This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

\floatsetup

[table]capposition=top

On the sum of the first two largest signless Laplacian eigenvalues of a graph111 Corresponding author Email: [email protected] (C.-X. He).
This research is supported by National Natural Science Foundation of China (No.12271362).

Zi-Ming Zhou College of Science, University of Shanghai for Science and Technology, Shanghai, P. R. China. Chang-Xiang He College of Science, University of Shanghai for Science and Technology, Shanghai, P. R. China. Hai-Ying Shan School of Mathematical Science, Tongji University, Shanghai, P. R. China.
Abstract

For a graph GG, let S2(G)S_{2}(G) be the sum of the first two largest signless Laplacian eigenvalues of GG, and f(G)=e(G)+3S2(G)f(G)=e(G)+3-S_{2}(G). Oliveira, Lima, Rama and Carvalho conjectured that K1,n1+K^{+}_{1,n-1} (the star graph with an additional edge) is the unique graph with minimum value of f(G)f(G) on nn vertices. In this paper, we prove this conjecture, which also confirm a conjecture for the upper bound of S2(G)S_{2}(G) proposed by Ashraf et al.

Keywords: (Signless) Laplacian matrix; (Signless) Laplacian eigenvalues; Sum of (Signless) Laplacian eigenvalues

1 Introduction

Graphs in this paper are connected and simple. Let G=(V(G),E(G))G=(V(G),E(G)) be a graph, the number of its vertices and edges are denoted by n(G)n(G) (or nn for short) and e(G)e(G), respectively. Let A(G)A(G) be the adjacency matrix of GG and D(G)=diag(d(v1),d(v2),,d(vn))D(G)=diag(d(v_{1}),d(v_{2}),\ldots,d(v_{n})) be the diagonal matrix of vertex degrees of GG. The Laplacian matrix and the signless Laplacian matrix of GG are defined as L(G)=D(G)A(G)L(G)=D(G)-A(G) and Q(G)=D(G)+A(G)Q(G)=D(G)+A(G), respectively. It is well known that both L(G)L(G) and Q(G)Q(G) are positive semidefinite and hence their eigenvalues are nonnegative real numbers. The eigenvalues of L(G)L(G) and Q(G)Q(G) are called Laplacian eigenvalues and signless Laplacian eigenvalues of GG, respectively, and are denoted by μ1(G)μ2(G)μn(G)\mu_{1}(G)\geq\mu_{2}(G)\geq\cdots\geq\mu_{n}(G) (or μ1μ2μn\mu_{1}\geq\mu_{2}\geq\cdots\geq\mu_{n} for short) and q1(G)q2(G)qn(G)q_{1}(G)\geq q_{2}(G)\geq\cdots\geq q_{n}(G) (or q1q2qnq_{1}\geq q_{2}\geq\cdots\geq q_{n} for short), respectively.

In [4], Brouwer proposed the following conjecture for Laplacian eigenvalues.

Conjecture 1.

( [4]) For any graph GG with nn vertices and for any k{1,2,,n}k\in\{1,2,\ldots,n\},

i=1kμi(G)e(G)+(k+12).\sum_{i=1}^{k}\mu_{i}(G)\leq e(G)+\binom{k+1}{2}.

It has been proved that Conjecture 1 is true for several classes of graphs such as trees [11], threshold graphs [11], unicyclic graphs [7], [15], bicyclic graphs [7], regular graphs [13] and split graphs [13]. Recently, Cooper [5] showed that Conjecture 1 is true for planar graphs when k11k\geq 11 and for bipartite graphs when k32nk\geq\sqrt{32}n. For k=2k=2, Haemers et al.[11] showed that Conjecture 1 is true. In [12], Li and Guo proposed the full Brouwer’s Laplacian spectrum conjecture and proved the conjecture holds for k=2k=2 with the unique extremal graph.

Let Sk(G)S_{k}(G) be the sum of the first kk largest signless Laplacian eigenvalues of GG. Ashraf et al.[3] proposed the following conjecture.

