let P be the diametral path of T, connecting u0 to ud. Then the eccentricity of each vertex w of T is equal to max{d(w,u0),d(w,ud)}. Let z∉{u0,ud} be a vertex of P and let Tz be a maximal subtree of T which contains z but no other vertex of P. We may assume that z can be selected such that εTz(z)=k≥2, for otherwise T is a caterpillar. Let u be vertex of Tz with d(u,z)=k−1 and let v be the neighbor of u with d(v,z)=k−2. Let S=N(u)∖v and let s=|S|. Note that s≥1. Let T′ be the tree from T by replacing the edges between u and the vertices of S with the edges between v and the vertices of S. Then we have
¯ξc(T)−¯ξc(T′)=∑w∈V(T)εT(w)(n−1−dT(w))−∑w∈V(T′)εT′(w)(n−1−dT′(w))=εT(u)(n−1−dT(u))+εT(v)(n−1−dT(v))+∑w∈SεT(w)(n−1−dT(w))−εT′(u)(n−1−dT′(u))−εT′(v)(n−1−dT′(v))−∑w∈SεT′(w)(n−1−dT′(w))≥s(εT(v)−εT(u))+∑w∈S(n−1−dT(w)). |
Note that εT(u)=εT(v)+1, εT(w)≥εT′(w)+1 for w∈S and since d≥2, we get that dT(w)<n−2. Hence
¯ξc(T)−¯ξc(T′)≥−s+(n−1)s−∑w∈SdT(w)=(n−2)s−∑w∈SdT(w)>(n−2)s−(n−2)s=0. |
Therefore,
¯ξc(T)>¯ξc(T′). |
If T′ is not a caterpillar, we can repeat the construction as many times as required to arrive at a caterpillar. Since at each step the value of ¯ξc(T) is decreased. Thus the claim is proven.
Since T0 is the extremal tree with diametral path P=u0u1,⋯,ud. By claim 1, we conclude that all vertices of V(T)∖V(P) must be pendent vertices attached at some vertices of P, we denote this tree T1. We now consider the case when d is even, if there exists some vertex ui (i≠d2) of P with pendent vertices (say w1,w2⋯,wtt≥1) attached. Let
T2=T1−{uiw1,uiw2,⋯,uiwt}+{ud2w1,ud2w2,⋯,ud2wt}. |
By the definition of ¯ξc, we have
¯ξc(T1)−¯ξc(T2)=t∑k=1εT1(wk)(n−1−dT1(wk))+εT1(ui)(n−1−dT1(ui))+εT1(ud2)(n−1−dT1(ud2))−t∑k=1εT2(wk)(n−1−dT2(wk))−εT2(ui)(n−1−dT2(ui))−εT2(ud2)(n−1−dT2(ud2)). |
As dT1(wk)=dT2(wk)=1, (k=1,2,⋯,t) and εT1(wk)>εT2(wk), εT1(ui)=εT2(ui)>εT1(ud2)=εT2(ud2),εT1(wk)−εT2(wk)=εT1(ui)−εT2(ud2).
Then, we have
¯ξc(T1)−¯ξc(T2)=tεT1(wk)(n−2)−tεT2(wk)(n−2)−tεT1(ui)+tεT1(ud2)≥t(n−2)(εT1(wk)−εT2(wk))−t(εT1(ui)−εT2(ud2))=[(n−2)t−t](εT1(wk)−εT2(wk))=(n−3)t(εT1(wk)−εT2(wk))>0. |
Therefore,
¯ξc(T1)>¯ξc(T2). |
Continue this procedure, forming new trees, untill all the pendent vertices in T1 are adjacent to ud2. If d is odd, similarly, the extremal graph must be the graph obtained from a path Pd+1 by some pendent vertices attched on the center of Pd+1. That is to say, T0≅Vn,d. This completes the proof.
