• No results found

On the sum of the two largest Laplacian eigenvalues of unicyclic graphs

N/A
N/A
Protected

Academic year: 2020

Share "On the sum of the two largest Laplacian eigenvalues of unicyclic graphs"

Copied!
8
0
0

Loading.... (view fulltext now)

Full text

(1)

R E S E A R C H

Open Access

On the sum of the two largest Laplacian

eigenvalues of unicyclic graphs

Yirong Zheng

1,2*

, An Chang

2

and Jianxi Li

2,3

*Correspondence:

yrzheng@xmut.edu.cn 1School of Applied Mathematics,

Xiamen University of Technology, Xiamen, Fujian, P.R. China 2Center for Discrete Mathematics,

Fuzhou University, Fuzhou, Fujian, P.R. China

Full list of author information is available at the end of the article

Abstract

LetGbe a simple graph andS2(G) be the sum of the two largest Laplacian

eigenvalues ofG. Guanet al.(J. Inequal. Appl. 2014:242, 2014) determined the largest value forS2(T) among all trees of ordern. They also conjectured that among all

connected graphs of ordernwithm(n≤m≤2n– 3) edges,Gm,nis the unique graph which has maximal value ofS2(G), whereGm,nis a graph of ordernwithmedges which hasmn+ 1 triangles with a common edge and 2n–m– 3 pendent edges incident with one end vertex of the common edge. In this paper, we confirm the conjecture withm=n.

MSC: 05C50; 15A48

Keywords: Laplacian eigenvalue; sum of eigenvalues; largest eigenvalue; unicyclic graph

1 Introduction

LetG= (V,E) be a simple connected graph with vertex setV(G) and edge setE(G). Its order is|V(G)|, denoted byn(G) (ornfor short), and its size is|E(G)|, denoted bym(G) (ormfor short). For a vertexvV(G), letN(v) be the set of all neighbors ofvinGand letd(v) =|N(v)|be the degree ofv. Particularly, denote by(G) the maximum degree ofG. A pendent vertex is a vertex with degree one. The diameter of a connected graphG, denoted byd(G), is the maximum distance among all pairs of vertices inG. LetSnandPn be the star and the path of ordern, respectively. LetSka,bbe the tree of ordernobtained from two starsSa+,Sb+by joining a path of lengthkbetween their central vertices (see Figure ). For all other notions and definitions, not given here, see, for example, [] or [] (for graph spectra).

LetA(G) andD(G) be the adjacency matrix and the diagonal matrix of vertex degrees of G, respectively. The matrixL(G) =D(G) –A(G) is calledLaplacianmatrix ofG. The Lapla-cianmatrix is an important topic in the theory of graph spectra. We use the notationIn for the identity matrix of ordernand denote byφ(G,x) =det(xInL(G)) theLaplacian characteristic polynomial ofG. It is well known thatL(G) is positive semidefinite sym-metric and singular. Denote its eigenvalues byμ(G)≥μ(G)≥ · · · ≥μn(G) (or simply μ≥μ≥ · · · ≥μn sometimes for convenience) which are always enumerated in non-increasing order and repeated according to their multiplicity. Note that each row sum of L(G) is  and, therefore,μn(G) = . Fiedler [] showed that the second smallest eigenvalue μn–(G) ofL(G) is  if and only ifGis disconnected. Thus the second smallest eigenvalue

(2)

Figure 1 Gn+1,n.Gm,nis a graph of ordernwithmedges which hasmn+ 1

triangles with a common edge and 2nm– 3 pendent edges incident with one end vertex of the common edge.

ofL(G) is popularly known as the algebraic connectivity ofG. The largest eigenvalueμ(G) ofL(G) is usually called theLaplacianspectral radius of the graphG. Recently, some of the research has been focused onμ,μorun–(see [–]).