Conjecture 2.

([3]) For any graph GG with nn vertices and for any k{1,2,,n}k\in\{1,2,\ldots,n\},

Sk(G)e(G)+(k+12).S_{k}(G)\leq e(G)+\binom{k+1}{2}.

For bipartite graphs, Conjecture 2 is equivalent to Conjecture 1 since L(G)L(G) and Q(G)Q(G) are similar [6]. Ashraf et al.[3] proved that Conjecture 2 is true for regular graphs, they also proved that Conjecture 2 is true for k=2k=2, but the key lemma ([3], Lemma 13) they used is incorrect which has a counterexample in[17]. Zheng[17] proved that Conjecture 2 is true for all connected triangle-free graphs when k=2k=2. Therefore, for general graphs Conjecture 2 is still open when k=2k=2.

For a graph GG, define

f(G)=e(G)+3S2(G).f(G)=e(G)+3-S_{2}(G).

Oliveira et al.[14] conjectured that K1,n1+K^{+}_{1,n-1} (the star graph with an additional edge) minimizes f(G)f(G) among all graphs GG on nn vertices.

Conjecture 3.

([14]) For any graph GG on n9n\geq 9 vertices. Then

f(G)f(K1,n1+).f(G)\geq f(K^{+}_{1,n-1}).

Equality if and only if GK1,n1+G\cong K^{+}_{1,n-1}.

In this paper, we give a proof of Conjecture 3.

Theorem 1.1.

For any graph GG with nn vertices, f(G)f(K1,n1+)f(G)\geq f(K^{+}_{1,n-1}) with equality if and only if GK1,n1+G\cong K^{+}_{1,n-1}.

By computing the value of f(K1,n1+)f(K^{+}_{1,n-1}), one can obtain a tight upper bound of S2(G)S_{2}(G) which confirm Conjecture 2 in the case k=2k=2.

The organization of the remaining of the paper is as follows. In Section 2, we give some lemmas which are useful in the proof of Theorem 1.1. In Section 3, we give the proof of Theorem 1.1.

2 Preliminaries

We begin this section with some definitions and notation. For vertex vV(G)v\in V(G), the set of its neighbors in XV(G)X\subset V(G), denoted by NX(v)N_{X}(v), for convenience, abbreviate NV(G)(v)N_{V(G)}(v) as N(v)N(v). For two graphs G1G_{1} and G2G_{2}, the union of G1G_{1} and G2G_{2}, denoted by G1G2G_{1}\cup G_{2}, is the graph with vertex set V(G1)V(G2)V(G_{1})\cup V(G_{2}) and edge set E(G1)E(G2)E(G_{1})\cup E(G_{2}). If HH is a subgraph of GG, let GE(H)G-E(H) denote the graph obtained from GG by removing the edges in HH.

For any square matrix MM, we use P(M,x)P(M,x) to denote the characteristic polynomial of MM. For graph GG, we write the characteristic polynomial of its signless Laplacian matrix P(Q(G),x)P(Q(G),x) as PQ(G,x)P_{Q}(G,x).

HH-join operation of graphs: Let HH be a graph with V(H)={1,2,,k}V(H)=\{1,2,...,k\}, and GiG_{i} be disjoint rir_{i}-regular graphs of order ni(i=1,2,,k)n_{i}\ (i=1,2,...,k). Then the graph H{G1,G2,,Gk}\bigvee_{H}\{G_{1},G_{2},...,G_{k}\} is formed by taking the graphs G1,G2,,GkG_{1},G_{2},...,G_{k} and joining every vertex of GiG_{i} to every vertex of GjG_{j} whenever ii is adjacent to jj in HH. Define Ni=jNH(i)njN_{i}=\sum_{j\in N_{H}(i)}n_{j} for each iV(H)i\in V(H) and

(Qπ)ij={2ri+Ni,i=j,nj,ijE(H),0,otherwise.\displaystyle(Q^{\pi})_{ij}=\begin{cases}2r_{i}+N_{i},&i=j,\\ n_{j},&ij\in E(H),\\ 0,&otherwise.\end{cases}
Lemma 2.1 ([16]).

