• 検索結果がありません。

The Number Of Spanning Trees And Chains Of Graphs ∗

N/A
N/A
Protected

Academic year: 2022

シェア "The Number Of Spanning Trees And Chains Of Graphs ∗ "

Copied!
7
0
0

読み込み中.... (全文を見る)

全文

(1)

The Number Of Spanning Trees And Chains Of Graphs

Sheng Ning Qiao

, Bing Chen

Received 3 September 2007

Abstract

Let t(G) denote the number of spanning trees of a graphG. Achainof two connected vertices u, v(dG(u), dG(v) ≥3) in G, denoted byLk, is defined as a path ofGanddG(p) = 2 for allp∈V(Lk)− {u, v}, wherekis the length of the path. In this paper, we investigate the relationship between t(G) and Lk of a graphG. In particular, the relationship betweent(G) andLkofτ-optimal graph Gis considered.

1 Introduction

We use Bondy and Murty [2] for terminology and notations not defined here and consider finite connected graphs only. Aspanning subgraph of a graphG= (V, E) is a subgraph with vertex setV. Aspanning treeis a spanning subgraph that is a tree. Let Γ(n, m) denote the collection of allnverticesmedges graphs with no loops. Lett(G) denote the number of spanning trees of a graph G. Spanning trees have been found to be structures of paramount importance in both theoretical and practical problems.

As a result the number of spanning trees of a connected graph has been the focus for extensive attention in graph theoretical research.

A graph G∈Γ(n, m) is called τ-optimal if t(G)≥t(H) for allH ∈ Γ(n, m). An open extremal problem, with applications to the synthesis of reliable networks, is the characterization ofτ-optimal graphs [1, 3, 4, 5, 6, 7]. In [5], authors introduced a lower bound for the trace of the k-th power of the Laplacian matrix of a graph in terms of its degree sequence. Using this inequality they developed an upper bound for the number of spanning trees of a graph in terms of the degree sequence of its completment that is sharp for, and only for, complete multipartite graphs. In [6], authors develop a powerful refinement of the upper bounding technique for the number of spanning trees.

The improved bound yields a new technique to characterize many hitherto unknown types ofτ-optimal graphs.

Mathematics Subject Classifications: 05C05, 05C85.

Department of Applied Mathematics, Northwestern Polytechnical University, Xi’an, Shaanxi 710072, P. R. China

Department of Aplied Mathematics, Xi’an University of Technology, Xi’an, Shaanxi 710048, P. R.

China

10

(2)

We consider the reliability of graphs for which edges fail independently of each other with a constant probabilityq. A standard formula for the reliability of a graphGis

R(G, q) = Xm

i=n1

Ni(G)qmi(1−q)i,

where Ni(G) denotes the number of connected spanning subgraphs ofGwith iedges.

ClearlyNn1(G) =t(G). Suppose thatG, H∈Γ(n, m). We have R(G, q)−R(H, q) = qnm1(1−q)1n

×

"

t(G)−t(H) + Xm

i=n

(Ni(G)−Ni(H))qni1(1−q)in+1

# .

Ift(G)> t(H), thenR(G, q)> R(H, q) forq→1. Thusτ-optimal graphs are uniformly most reliable in Γ(n, m) forq→1.

In this paper, we investigate the relationship between the number of spanning trees and chains of a graph. In particular, the relationship between the number of spanning trees and chains ofτ-optimal graphs is considered.

2 Number of Spanning Trees and Chains of Graphs

Achainof two connected verticesu, v(dG(u), dG(v)≥3) inG, denoted byLk, is defined as a path ofGanddG(p) = 2 for allp∈V(Lk)−{u, v}, wherekis the length of the path.

Ifk= 1, thenL1 is trivial, i.e., an edge. Two chainsLk1, Lk2 are said to be parallel if Lk1, Lk2 meet only in two common endpoints. LetG−Lk=G[V(G)−V(Lk) +{u, v}]

and G/Lk= ((G−Lk) +uv)/uv, whereu, v are two endpoints ofLk.