LetSk(G) =i=ki=μi(G) be the sum of theklargestLaplacianeigenvalues ofG. Brouwer conjectured thatSk(G)≤m(G) +k+ fork= , , . . . ,n. The conjecture is still open, but some advances on the conjecture have been achieved (see [–]). Specially, fork= , Haemerset al. [] proved thatS(G)≤m(G) +  for any graphG. When Gis a tree, Fritscheret al.[] improved this bound by givingS(T)≤m(T) +  – n(T) , which im-plies that Haemers’ bound is always not attainable for trees. Therefore, it is interesting to determine which tree has maximal value ofS(T) among all trees of ordern. Guanet al.[] proved thatS(T)≤S(Tn–

 ,n–

) for any tree of ordern≥, and the equality

holds if and only ifT∼=Tn–  ,n–

. For any graphGof ordernwithmedges, note that μ(G)≤n(G). Then Haemers’ bound is clearly not attainable when n(G) <m(G) + . For m(G) + ≤n(G), Guanet al.[] showed thatS(Gm,n) =m(Gm,n) + , whereGm,nis a graph of ordernsizemwhich hasmn+  triangles with a common edge and nm–  pendent edges incident with one end vertex of the common edge (Gn+,n is illustrated in Figure ). This indicates that Haemers’ bound is always sharp for connected graphs (m≤n– ) other than trees. The following conjecture on the uniqueness of the extremal graph is also presented in [].

Conjecture .[] Among all connected graphs of order n with m edges(nm≤n– ), Gm,nis the unique graph with maximal value of S(G),that is,S(Gm,n) =m(Gm,n) + .

In this paper, we confirm Conjecture . withm=n.

2 Preliminaries

In this section, we present some lemmas which will be useful in the subsequent sections. Forμ(G), the following results are well known.

Lemma . Let G be a connected graph of order n,di=d(vi)and mi=

vjN(vi)dj/di.Then () []μ(G)≤n(G)with equality if and only if the complement ofGis disconnected; () []μ(G)≤max{di+mi|viV(G)}.

Lemma .[] Let G be a connected graph of order n≥with m edges.Then

μ(G) <max

(G),mn–  

+ .

Lemma .[] Let T be a tree of order n with d(T)≥.Thenμ(T) <n– ..

(3)

Proof For any treeT of ordernwithd(T)≥, it follows thatn≥ and(T)≤n– . That is,(T) + ≤n–  andmn– +  =n– + ≤n– . Then the result follows from

Lemma ..

The following theorem from matrix theory plays a key role in our proofs. We denote the eigenvalues of a symmetric matrixMof ordernbyλ(M)≥λ(M)≥ · · · ≥λn(M).

Theorem .[] Let A and B be two real symmetric matrices of size n.Then, for any ≤kn,

k

i=

λi(A+B)≤ k

i= λi(A) +

k

i= λi(B).

From Theorem ., the following lemma is immediate.

Lemma . Let G, . . . ,Grbe some edge disjoint graphs.Then,for any k,

Sk(G∪ · · · ∪Gr)≤ r

i= Sk(Gi).

The following lemma can be found in [] and is well known as the interlacing theorem ofLaplacianeigenvalues.

Lemma .[] Let G be a graph of order n and let Gbe the graph obtained from G by inserting a new edge into G.Then the Laplacian eigenvalues of G and Ginterlace,that is,

μ

Gμ(G)≥ · · ·μn

Gμn(G) = .

Lemma .[] Let A be a symmetric matrix of order n with eigenvaluesλ≥λ≥ · · · ≥λn and B be a principal submatrix of A of order m with eigenvaluesμ≥μ≥ · · · ≥μm.Then the eigenvalues of B interlace the eigenvalues of A,that is,λiμiλn–m+ifor i= , . . . ,m. Specially,for vV(G),let Lv(G)be the principal submatrix of L(G)formed by deleting the row and column corresponding to vertex v,then the eigenvalues of Lv(G)interlace the eigenvalues of L(G).

Lemma . For any tree T of order n, if there exists an edge eE(T) such that

min{e(T),e(T)} ≥ and min{d(T),d(T)} ≥ ormax{d(T),d(T)} ≥,then S(T) < n(T) + ,where T,Tare the two components of Te.

Proof LetTe=T∪T. By Lemma ., it suffices to show thatS(T∪T) <n(T) – . If μ(T) =μ(T) andμ(T) =μ(T) (orμ(T) =μ(T) andμ(T) =μ(T)), then the result follows sinceS(T∪T) =S(Ti) <m(Ti) + m(T) =n(T) –  fori= , . In what follows, we assume thatS(T∪T) =μ(T) +μ(T). Ifmin{d(T),d(T)} ≥, then Lemma . implies thatS(T∪T) < (n(T) – .) + (n(T) – .) =n(T) – . Ifmax{d(T),d(T)} ≥, sayd(T)≥, then Lemmas . and . imply thatS(T∪T) <n(T) + (n(T) – ) =

n(T) – . This completes the proof.