Let HH be a graph with V(H)={1,2,,k}V(H)=\{1,2,...,k\}, and GiG_{i} be a rir_{i}-regular graph of order ni(i=1,2,,k)n_{i}\ (i=1,2,...,k). If G=H{G1,G2,,Gk}G=\bigvee_{H}\{G_{1},G_{2},...,G_{k}\}, then

PQ(G,x)=P(Qπ,x)i=1kPQ(Gi,xNi)x2riNi.\displaystyle P_{Q}(G,x)=P(Q^{\pi},x)\prod_{i=1}^{k}\frac{P_{Q}(G_{i},x-N_{i})}{x-2r_{i}-N_{i}}.

Lexicographically ordered kk-sets: Fix a number kk between 1 and nn, inclusive. Consider all possible sets of kk distinct numbers between 1 and nn. Order the elements of each set by using a lexicographic or dictionary order. Denote these sets by S1,S2,,S(nk)S_{1},S_{2},\ldots,S_{\tbinom{n}{k}} so that S1<S2<<S(nk)S_{1}<S_{2}<\ldots<S_{\tbinom{n}{k}}.

Compound matrix: Let Ai,jA_{i,j} be the k×kk\times k matrix formed by the intersection, in AA, of the rows whose numbers are in the set SiS_{i} and the columns whose numbers are in the set SjS_{j}. Let AA be a complex matrix of order nn, define Ck(A)C_{k}(A) as its kk-th compound matrix, which is the matrix of size (nk)×(nk)\tbinom{n}{k}\times\tbinom{n}{k} whose (i,j)(i,j)-th entry is the determinant of the matrix Ai,jA_{i,j}.

Additive compound matrix: The matrix Ck(I+tA)C_{k}(I+tA) is an (nk)×(nk)\tbinom{n}{k}\times\tbinom{n}{k} matrix, each entry of which is a polynomial in tt of degree at most kk. The kk-th additive compound matrix is defined as

Δk(A)=ddtCk(In+tA)|t=0.\displaystyle\Delta_{k}(A)=\frac{d}{dt}C_{k}(I_{n}+tA)\big{|}_{t=0}.

For additive compound matrix, the following lemma is a well-known result.

Lemma 2.2 ([9]).

If λ1,λ2,,λn\lambda_{1},\ \lambda_{2},\ldots,\lambda_{n} are all eigenvalues of AA, then the eigenvalues of Δk(A)\Delta_{k}(A) are the (nk)\tbinom{n}{k} distinct sums of the λi\lambda_{i} taken kk at a time.

Corollary 2.1.

If q1,q2,,qnq_{1},\ q_{2},\ldots,q_{n} are signless eigenvalues of GG, and P(Qπ,x)P(Q^{\pi},x) contains the largest two roots of PQ(G,x)P_{Q}(G,x), then the largest eigenvalue of Δ2(Qπ)\Delta_{2}(Q^{\pi}) equals S2(G)S_{2}(G).

The following result which can be found in matrix theory which plays an important role in our proofs. For a symmetric matrix AA of order nn, let λ1(A)λn(A)\lambda_{1}(A)\geq\cdots\geq\lambda_{n}(A) be its eigenvalues.

Lemma 2.3 ([8]).

Let AA and BB be two real symmetric matrices of size nn. Then for any 1kn1\leq k\leq n,

i=1kλi(A+B)i=1kλi(A)+i=1kλi(B).\sum^{k}_{i=1}\lambda_{i}(A+B)\leq\sum^{k}_{i=1}\lambda_{i}(A)+\sum^{k}_{i=1}\lambda_{i}(B).

Apparently, λ1(Δk(A))=i=1kλi(A)\lambda_{1}(\Delta_{k}(A))=\sum^{k}_{i=1}\lambda_{i}(A), then the above lemma can be directly obtained from the fact that λ1(Δk(A+B))λ1(Δk(A))+λ1(Δk(B)).\lambda_{1}(\Delta_{k}(A+B))\leq\lambda_{1}(\Delta_{k}(A))+\lambda_{1}(\Delta_{k}(B)). An immediate result of Lemma 2.3 is the following corollary which will be used frequently.