For even d,
¯ξc(Vn,d)=d2−1∑i=1(d−i)(2n−6)+n2d−nd2−nd+2n2−6n−4d+3d2+42=n(n−2)(d+2)2−(n−2)(d2+3d+2)2+d(d−2)(3n−7)4+2d(n−2). |
Let f(x)=n(n−2)(x+2)2−(n−2)(x2+3x+2)2+x(x−2)(3n−7)4+2x(n−2), (x≥2). It remains to determine which value of x minimizes f(x). For this, we use the first and second derivative test. Noting that
f′(x)=n(n−2)2−(n−2)(2x+3)2+(x−1)(3n−7)2+2(n−2). |
f″(x)=−(n−2)+3n−72=n−32>0. |
f″(x) is positive for x≥2, then f′(x) is an increasing function for x≥2. So
f′(x)>f′(2)=n(n−2)2−7(n−2)2+(3n−7)2+2(n−2)=n2−2n−12=(n−1)2−22>0. |
Therefore, f(x) is an increasing function for x≥2.
Then
¯ξc(Vn,d)<¯ξc(Vn,d+1). |
For odd d, we obtain the same result. Therefore, we obtain the following chain inequality
¯ξc(Pn)=¯ξc(Vn,n−1)>¯ξc(Vn,n−2)>⋯>¯ξc(Vn,2)=¯ξc(Sn). | (3) |
We denote by H(p,n,q) one double starlike tree which is obtained by attaching the centers of two stars K1,p and K1,q to the ends of path Pd−2, respectively, where p+q=n−d−1. The broom graph Bn,d consists of a path Pd−1, together with n−d pendent vertices all adjacent to the same pendent vertex of Pd−1, obviously, Bn,d=H(0,n,n−d−1), see Figure 3.
For any tree T∈H(p,n,q), we have ¯ξc(T)=¯ξc(Bn,d)=f2(n,d). Let
f2(n,d)={∑d2−1i=1(d−i)(2n−6)+6d2−3nd+2n2d−2nd2+2n−2−7d2if n is even,∑d−12i=1(d−i)(2n−6)+3d2−2nd+n2d−nd2+n−1−2dif n is odd, |
Theorem 4.2. If T is a tree of order n and diameter d, then
¯ξc(T)≤f2(n,d). |
The equality holds if and only if T≅H(p,n,q).
Proof. Let P=u0u1,⋯,ud be a diametral path in T. Asssume that T is not the graph H(p,n,q), then there exists a pendent vertex v of T,v≠u0 such that v is adjacent to a vertex u, where u≠ud−1 and u≠u1 (It is possible that u lies on P). Denote by {v1,v2,⋯,vk} be the set of pendent vertices which are adjacent to u and vi≠u0 for i=1,2,⋯,k. Let
T′=T−{uv1,uv2,⋯,uvk}+{ud−1v1,ud−1v2,⋯,ud−1vk}. |
Note that T′ has the same order and diameter as T. We will show that T′ has a larger eccentric connectivity coindex than T.
¯ξc(T′)−¯ξc(T)=k∑i=1εT′(vi)(n−1−dT′(vi))+εT′(u)(n−1−dT′(u))+εT′(ud−1)(n−1−dT′(ud−1))−k∑i=1εT(vi)(n−1−dT(vi))−εT(u)(n−1−dT(u))−εT(ud−1)(n−1−dT(ud−1)). |
As dT′(vi)=dT(vi)=1, (i=1,2,⋯,k) and εT′(vi)>εT(vi), εT′(u)=εT(u)<εT′(ud−1)=εT(ud−1),εT′(vi)−εT(vi)=εT′(ud−1)−εT(u). Then, we have
¯ξc(T′)−¯ξc(T)=k(n−2)(εT′(vi)−εT(vi))−k(εT′(ud−1)−εT(u))=k(n−2)(εT′(vi)−εT(vi))−k(εT′(vi)−εT(vi))=k(n−3)(εT′(vi)−εT(vi))>0. |
Then
¯ξc(T′)>¯ξc(T). |
Continue this procedure, forming new trees untill all vertices outside P having degree one are adjacent to xd−1. Thus a tree H(p,n,q) of order n and diameter d is obtained, one has
¯ξc(Bn,d)=¯ξc(H(P,n,q))≥¯ξc(T). |
This completes the proof.