(4)
[image:4.595.118.480.81.174.2]

Figure 2 Two trees of ordern.

Proof Without loss of generality, we assume thatvvvvvvis a path of length  inT. We now consider the following three cases.

Case .min{d(v),d(v)} ≥.

LetT,Tbe the two components ofTvv. Then the result follows from Lemma .. Case .d(v) =d(v) = .

Note thatd(T) = . ThenT is isomorphic toSa,b (see Figure ), wherea,b≥ anda+ b+  =n. By an elementary calculation, we haveφ(Sa,b ,λ) =λ(λ– )n–f(λ), wheref

(λ) = λ– (n+ )λ+ (n+ab– )λ– (n+ ab– )λ+n. Denoted byλ

≥λ≥λ≥λ>  are the four roots off(λ) = . Then we haveλ+λ+λ+λ=n+  since

i=n

i=μi= m. Note thatSa,b containsPas a subgraph. Then Lemma . implies thatλ≥μ(P) = . Thus,S(T) =λ+λ=n+  – (λ+λ) <n+  –λ<n– , as desired.

Case .d(v)≥ andd(v) =  (ord(v)≥ andd(v) = ).

Without loss of generality, we assume thatd(v)≥ andd(v) = . Ifd(v)≥, letTvv=T∪T; if there isP=uvv attached tov, letTvv=T∪T, whereu,v= vi(i= , , . . . , ). Then the result follows from Lemma .. We now assume thatd(v) = d(v) =  and all the neighbors of v except forv andvare pendent vertices. That is, T is isomorphic toHa,b, whereHa,b (see Figure ) is the tree of ordernobtained from vvvvvvby attachingaandb–  pendent vertices tovandv, respectively, where a,b≥ anda+b+  =n. Note that the matrix ·InL(Ha,b) hasaandbdifferent identical rows. Then the multiplicity of eigenvalue  is at leastn– . Letλ≥λ≥λ≥λ≥λ≥ λ>λ=  be the other seven eigenvalues. Thenλ+λ+λ+λ+λ+λ+λ=n+  since

i=n

i=μi= m. Fora,b≥,Ha,bcontainsH,as a subgraph. Then by Lemma . we have λ≥μ(H,) = . andλ≥μ(H,) = .. Therefore,S(T) =λ+λ<n+  – (λ+ λ) <n+ , as required. Ifa= , then by Lemmas . and . we haveμ(H,b)≤(n– ) +n–n– andμ(H,b)≤μ(Lv(H,b)) = .. That is,S(T) =S(H,b) =μ(H,b)+μ(H,b) <n+, as

required. Similarly, ifb= , then by Lemmas . and . we haveμ(Ha,)≤(n–)+n–n–and μ(Ha,)μ(Lv(Ha,)) = .. It follows thatS(T) =S(Ha,) =μ(Ha,) +μ(Ha,) <n+ ,

as required.

From the discussion above, the proof is completed.

Lemma . Let T be a tree of order n with d(T)≥.Then S(T) <n(T) + .

Proof We now consider the following two cases. Case .d(T)≥.

Letvvvvvvvvbe a path of length  inTandTvv=T∪T. Then the result follows from Lemma ..

Case .d(T) = .

(5)
[image:5.595.115.481.79.172.2]

Figure 3 An example of firefly graph.

T∪T (orTvv=T∪T); ifd(v)≥ (ord(v)≥), letTvv=T ∪T (or Tvv=T∪T). In each of the above cases, by Lemma ., we haveS(T) <n(T) + .

This completes the proof.

A firefly graphFs,t,n–s–t–(s≥,t≥ andn– s– t– ≥) is a graph withnvertices that consists ofstriangles,tpendent paths of length  andn– s– t–  pendent edges, sharing a common vertex. An example of a firefly graphF,,is illustrated in Figure . ClearlyF,,n–∼=SnandF,,n–∼=Gn,n.