Corollary 2.2.

Let GG be a graph of order nn such that there exist edge disjoint subgraphs of GG, say G1,,GrG_{1},\cdots,G_{r}, with E(G)=i=1rE(Gi)E(G)=\bigcup^{r}_{i=1}E(G_{i}). Then Sk(G)i=1rSk(Gi)S_{k}(G)\leq\sum^{r}_{i=1}S_{k}(G_{i}) for any integer kk with 1kn1\leq k\leq n.

The following result is also straightforward.

Lemma 2.4.

For some kk, if Theorem1.1 holds for GG and HH, then it does for GHG\cup H.

The next lemma is the key to our approach. It gives a sufficient condition for the result of Theorem1.1, that holds for almost all graphs.

Lemma 2.5.

If GG is a graph with a nonempty subgraph HH for which Sk(H)e(H)S_{k}(H)\leq e(H), then Sk(G)<e(G)S_{k}(G)<e(G).

Proof.

Assume that GG is a counterexample with a minimum possible number of edges. By Corollary 2.2, we have e(G)Sk(G)Sk(H)+Sk(GH)e(G)\leq S_{k}(G)\leq S_{k}(H)+S_{k}(G-H). This implies that Sk(GH)e(GH)S_{k}(G-H)\geq e(G-H), since e(G)e(H)+e(GH)e(G)\geq e(H)+e(G-H) and Sk(H)e(H)S_{k}(H)\leq e(H), which contradicts the minimality of e(G)e(G). ∎

Corollary 2.3.

If GG is a graph with a nonempty subgraph HH for which S2(H)e(H)S_{2}(H)\leq e(H), then S2(G)<e(G)S_{2}(G)<e(G) and f(G)>3f(G)>3.

Lemma 2.6 ([1]).

For any graph GG with at most 10 vertices, f(G)f(K1,n1+)f(G)\geq f(K^{+}_{1,n-1}) with equality if and only if GK1,n1+G\cong K^{+}_{1,n-1}.

Lemma 2.7 ([6]).

Let nn be a positive integer.

  • (1)

    The signless Laplacian eigenvalues of the complete graph KnK_{n} are 2n22n-2 and n2n-2 with multiplicities 1 and n1n-1, respectively.

  • (2)

    The signless Laplacian eigenvalues of the star K1,n1K_{1,n-1} are nn, 1 and 0 with multiplicities 1, n2n-2, and 1, respectively.

Lemma 2.8 ([10]).

Let TT be a tree with nn vertices and let 1kn1\leq k\leq n. Then the sum of the kk largest Laplacian eigenvalues of TT satisfies i=1kμi(G)(n1)+2k12k2n\sum_{i=1}^{k}\mu_{i}(G)\leq(n-1)+2k-1-\frac{2k-2}{n}. Moreover, equality is achieved only when k=1k=1 and TT is a star on nn vertices.

Using the fact that signless Laplacian matrix and Laplacian matrix are similar for a bipartite graph, from the above lemma, we have the following result.

Corollary 2.4.

Let GG be a tree with nn vertices, then f(G)2nf(G)\geq\frac{2}{n}.

Lemma 2.9 ([6]).

Let GG be a graph with nn vertices and let GG^{\prime} be a graph obtained from GG by inserting a new edge into GG. Then the signless Laplacian eigenvalues of GG and GG^{\prime} interlace, that is,

q1(G)q1(G)qn(G)qn(G).q_{1}(G^{\prime})\geq q_{1}(G)\geq\cdots\geq q_{n}(G^{\prime})\geq q_{n}(G).
Lemma 2.10.