Similar to previous discussion, we have ¯ξc(Bn,d)<¯ξc(Bn,d+1) for any d≥2. Therefore, it follows that
¯ξc(Pn)=¯ξc(Bn,n−1)>¯ξc(Bn,n−2)>⋯>¯ξc(Bn,2)=¯ξc(Sn). | (4) |
Summarizing theorems 4.1, 4.2 and these inequlities (3), (4), we have the following result.
Theorem 4.3. Let T be a tree on n vertices. Then
2n2−6n+4≤¯ξc(T)≤f2(n,n−1). |
Where f2(n,n−1) as mentioned in theorem 4.2. The left quality holds if and only if T≅Sn and the right equality holds if and only if T≅Pn.
In this section, we consider the minimum ¯ξc of unicyclic graphs with given diameter. Let Gdn be the unicyclic graph with order n and diameter d. First, for even d, let V1n,d be the graph obtained from Pd+1=u0u1.⋯,ud by attaching n−d−1 pendent edges to ud2 and adding an edge between ud2+1 and one of the attached pendent vertices of ud2. Let V2n,d be the graph obtained from Pd+1=u0u1.⋯,ud by attaching n−d−1 pendent edges to ud2 and adding an edge between two attached pendent vertices of ud2, (see Figure 4).
For odd d, let V3n,d be the unicyclic graphs in which there are s,t (s+t=n−d−2) pendent vertices adjacent to ud−12 and ud+12 of diametrel path respectively. Let V4n,d be the unicyclic graphs in which there are p,q (p+q=n−d−3) pendent vertices adjacent to ud−12 and ud+12 of diametrel path respectively, (see Figure 4).
By direct calculation, for even d,
¯ξc(V1n,d)=¯ξc(V2n,d)=d2−1∑i=1(d−i)(2n−6)+2d(n−2)+d(d−2)2+(d+2)((n−2)(n−d)−n)2. |
For odd d, any G∈V3n,d, we have
¯ξc(G)=d−12−1∑i=1(d−i)(2n−6)+2d(n−2)+(d+1)(2n+d−9)2+(n−2)(d+3)(n−d−2)2. |
If G1∈V3n,d and G2∈V4n,d. By direct calculation we find that, ¯ξc(G1)<¯ξc(G2). Let
f3(n,d)={∑d2−1i=1(d−i)(2n−6)+3d2−6d+n2d−nd2+2n2−6n−nd2if n is even,∑d−12−1i=1(d−i)(2n−6)+3d2−nd−6d−10n+n2d−nd2+3n2+32if n is odd. |
Theorem 5.1. Let G be a unicyclic graph on n(≥7) vertices with diameter d≥2. Then
¯ξc(G)≥f3(n,d). |
The equality holds if and only if G≅V1n,d or G≅V2n,d for even d and G≅V3n,d for odd d.
Proof. Choose G0 in Gdn such that ¯ξc(G0) is as small as possible. Let Pd+1=u0u1,⋯,ud be a diametral path and Ck be the unique cycle in G0. Similar as the proof of theorem 4.1 all vertices in V(G0)∖{V(P)∪V(Ck)} must be pendent vertices and adjacent to some vertices of V(P)∪V(Ck), we denote it by G∗. First, we consider when the diameter d is even, we proceed by considering the following possible cases.
Case 1. |V(Ck)∩V(Pd+1)|=1.
In this case, let V(Ck)∩V(Pd+1)=ui. In the following, we show three facts.
Fact 1. All vertices in V(G∗)∖{V(P)∪V(Ck)} must be adjacent to ud2 of P.