Lemma . [] The second largest Laplacian eigenvalue of F,t,n–t– satisfies μ(F,t,n–t–) = .

Note that fort≥, the complement ofF,t,n–t–is connected. Hence Lemma . im-plies thatμ(F,t,n–t–) <n(F,t,n–t–). This together with Lemma . implies the follow-ing lemma.

Lemma . For t≥,S(F,t,n–t–) <n(F,t,n–t–) + .

3 Main result

A unicyclic graph is a connected graph whose number of edgesmis equal to the number of verticesn. It is easy to see that each unicyclic graph can be obtained by attaching rooted trees to the vertices of a cycleCkfor somek. Thus ifR, . . . ,Rkarekrooted trees (of orders n, . . . ,nk, say), then we adopt the notationU(R, . . . ,Rk) to denote the unicyclic graphG (of ordern=n+· · ·+nk) obtained by attaching the rooted treeRito the vertexviof a cycleCk=vv· · ·vkv(i.e., by identifying the root ofRiwith the vertexvifori= , . . . ,k). Denote bye(vi) the maximum distance betweenviand any vertex ofRi. In the special case whenRiis a rooted starK,ai with the center of the star as its root (that is,e(vi) = ), we

will simplify the notation by replacingRiby the numberai.

The following lemma is immediate from Lemmas ., . and ..

Lemma . For any unicyclic graph G of order n with m edges,if there exists an edge eE(G)such that Ge is a tree with d(Ge)≥,then S(G) <m(G) + .

Letuvwu be a triangle andTa,b,c be the graph obtained by attachinga,b,c pendent vertices tou,v,w, respectively, wherea+b+c=n– andabc≥. Note thatTn–,,∼= Gn,n. Denote byQa,bthe graph obtained by attachingaandbpendent vertices to two non-adjacent vertices of a quadrangle, respectively, wherea+b=n–  andab≥.Ta,b,c andQa,bare illustrated in Figure .

(6)
[image:6.595.119.478.82.185.2]

Figure 4 Two unicyclic graphs of ordern.Ta,b,cis a unicyclic graph of ordernobtained by attachinga,b,c

pendent vertices to three vertices of a triangleuvw, respectively, andQa,bis a unicyclic graph of ordern

obtained by attachingaandbpendent vertices to two non-adjacent vertices of a quadrangle, respectively.

Proof Note that the matrix ·InL(Ta,b,c) has a, band cdifferent identical rows. So the multiplicity of eigenvalue  is at leastn– . Letλ≥λ≥λ≥λ≥λ>λ=  be the other six eigenvalues ofL(Ta,b,c). Then we haveλ+λ+λ+λ+λ=n+  since

i=n

i=μi= m(G). Forc≥,Ta,b,ccontainsT,,as a subgraph. Then by Lemma . we haveλ≥μ(T,,) = .. Therefore,S(Ta,b,c) =λ+λ=n+–(λ+λ+λ) <n+–λ≤ n+  =m(Ta,b,c) + , as required. Forc=  andb≥,Ta,b,ccontainsT,, as a subgraph. Then by Lemma . we haveλ≥μ(T,,) = .. That is,S(Ta,b,) =n+  – (λ+λ+λ) < n+  –λ≤n+  =m(Ta,b,c) + , as required. Ifc=  andb= , then by Lemmas . and ., we haveμ(Ta,,)n–  + n– andμ(Ta,,)μ(Lu(Ta,,)) = .. It follows that S(Ta,,) =μ(Ta,,) +μ(Ta,,) <m(Ta,,) + , as required. Ifc=  andb= , then by Lem-mas . and ., we haveμ(Ta,,)≤n–  +n– andμ(Ta,,)≤μ(Lu(Ta,,)) = .. That is,S(Ta,,) =μ(Ta,,) +μ(Ta,,) <m(Ta,,) +  forn≥ (that is,a≥), as required. A direct calculation shows thatS(T,,) <m(T,,) +  (orS(T,,) <m(T,,) + ). Sim-ilarly, ifc=  andb= , then by Lemmas . and ., we haveμ(Ta,,)≤n–  +n– and μ(Ta,,)μ(Lu(Ta,,)) = .. That is,S(Ta,,) =μ(Ta,,) +μ(Ta,,) <m(Ta,,) + , for n≥ (that is,a≥), as required. A direct calculation shows thatS(T,,) <m(T,,) +  (orS(T,,) <m(T,,) +  orS(T,,) <m(T,,) + ). Next, we assumec= . By an ele-mentary calculation, we haveφ(Ta,b,,λ) =λ(λ– )n–f(λ), wheref(λ) =λ– (n+ )λ+ (n+ab+ )λ– (n+ ab+ )λ+ n. Letx≥x≥x≥xbe the roots off(λ) = . Then

x+x+x+x=n+ , (.)