Let GG be a graph with nn vertices. Suppose that there exist two non-adjacent vertices u,vV(G)u,v\in V(G) such that qk(G)d(u)+d(v)+2q_{k}(G)\geq d(u)+d(v)+2 for some integer kk, 1kn1\leq k\leq n. If GG^{\prime} is the graph obtained from GG by inserting an edge e=uve=uv into GG, then Sk(G)<Sk(G)+1S_{k}(G^{\prime})<S_{k}(G)+1.

Proof.

For i=1,,ni=1,\ldots,n, define σi=qi(G)qi(G)\sigma_{i}=q_{i}(G^{\prime})-q_{i}(G). By Lemma 2.9, σi0\sigma_{i}\geq 0, for any ii. Let d1dnd_{1}\geq\cdots\geq d_{n} and d1dnd_{1}^{\prime}\geq\cdots\geq d_{n}^{\prime} be vertex degrees of GG and GG^{\prime}, respectively. Recall that for any graph HH, considering the trace of the matrix Q2(H)Q^{2}(H), we have

i=1|V(H)|qi2(H)=vV(H)d2(v)+2e(H).\sum^{|V(H)|}_{i=1}q^{2}_{i}(H)=\sum_{v\in V(H)}d^{2}(v)+2e(H).

Applying this fact, we have

i=1nqi2(G)\displaystyle\sum^{n}_{i=1}q^{2}_{i}(G^{\prime}) =i=1ndi2+2e(G)\displaystyle=\sum^{n}_{i=1}d_{i}^{\prime 2}+2e(G^{\prime})
=i=1ndi2+2e(G)+2d(u)+2d(v)+4\displaystyle=\sum^{n}_{i=1}d_{i}^{2}+2e(G)+2d(u)+2d(v)+4
=i=1nqi2(G)+2(d(u)+d(v)+2).\displaystyle=\sum^{n}_{i=1}q^{2}_{i}(G)+2(d(u)+d(v)+2).

This yields that

2qk(G)i=1kσi\displaystyle 2q_{k}(G)\sum^{k}_{i=1}\sigma_{i} i=1k2σiqi(G)\displaystyle\leq\sum^{k}_{i=1}2\sigma_{i}q_{i}(G)
<i=1nqi2(G)i=1nqi2(G)\displaystyle<\sum^{n}_{i=1}q^{2}_{i}(G^{\prime})-\sum^{n}_{i=1}q^{2}_{i}(G)
=2(d(u)+d(v)+2).\displaystyle=2(d(u)+d(v)+2).

Since qk(G)d(u)+d(v)+2q_{k}(G)\geq d(u)+d(v)+2, Sk(G)Sk(G)=i=1kσi<1S_{k}(G^{\prime})-S_{k}(G)=\sum^{k}_{i=1}\sigma_{i}<1, and the result follows. ∎

In particular, if Sk(G)<e(G)+(k+12)S_{k}(G)<e(G)+\binom{k+1}{2}, then Sk(G)<e(G)+(k+12)S_{k}(G^{\prime})<e(G^{\prime})+\binom{k+1}{2}.

3 Proof of Theorem1.1

Amaro et al.[1] pointed out that Theorem 1.1 holds for all graphs with at most 10 vertices by computational checking. Assuming the graph we are considering has at least 11 vertices in the next. We firstly present Lemma 3.1 which indicates the graph K1,n1+K^{+}_{1,n-1} is the candidate to be extremal to f(G)f(G).

Lemma 3.1.

Let GG be isomorphic to K1,n1+K^{+}_{1,n-1}. Then 1.3n<f(G)<1.5n\frac{1.3}{n}<f(G)<\frac{1.5}{n}.

Proof.

The graph GG has an obvious partition G=H{G1,G2,G3}G=\bigvee_{H}\{G_{1},G_{2},G_{3}\}, and the corresponding quotient matrix is

Qπ=(3102n1n3011).\displaystyle Q^{\pi}=\left(\begin{array}[]{ccc}3&1&0\\ 2&n-1&n-3\\ 0&1&1\end{array}\right).