Proof of Fact 1. If there exists a vertex us (with s≠d2) of P with pendent vertices, say w1,w2,⋯,wk attached in G∗. Let
G1=G∗−{usw1,usw2,⋯,uswk}+{ud2w1,ud2w2,⋯,ud2wk}. |
By a similar approach in the proof of theorem 4.1, we get ¯ξc(G∗)>¯ξc(G1), a contradiction to our choice of G∗. Similarly, we conclude that there is no pendent vertices attached the vertex of cycle Ck in G∗ other than ui. That is to say, each of the vertices other ui on cycle Ck in G∗ is of degree 2. This completes the proof of fact 1.
Fact 2. The length of the cycle Ck is equal to 3, i.e., k=3 in G∗.
Proof of Fact 2. If the length of the cycle k≠3.
Let Ck=uiv1v2,⋯,vk−1ui and NG∗(ui)∩V(Ck)={v1,vk−1}. Let
G2=G∗−E(Ck)+v1vk−1+{uiv1,uiv2,⋯,uivk−1}. |
Clearly, G2 is in Gdn and C3 is the unique cycle in G2. It is routine to check that
¯ξc(G∗)−¯ξc(G2)=k−2∑j=2εG∗(vj)(n−1−dG∗(vj))+εG∗(ui)(n−1−dG∗(ui))+εG∗(v1)(n−1−dG∗(v1))+εG∗(vk−1)(n−1−dG∗(vk−1))−k−2∑j=2εG2(vj)(n−1−dG2(vj))−εG2(ui)(n−1−dG2(ui))−εG2(v1)(n−1−dG2(v1))−εG2(vk−1)(n−1−dG2(vk−1)). |
Note that εG∗(ui)=εG2(ui),εG∗(v1)=εG2(v1), εG∗(vk−1)=εG2(vk−1) and εG∗(vj)≥εG2(vj)>εG∗(ui) for j=2,⋯,k−2 and εG∗(v2)=εG∗(ui)+2, εG2(v2)=εG2(ui)+1.
Therefore,
¯ξc(G∗)−¯ξc(G2)>(k−3)εG∗(ui)+k−2∑j=2εG∗(v2)(n−3)−k−2∑j=2εG2(v2)(n−2)>(k−3)εG∗(ui)+(k−3)(n−3)εG∗(v2)−(k−3)(n−2)εG2(v2)=(k−3)εG∗(ui)+(k−3)(n−3)(εG∗(ui)+2)−(k−3)(n−2)(εG∗(ui)+1)=(k−3)εG∗(ui)+(k−3)(−εG∗(ui)+n−4)=(k−3)(n−4)>0. |
Then ¯ξc(G∗)>¯ξc(G2), a contradiction. This completes the proof of fact 2.
Fact 3. ui is the center of the diametral path P in G∗.
Proof of Fact 3. Based on fact 2, we know that Pd+1=u0u1,⋯,ud and Ck=uiv1vk−1. If ui≠ud2, we assume without loss of generality that dG∗(u0,ui)<d2, that is to say, dG∗(u0,ui)<dG∗(ui+1,ud). Move the triangle and all the pendent edges from ui to ud2 in G∗ and denote the result graph by G3. It is routine to check that G3 in Gdn. Since
εG∗(ui)−εG∗(ud2)=εG∗(v1)−εG3(v1)=εG∗(vk−1)−εG3(vk−1)=εG∗(v2)−εG3(v2). |
We have
¯ξc(G∗)−¯ξc(G3)=εG∗(ui)(n−1−dG∗(ui))+εG∗(ud2)(n−1−dG∗(ud2))+k−1∑i=1εG∗(vi)(n−1−dG∗(vi))−εG3(ui)(n−1−dG3(ui))−εG3(ud2)(n−1−dG3(ud2))−k−1∑i=1εG3(vi)(n−1−dG3(vi))>(k−1)(εG∗(ud2)−εG∗(ui))+(n−3)εG∗(v1)+(n−3)εG∗(vk−1)+(k−3)(n−2)εG∗(v2)−(n−3)εG3(v1)−(n−3)εG3(vk−1)−(k−3)(n−2)εG3(v2)=[−(k−1)+2(n−3)+(n−2)(k−3)](εG∗(ui)−εG∗(ud2))>(k−1)(n−4)(εG∗(ui)−εG∗(ud2))>0. |
Then ¯ξc(G∗)>¯ξc(G3), a contradiction again. This completes the proof of fact 3.