xx+xx+xx+xx+xx+xx= n+ab+ , (.)

xxx+xxx+xxx+xxx= n+ ab+ . (.)

If

x+x=n+ , (.)

then, by (.), we have

x+x= . (.)

From (.)-(.) it follows that

(7)

(n+ )xx+ xx= n+ ab+ . (.)

By (.) and (.), we have

xx= . (.)

Combining (.) and (.), we have

x=x= . (.)

Then f() = –ab= , which implies that b= . Therefore, if b≥, thenS(Ta,b,) < m(Ta,b,) + . A direct calculation shows thatS(Tn–,,) =m(Tn–,,) + . This completes

the proof.

Lemma . For Qa,b,S(Qa,b) <m(Qa,b) + .

Proof By a direct calculation, we have φ(Qa,b,λ) =λ(λ– )n–f(λ), where f(λ) = (λ )(λ– (n+ )λ+ (n+ab+ )λ– (n+ ab– )λ+ n). Letλ

≥λ≥λ≥λ≥λ>  be the five roots off(λ) = . Then we haveλ+λ+λ+λ+λ=n+  since

i=n

i=μi= m(G). Forb≥,Qa,bcontainsQ,as a subgraph. Then by Lemma . we haveλ≥μ(Q,) =  andλ≥μ(Q,) = .. Therefore,S(Qa,b) =λ+λ=n+  –λ–λ–λ<n+ . In what follows, we assumeb= . Sincef() = , we can rewriteφ(Qa,,λ) =λ(λ– )n–f(λ), where f(λ) = (λ– )f(λ). Letλλλλ>  be the four roots off(λ) = . Then we have λ+λ+λ+λ=n+  sincei=ni=μi= m(G). Fora≥,Qa,containsQ,as a subgraph. Then by Lemma . we haveλμ(Q,) = . Thus,S(Qa,b) =λ+λ=n+  –λ–λ< n+ . Ifa= , thenS(Qa,b) =S(C) <m(C) +  by a straight calculation. This completes

the proof.

Now, we come to the main results of this paper.

Theorem . For any unicyclic graph G, S(G)≤m(G) + with equality if and only if G∼=T(n– , , ).

Proof For any unicyclic graphG, we assume that Ck =vv· · ·vkv is the unique cycle in G(for somek) andGhas the form U(R, . . . ,Rk). Fork≥,Gvv is a tree with d(Gvv)≥. Then by Lemma . we haveS(G) <m(G) + . We now consider the fol-lowing two cases.

Case .k= .

LetC=vvvvvbe the unique cycle inG. If there existe(vi)≥, saye(v)≥, then Gvvis a tree withd(Gvv)≥; if there are two adjacent vertices inC=vvvvv, sayvandv, such thate(v)≥ ande(v)≥, thenGvvis a tree withd(Gvv)≥. Then by Lemma . we haveS(G) <m(G) + . We now assumeG∼=Qa,b(see Figure ). Then the result follows from Lemma ..

Case .k= .

(8)

m(G) + . We now assumeG∼=F,t,n–t–(t≥) orG∼=Ta,b,c. Then the result follows from Lemma . or Lemma ..

This completes the proof.

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

YZ carried out the proofs of the main results in the manuscript. AC and JL participated in the design of the study and drafted the manuscript. All the authors read and approved the final manuscript.

Author details

1School of Applied Mathematics, Xiamen University of Technology, Xiamen, Fujian, P.R. China.2Center for Discrete

Mathematics, Fuzhou University, Fuzhou, Fujian, P.R. China. 3School of Mathematics and Statistics, Minnan Normal University, Zhangzhou, Fujian, P.R. China.