THEOREM 1. LetLk(k≥1) be a chain of a graphG. Thent(G−Lk)≤t(G) and t(G/Lk)≤t(G).

PROOF. We provet(G) =kt(G−Lk) +t(G/Lk) first. Letuandv be end vertices ofLk andG=G−Lk+uv. Then

t(G) =t(G−uv) +t(G/uv).

Since every spanning tree of G that does not contain uv yields k spanning trees of G, each of which does not contain Lk, and conversely, kt(G−Lk) is the number of spanning trees ofGthat does not contain Lk.

Now to each spanning treeT ofGthat containsuv, there corresponds a spanning tree T /Lk of G/Lk. This correspondence is clearly a bijection. Therefore t(G/Lk) is precisely the number of spanning trees ofGthat containLk. It follows that

t(G) =kt(G−Lk) +t(G/Lk).

Since t(G−Lk) ≥ 0 and t(G/Lk) ≥ 0, it is easy to have t(G−Lk) ≤ t(G) and t(G/Lk)≤t(G).

THEOREM 2. LetLk(k >3) be a chain of a graphGand u, vare two endpoints of Lk. Suppose that Lk does not contain and cut edges of Gand w∈V(G)−V(Lk)

(3)

withwu, wv6∈E(G). We construct two chainsLk1(k1=bk/2c), Lk2(k2=k−k1), such that w, uandw, v are two endpoints ofLk1, Lk2, respectively. Then we have

t(G)< t(G−Lk+{Lk1, Lk2}).

PROOF. Let G = G−Lk +{Lk1, Lk2}. By the proof of Theorem 1, we have t(G) =kt(G−Lk) +t(G/Lk). Similarly, we have

t(G) = t(G−Lk1−Lk2)k1k2+k1t((G/Lk2)−Lk1) +k2t((G/Lk1)−Lk2) +t(G/Lk1/Lk2).

Since k1 =bk/2c, k2= k−k1 andk >3, we have k1k2 ≥k. Let Ge = G−Lk+uv andG=G−Lk1−Lk2+uw. LetT be a spanning tree ofGewhich containsuv, then T−uv+uw is a spanning tree ofGwhich does not containvw, which implies

t(G/Lk) =t((G/Lk1)−Lk2).

Combined with t(G−Lk) =t(G−Lk1−Lk2), t(G/Lk1/Lk2)>0 andt((G/Lk2)− Lk1)>0,we havet(G)< t(G).

LetGe=G−Lk+uv,G1=G−Lk1−Lk2+uwandG2=G−Lk1−Lk2+vw. Let T be a spanning tree ofGwhich containsuv, then one of the following results holds:

(1)T−uv+uwis a spanning tree ofG1 which does not containvw, which implies t(G/Lk) =t((G/Lk1)−Lk2);

(2)T−uv+vwis a spanning tree ofG2 which does not containuw, which implies t(G/Lk) =t((G/Lk2)−Lk1).

Combined witht(G−Lk) =t(G−Lk1−Lk2),t((G/Lk1)−Lk2)>0,t((G/Lk2)−

Lk1)>0 and t(G/Lk1/Lk2)>0, we havet(G)< t(G).

3 Number Of Spanning Trees and Chains of τ -Optimal Graphs

We have the following result.

THEOREM 3. LetGbe aτ−optimal graph andLk1(k1>0), Lk2(k2>0) are two chains of G. Then

(a)t((G−Lk1)/Lk2)≤t(G−Lk2),and (b)t((G−Lk1)/Lk2)≤t(G/Lk1).

PROOF of (a). We prove by contradiction. LetGbe aτ-optimal graph, and assume that there are two chains Lk1 andLk2 ofGwith

t((G−Lk1)/Lk2)> t(G−Lk2).