Case 2. P and Ck are vertex and edge disjoint.
Let Q=uiz1z2,⋯,zs−1zs be a path connecting path P and Ck. By the similar approach in the proof of theorem 4.1, we can contract the whole path Q (i.e., ui and zs coincide and then attaching suitable number of pendent vertices at ui) to get a new graph G4 with P and Ck having exactly one vertex in common. So the following procedure similar as case 1.
By the discussion as above, we obtain that for all G in Gdn. If the unique cycle and the diameter in G have no edges in common, then the extremal graph is V2n,d when d is even. When d is even, the diameter path has only one center, while when d is odd, the diameter path has two centers. In either case, the pendent vertices in the minimum graph except the endpoints of the diameter path are all adjacent to the center of the diameter path, the proof process is exactly the same. Here we omit the proof when d is odd. Hence, the extremal graph is V4n,d when d is odd. As is depicted in Figure 4.
Case 3. |V(Ck)∩V(Pd+1)|≥2.
If there exist common edges between Pd+1 and Ck, then we get that all vertices in V(G0)∖{V(P)∪V(Ck)} must be adjacent to ud2 of P similar as case 1. We denote it by G∗. Let Pd+1=u0u1,⋯,ud and Ck=uiui+1,⋯,ujy1y2,⋯,ymui. In this case, we first show that the unique cycle Ck contained in G∗ is just C3, that is to say, k=3, otherwise, we assume that k≥4. First we consider the diameter d is even. Let
G5=G∗−{uiym,ymym−1,⋯,y2y1,y1uj}+{ud2ym+ud2ym−1,⋯,ud2y2,ud2y1}+{ud2+1y1}. |
By the definition of ¯ξc and bearing in mind that it is possible uj−1=ui. As εG∗(yi)≥εG5(yi), i=1,⋯,m and εG5(y1)=εG5(y2)=εG5(y3)=⋯=εG5(ym). εG5(ui)=εG∗(ui) and εG5(uj+1)=εG∗(uj+1), εG5(yi)≥εG5(ud2)+1, εG∗(yi)≥εG∗(ud2)+2, εG5(ud2)=εG∗(ud2), εG5(ud2+1)=εG∗(ud2+1),
Moreover, dG∗(yi)=2, dG5(yi)=1, i=2,⋯,mdG∗(y1)=2, dG5(y1)=2 and dG5(ud2)=dG∗(ud2)+m, dG5(ud2+1)=dG∗(ud2+1)+1
It follows that
¯ξc(G∗)−¯ξc(G5)=m∑i=2εG∗(yi)(n−3)+εG∗(y1)(n−3)+εG∗(ui)(n−4)+εG∗(uj)(n−4)+εG∗(ud2)(n−1−dG∗(ud2))+εG∗(ud2+1)(n−1−dG∗(ud2+1))−m∑i=2εG5(yi)(n−2)−εG5(y1)(n−3)−εG5(ui)(n−3)−εG5(uj)(n−3)−εG5(ud2)(n−1−dG5(ud2))−εG5(ud2+1)(n−1−dG5(ud2+1))=m∑i=2[εG∗(yi)(n−3)−εG5(yi)(n−2)]−εG∗(ui)−εG∗(uj)+mεG∗(ud2)+εG∗(ud2+1)>(m−1)[(εG∗(ud2)+2)(n−3)−(n−2)(εG∗(ud2)+1)]−εG∗(ui)−εG∗(uj)+(m−1)εG∗(ud2)+εG∗(ud2)+εG∗(ud2+1)>(m−1)(n−4−εG∗(ud2))+d−εG∗(ui)−εG∗(uj)+(m−1)εG∗(ud2)=(m−1)(n−4)+d−(εG∗(ui)+εG∗(uj))>(m−1)(n−4)+d−2d=(m−1)(n−4)−d>(m−1)(n−4)−(n−2)>2(n−4)−(n−2)=n−6>0. |
Therefore,
¯ξc(G∗)>¯ξc(G5), |
which contradicts the chioce of G∗.