Acknowledgements

The authors would like to thank the anonymous referees for their constructive corrections and valuable comments on this paper, which have considerably improved the presentation of this paper. This project is supported by NSF of China (Nos. 11471077, 11301440), China Postdoctoral Science Foundation (No. 2014M551831), NSF of Fujian (No. 2014J01020), the Foundation to the Educational Committee of Fujian (JA13240, JA15381, JB13155).

Received: 31 March 2015 Accepted: 23 August 2015

References

1. Brouwer, A, Haemers, W: Spectra of Graphs. Springer, New York (2012) 2. Godsil, C, Royle, G: Algebraic Graph Theory. Springer, New York (2001)

3. Fiedler, M: Algebraic connectivity of graphs. Czechoslov. Math. J.23, 298-305 (1973)

4. Anderson, W, Morley, T: Eigenvalues of the Laplacian of a graph. Linear Multilinear Algebra18, 141-145 (1985) 5. Guo, J-M: On the second largest Laplacian eigenvalue of trees. Linear Algebra Appl.404, 251-261 (2005)

6. Li, J, Guo, J-M, Shiu, WC: On the second largest Laplacian eigenvalue of graphs. Linear Algebra Appl.438, 2438-2446 (2013)

7. Liu, M, Liu, B, Cheng, B: Ordering (signless) Laplacian spectral radii with maximum degree of graphs. Discrete Math.

338, 159-163 (2015)

8. Merris, R: A note on the Laplacian graph eigenvalues. Linear Algebra Appl.285, 33-35 (1990)

9. Zhang, X-D, Li, J: The two largest Laplacian eigenvalues of trees. J. Univ. Sci. Technol. China28, 513-518 (1998) 10. Du, Z, Zhou, B: Upper bounds for the sum of Laplacian eigenvalues of graphs. Linear Algebra Appl.436, 3672-3683

(2012)

11. Guan, M, Zhai, M, Wu, Y: On the sum of two largest Laplacian eigenvalue of trees. J. Inequal. Appl.2014, 242 (2014) 12. Haemers, W, Mohammadian, A, Tayfeh-Rezaie, B: On the sum of Laplacian eigenvalues of graphs. Linear Algebra Appl.

432, 2214-2221 (2010)

13. Rocha, I, Trevisian, V: Bounding the sum of the Laplacian graph eigenvalues of graphs. Discrete Appl. Math.170, 95-103 (2014)

14. Wang, S, Huang, Y, Liu, B: On the conjecture for the sum of the Laplacian graph eigenvalues. Math. Comput. Model.

56, 60-68 (2012)

15. Fritscher, E, Hoppen, C, Rocha, I, Trevisan, V: On the sum of the Laplacian eigenvalues of a tree. Linear Algebra Appl.

435, 371-399 (2011)

Figure

Figure 2 Two trees of order n.
Figure 3 An example of firefly graph.
Figure 4 Two unicyclic graphs of order n. Ta,b,c is a unicyclic graph of order n obtained by attaching a, b, cpendent vertices to three vertices of a triangle uvw, respectively, and Qa,b is a unicyclic graph of order nobtained by attaching a and b pendent vertices to two non-adjacent vertices of a quadrangle, respectively.

References

Related documents

For each climate risk indicator, open-access tools were utilized to source historical, current and projection data to represent the overall climate risk for each facility of a list

It is expressed on a fraction of pro-B (pre-BI) cells and most large pre-B(II) cells, and has been implicated in IgH chain allelic exclusion and down-regulation of the

Chemical composition and antioxidant activity of essential oil of leaves and flowers of Alternanthera sessilis red from Sabah.. Mohammad Shaheen Khan 1* , Samina Khan Yusufzai 2

This approach will also enable to assess early changes on organisms’ fitness, at organizational levels below the individual that can not become apparent in other types of

There shall be a total of five Board members representing Aboriginal organizations, one for Aboriginal organizations in each of British Columbia, Alberta,

Wang, “A dual-polarized pattern reconfigurable yagi patch antenna for microbase stations,” IEEE Transactions on Antennas and Propagation, vol. Ka- marudin,

Annex 1 of this report includes some of the available IT security data collected and published by: the Australian Bureau of Statistics (Business Use of Information Technology