Letuandvbe end vertices ofLk1. We construct a new graphG from (G−Lk1)/Lk2

by adding a chainLk in (G−Lk1)/Lk2, withu, vas end vertices andk=k1+k2. Then we have |V(G)| =|V(G)|and |E(G)| =|E(G)|. Since Gis a τ-optimal graph, we

(4)

havet(G)≥t(G). Sincek=k1+k2 > k2, we may select a chainLp from Lk inG with p=k2, starting fromu, and so

t(G) =t(G−Lp)p+t(G/Lp).

Note that k −p = k1, which implies that G/Lp = G/Lk2 and (G −Lp)/Lq = (G−Lk1)/Lk2, whereLq =Lk−Lp. By Theorem 1, we have

t((G−Lp)/Lq)≤t(G−Lp), so

t(G−Lp)≥t((G−Lk1)/Lk2).

Therefore, we have

t(G) ≥ t(G)

= t(G−Lp)p+t(G/Lp)

≥ t((G−Lk1)/Lk2)p+t(G/Lk2)

> t(G−Lk2)p+t(G/Lk2)

= t(G), a contradiction.

PROOF of (b). We prove by contradiction. Let G be a τ-optimal graph, and assume that there are two chainsLk1 andLk2 ofGwith

t(G−Lk1)> t(G/Lk1).

Let uand v be end vertices ofLk2. We construct a new graph G from G−Lk1 by adding a chainLk in (G−Lk1)/Lk2, withu, vas end vertices andk=k1. Then we have

|V(G)|=|V(G)|and|E(G)|=|E(G)|. SinceGis aτ-optimal graph, we havet(G)≥ t(G) and t(G) =t(G−Lk)k+t(G/Lk). Note thatG/Lk/Lk2 = (G−Lk1)/Lk2

and G−Lk =G−Lk1. By Theorem 1, we have t(G/Lk/Lk2)≤t(G/Lk), so

t(G/Lk)≥t((G−Lk1)/Lk2).

Therefore, we have

t(G) ≥ t(G)

= t(G−Lk)k+t(G/Lk)

≥ t(G−Lk1)k1+t((G−Lk1)/Lk2)

> t(G−Lk1)k1+t(G/Lk1)

= t(G), a contradiction.

(5)

The following results may be useful.

LEMMA 1. [1] If 3≤n≤e, thenτ-optimal graphs in Γ(n, e) are two connected.

LEMMA 2. [1] LetGbe aτ−optimal graph and 6≤n+ 2≤e. If there exit two parallel chainsLk1, Lk2 inG, thenk1=k2= 1.

LEMMA 3. [4] LetGbe a connected graph and u, v∈V(G), dG(u) = dG(v) = 2.

If u6∈NG(v), then

t(G)≤t(G/{u, v})

and the equality holds if and only if NG(u) =NG(v), whereG/{u, v}= (G+uv)/uv.

LEMMA 4. [1] Let G be a τ− optimal graph. If there exit two parallel chains Lk1, Lk2 inG, then|k1−k2| ≤1.

LEMMA 5. Ifεis an edge ofG, thent(G) =t(G−ε) +t(G/ε).

THEOREM 4. If 6≤n+ 2< e,1< k <3n−2e+ 2, then bt(n, e)>3bt(n−k+ 1, e−k),

where n, e, kare positive integer numbers and ˆt(n, e) denotes the number of spanning trees ofτ-optimal graphs in Γ(n, e).

PROOF. Let G0 ∈Γ(n−k+ 1, e−k) be a τ− optimal graph. By Lemma 1, we know thatG0 is two connected. Since 1< k <3n−2e+ 2, we obtain that the number of degree two vertices in G0 is at least two. Without loss of generality, we assume u, v∈V(G0) and|NG0(u)|=|NG0(v)|= 2.

We distinguish two cases:

Case 1. u6∈NG0(v).

Since 6≤n−k+ 3≤e−k, by Lemma 2, we haveNG0(u)6=NG0(v). We construct graph Gas follows,

V(G) =V(G0)∪ {p1, p2, ..., pk1}, E(G) =E(G0)∪ {(up1),(p1p2), ...,(pk1v)},

where u, vare two endpoints ofLk. ClearlyG∈Γ(n, e). By Lemma 3, we have bt(n, e) ≥ t(G)

= t(G−Lk)k+t(G/Lk)

= t(G0)k+t(G0/{u, v})

> 3t(G0)

= 3ˆt(n−k+ 1, e−k).