Hence, the structure of G∗ can be described as follows: its unique cycle C3 and its diametral path Pd+1=u0u1,⋯,ud have only one edge ud2ud2+1 in common and there are some pendent edges attached to ud2 in G∗.
Summarizing the discussion as in case 3, we obtain that for all G in Gdn, if the unique cycle and the diameter in G have edge in common. When d is even, then the graph say V1n,d with the minimum ¯ξc and when d is odd, we can similarly get that the extremal graph belongs to V3n,d.
Summarizing cases 1–3, when d is odd, for any G1∈V3n,d and G2∈V4n,d, we can easily obtain that ¯ξc(G1)<¯ξc(G2). Therefore, V3n,d achieved the minimum ¯ξc for d is odd.
When d is even, ¯ξc(V1n,d)=¯ξc(V2n,d), V1n,d and V2n,d obtain the minimum value of Eccentric connectivity coindex at the same time (See Figure 4).
This completes the proof.
In this paper, we first present the sharp lower bounds on ¯ξc for general connecteds graphs and present a structure of the extremal graphs for eccentric connectivity coindex over cacti graphs with n vertices and k cycles, then characterize the extremal trees with given order and diameter on the eccentric connectivity coindex. Moreover, we optimize the extremal structure of unicyclic graphs with given order and diameter. Along this line, some other interesting extremal problems on the eccentric connectivity coindex are valuable to be considered.
H. Z. Wang is the corresponding author and was partially supported by the National Nature Science Foundation of China (No.11971011) and Natural Science Foundation of Jiangsu Province of China (No. BK20191047). The authors would like to thank the anonymous referees for their careful reading and helpful suggestions that improve the presentation of the paper.
All authors in the paper have no conflict of interest.
[1] |
A. R. Ashrafi, M. Saheli, M. Ghorbani, The eccentric connectivity index of nanotubes and nanotori, J. Comput. Appl. Math., 235 (2011), 4561–4566. doi: 10.1016/j.cam.2010.03.001. doi: 10.1016/j.cam.2010.03.001
![]() |
[2] |
A. R. Ashrafi, T. Došlić, A. Hamzeh, The zagreb coindices of graph operations, Discrete Appl. Math., 158 (2010), 1571–1578. doi: 10.1016/j.dam.2010.05.0171. doi: 10.1016/j.dam.2010.05.0171
![]() |
[3] |
A. R. Ashrafi, T. Došlić, A. Hamzeh, Extremal graphs with respect to the zagreb coindices, MATCH Commun. Math. Comput. Chem., 65 (2011), 85–92. doi: 10.1136/jamia.2010.009928. doi: 10.1136/jamia.2010.009928
![]() |
[4] | J. A. Bondy, U. S. R. Murty, Graph theory with applications, 2 Eds., New York: Macmillan London and Elsevier, 1976. doi: 10.1137/1021086. |
[5] |
H. Dureja, S. Gupta, A. K. Madan, Predicting anti-HIV-1 activity of 6-arylbenzonitriles: Computational approache using superaugmented eccentric connectivity topochemical indices, J. Mol. Graph. Model., 26 (2008), 1020–1029. doi: 10.1016/j.jmgm.2007.08.008. doi: 10.1016/j.jmgm.2007.08.008