By Lemma 2.1 and direct calculation, the characteristic polynomial of Q(G)Q(G) is

PQ(G,x)=P(Qπ,x)(x1)n3.\displaystyle P_{Q}(G,x)=P(Q^{\pi},x)(x-1)^{n-3}.

The 2-th additive compound matrix of QπQ^{\pi} is

Δ2(Qπ)=(n+2n3014102n),\displaystyle\Delta_{2}(Q^{\pi})=\left(\begin{array}[]{ccc}n+2&n-3&0\\ 1&4&1\\ 0&2&n\end{array}\right),

and its characteristic polynomia is P(Δ2(Qπ),x)=x3(6+2n)x2+(n2+9n+9)x3n29n+4P(\Delta_{2}(Q^{\pi}),x)=x^{3}-(6+2n)x^{2}+(n^{2}+9n+9)x-3n^{2}-9n+4. By direct calculation we obtain

P(Δ2(Qπ),n+31.3n)\displaystyle P(\Delta_{2}(Q^{\pi}),n+3-\frac{1.3}{n}) >0,\displaystyle>0,
P(Δ2(Qπ),n+31.5n)\displaystyle P(\Delta_{2}(Q^{\pi}),n+3-\frac{1.5}{n}) <0,\displaystyle<0,
P(Δ2(Qπ),3)\displaystyle P(\Delta_{2}(Q^{\pi}),3) >0,\displaystyle>0,
P(Δ2(Qπ),0)\displaystyle P(\Delta_{2}(Q^{\pi}),0) <0.\displaystyle<0.

Since q22q_{2}\geq 2, by Lemma 2.9, P(Qπ,x)P(Q^{\pi},x) contains the largest two roots of PQ(G,x)P_{Q}(G,x). From Corollary 2.1 and the fact that e(G)=ne(G)=n, we have 1.3n<f(G)<1.5n\frac{1.3}{n}<f(G)<\frac{1.5}{n}. ∎

By Lemma 2.4, we may assume that GG is a connected graph with at least 11 vertices. Noting that for H=4K2H=4K_{2} or H=3K1,2H=3K_{1,2}, one has S2(H)=e(H)S_{2}(H)=e(H), using Corollary 2.3, we may assume GG contains neither H=4K2H=4K_{2} nor H=3K1,2H=3K_{1,2} as a subgraph. It is sufficient to consider graphs GG whose matching number m(G)m(G) is at most 3. In the following, we prove Theorem 1.1 for m(G)=1,2m(G)=1,2 and 3, respectively.

Lemma 3.2.

Let GG be a graph with m(G)=1m(G)=1. Then f(G)>1.5nf(G)>\frac{1.5}{n}.

Proof.

Let n=|V(G)|n=|V(G)|. Since m(G)=1m(G)=1, it is easy to check that either G=K1,k1(nk)K1G=K_{1,k-1}\cup(n-k)K_{1} for some kk, 1kn1\leq k\leq n or G=K3(n3)K1G=K_{3}\cup(n-3)K_{1}. By Lemma 2.7, the assertion holds. ∎

Let X,YV(G)X,\ Y\subset V(G), and E(X,Y)E(X,Y) (e(X,Y)e(X,Y)) be the set of edges (the number of edges) whose one endpoint is in XX and one endpoint is in YY. A pendant neighbor of GG is a vertex adjacent to a pendant vertex. Let p(G)p(G) be the number of pendant vertices of GG and q(G)q(G) be the number of pendant neighbors.

Lemma 3.3.

Let GG be a graph with m(G)=2m(G)=2. Then f(G)f(K1,n1+)f(G)\geq f(K^{+}_{1,n-1}) with equality if and only if GK1,n1+G\cong K^{+}_{1,n-1}.

Proof.

We may assume that GG is a connected graph with at least 11 vertices. First suppose that GG has no K3K_{3} as a subgraph, if GG is a tree, by Corollary 2.4, we have f(G)>1.5nf(G)>\frac{1.5}{n}. Since m(G)=2m(G)=2, GG is isomorphic to G5G_{5} in Fig. 1, using the 2-th additive compound matrix approach, we have f(G)>0.5>1.5nf(G)>0.5>\frac{1.5}{n}, details are deferred in the Appendix.