Case 2. u∈NG0(v).

Without loss of generality, we assume that the number of degree two vertices inG0 is two. Let a ∈ NG0(u), b ∈NG0(v). Since 4 + 3(n−k−1)−2e+ 2k ≥ 0 and the equality holds if and only ifk= 3n−2e+ 1, we havedG0(a) =dG0(b) = 3. By Lemma 4, we knowa6∈N(b).

Case 2.1. NG0(a)−u6=NG0(b)−v.

(6)

LetG00=G0− {u, v}+ (ab).By Lemma 3, we have

t(G00−(ab)) =t(G0− {u, v})< t((G0− {u, v})/{a, b}) =t(G00/(ab)).

We construct graphGas follows,

V(G) =V(G0)∪ {p1, p2, ..., pk1}, E(G) =E(G0)∪ {(up1),(p1p2), ...,(pk1b)},

where u, bare two endpoints ofLk. ClearlyG∈Γ(n, e). Analogously, we have bt(n, e) ≥ t(G)

= t(G−Lk)k+t(G/Lk)

= t(G0)k+ 2t(G00)

= t(G0)k+ 2t(G00−(ab)) + 2t(G00/(ab))

> t(G0)k+ 3t(G00−(ab)) +t(G00/(ab))

≥ 3t(G0)

= 3ˆt(n−k+ 1, e−k).

Case 2.2. NG0(a)−u=NG0(b)−v.

In this case,G0 is the graph as follows.

a

u v b

a

u

b

G ' H

v

By Lemma 5, we have

)= (G'

t + = + +

)= (H

t + = + +

Clearlyt(H)> t(G0), a contradiction. This means that this case is impossible.

Acknowledgment. The authors are very grateful to the referee for his valuable sug- gestions and comments, which helped to improve the presentation of the paper.

(7)

References

[1] F. T. Boesch, X. M. Li and C. Suffel, On the existence of uniformly optimally reliable networks, Networks, 21(1991) 181–194.

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

[3] B. Gilbert and W. Myrvold, Maximizing spanning trees in almost complete graphs, Networks, 30(1997), 23–30.

[4] X. M. Li, Some properties ofτ-optimal graphs, Proc. of ICCAS’89, 1989, 736–739, Nanjing.

[5] L. Petingi, F. T. Boesch and C. Suffel, On the characterization of graphs with maximun number of spanning tree, Discrete Math., 179(1998), 155–166.

[6] L. Petingi and J. Rodriguez, A new technique for the characterization of graphs with a maximun number of spanning trees, Discrete Math., 244(2002), 351–373.

[7] G. F. Wang, A proof of Boesch’s conjecture, Networks, 24(1994), 277–284.

参照

関連したドキュメント

We establish why expected value is insensitive to catastrophic risks see the study by Chichilnisky 1996, and use another criterion to evaluate risk based on axioms for choice

As a result of this computer-based market analysis, the following findings were made: 1 improvements in the forecast accuracy of fundamentalists can contribute to an increase in

The first bound for the 3- SAT threshold has been obtained by several authors as a direct application of the first moment method to the random variable giving the number of solutions

This paper gives spectral characterizations of two closely related graph functions: the Lov´asz number ϑ and a generalization ϑ 1 of Delsarte’s linear programming bound.. There are

Sopena, On the maximum average degree and the oriented chromatic number of a graph, Discrete Math. Brualdi

First, is there a combinatorial significance to the fact that essentially all studied sequences listed in the EIS [5] that have the Hankel transform {1, 1, 1, 1,…} and are related

The edges terminating in a correspond to the generators, i.e., the south-west cor- ners of the respective Ferrers diagram, whereas the edges originating in a correspond to the

The main result shows that if the boundary of the domain contains n 1 points which form extreme points of an n-simplex, then the equivalence of the Apollonian metric and its