![]() |
[6] |
J. Du, G. Su, J. Tu, I. Gutman, The degree resistance distance of cacti, Discrete Appl. Math., 188 (2015), 16–24. doi: 10.1016/j.dam.2015.02.022. doi: 10.1016/j.dam.2015.02.022
![]() |
[7] |
F. He, Z. Zhu, Cacti with maximum eccentricity resistance-distance sum, Discrete Appl. Math., 219 (2017), 117–125. doi: 10.1016/j.dam.2016.10.032. doi: 10.1016/j.dam.2016.10.032
![]() |
[8] |
H. Hua, Z. Miao, The total eccentricity sum of non-adjacent vertex pairs in graphs, Bull. Malays. Math. Sci. Soc., 42 (2019), 947–963. doi: 10.1007/s40840-017-0528-2. doi: 10.1007/s40840-017-0528-2
![]() |
[9] |
A. Ilić, I. Gutman, Eccentric connectivity index of chemical trees, MATCH Commun. Math. Comput. Chem., 65 (2011), 731–744. doi: 10.1080/07366299.2011.581101. doi: 10.1080/07366299.2011.581101
![]() |
[10] |
J. B. Liu, W. R. Wang, Y. M. Zhang, X. F. Pan, On degree resistance distance of cacti, Discrete Appl. Math., 203 (2016), 217–225. doi: 10.1016/j.dam.2015.09.006. doi: 10.1016/j.dam.2015.09.006
![]() |
[11] |
M. J. Morgan, S. Mukwembi, H. C. Swart, On the eccentric conectivity index of a graph, Discrete Math., 311 (2011), 1229–1234. doi: 10.1016/j.disc.2009.12.013. doi: 10.1016/j.disc.2009.12.013
![]() |
[12] |
V. Sharma, R. Goswami, A. K. Madan, Eccentric connectivity index: A novel highly discriminating topological descriptor for structure-property and structure-activity studies, J. Chem. Inf. Model., 37 (1997), 273–282. doi: 10.1002/chin.199727028. doi: 10.1002/chin.199727028
![]() |
[13] |
S. Sardana, A. K. Madan, Predicting anti-HIV activity of TIBO derivatives: a computational approach using a novel topological descriptor, J. Mol. Model., 8 (2002), 258–265. doi: 10.1007/s00894-002-0093-x. doi: 10.1007/s00894-002-0093-x
![]() |
[14] |
S. Wang, X. Yang, Y. Zhang, P. Phillips, J. Yang, T. Yuan, Identification of green, Oolong and black teas in China via wavelet packet entropy and fuzzy support vector machine, Entropy, 17 (2015), 6663–6682. doi: 10.3390/e17106663. doi: 10.3390/e17106663
![]() |
[15] |
H. Wang, H. Hua, D. Wang, Cacti with minimum, second-minimum and third-minimum kirchhoff indices, Math. Commun., 15 (2010), 347–358. doi: 10.1016/j.mcm.2010.06.018. doi: 10.1016/j.mcm.2010.06.018
![]() |
[16] |
H. Wang, L. Kang, More on the Harary index of cacti, J. Appl. Math. Comput., 43 (2013), 369–386. doi: 10.1007/s12190-013-0668-y. doi: 10.1007/s12190-013-0668-y
![]() |
[17] |
G. Yu, L. Feng, On connective eccentricity index of graphs, MATCH Commun. Math. Comput. Chem., 69 (2013), 611–628. doi: 10.3233/ICA-130440. doi: 10.3233/ICA-130440
![]() |
[18] |
X. Zhou, Y. Zhang, G. Ji, J. Yang, Z. Dong, S. Wang, et al., Detection of abnormal MR brains based on wavelet entropy and feature selection, IEEJ T. Electri. Electr. Eng., 11 (2016), 364–373. doi: 10.1002/tee.22226. doi: 10.1002/tee.22226
![]() |