Next suppose that GG has a subgraph H=K3H=K_{3} with V(H)={u,v,w}V(H)=\{u,v,w\}.

If every edge of GG has at least one endpoint in V(H)V(H), since m(G)=2m(G)=2, we only need to consider q(G)=0,1q(G)=0,1 and 2, respectively. Then GG is isomorphic to G1(1t1n2, 0t2,t3n3)G_{1}\ (1\leq t_{1}\leq n-2,\ 0\leq t_{2},\ t_{3}\leq n-3) or G2(7t=n4)G_{2}\ (7\leq t=n-4) in Fig. 1.

If there exists an edge e=abe=ab has no endpoint in V(H)V(H), let M=V(G){a,b,u,v,w}M=V(G)-\{a,b,u,v,w\}, since m(G)=2m(G)=2, MM is an independent set and e(V(H),M)=0e(V(H),M)=0, and all vertices in MM are adjacent to one of the endpoints of ee, say aa. Hence there are no edges between bb and V(H)V(H), we only need to consider |N(a)V(H)|=1|N(a)\cap V(H)|=1 and 3, respectively (isomorphic to G2G_{2} when |N(a)V(H)|=2|N(a)\cap V(H)|=2). Then GG is isomorphic to G3(7t=n4)G_{3}\ (7\leq t=n-4) or G4(7t=n4)G_{4}\ (7\leq t=n-4) in Fig. 1.

Refer to caption
Figure 1: Possible forms of graphs GG with m(G)=2m(G)=2

When GG1G\cong G_{1}, let V(Kti¯)={v1,,vti}V(\overline{K_{t_{i}}})=\{v_{1},\ldots,v_{t_{i}}\} (i=1,2,3)(i=1,2,3), and Gt1,t2,t3G_{t_{1},t_{2},t_{3}} be a graph of order nn (n=t1+t2+t3)(n=t_{1}+t_{2}+t_{3}), which consists of three independent sets Kti¯\overline{K_{t_{i}}} (i=1,2,3)(i=1,2,3), where t1t_{1} triangles share one common edge and t2,t3t_{2},t_{3} pendant vertices at each end of the common edge. The graph Gt1,t2,t3G_{t_{1},t_{2},t_{3}} has an obvious partition G=H{G1,G2,,Gk}(k=5)G=\bigvee_{H}\{G_{1},G_{2},...,G_{k}\}(k=5) where the corresponding quotient matrix is

Qπ=(200110101000101t1t20t1+t2+11t10t31t1+t3+1).\displaystyle Q^{\pi}=\left(\begin{array}[]{ccccc}2&0&0&1&1\\ 0&1&0&1&0\\ 0&0&1&0&1\\ t_{1}&t_{2}&0&t_{1}+t_{2}+1&1\\ t_{1}&0&t_{3}&1&t_{1}+t_{3}+1\end{array}\right).

By Lemma 2.1 and direct calculation, the characteristic polynomial of Q(Gt1,t2,t3)Q(G_{t_{1},t_{2},t_{3}}) is

PQ(Gt1,t2,t3,x)=P(Qπ(Gt1,t2,t3),x)(x2)t11(x1)t2+t32.\displaystyle P_{Q}(G_{t_{1},t_{2},t_{3}},x)=P(Q^{\pi}(G_{t_{1},t_{2},t_{3}}),x)(x-2)^{t_{1}-1}(x-1)^{t_{2}+t_{3}-2}.

Without loss of generality, we may assume t2t3t_{2}\geq t_{3}. Let

P1\displaystyle P_{1} =P(Δ2(Qπ(Gt1,t2,t3)),x+e(Gt1,t2,t3)+3),\displaystyle=P(\Delta_{2}(Q^{\pi}(G_{t_{1},t_{2},t_{3}})),x+e(G_{t_{1},t_{2},t_{3}})+3),
P2\displaystyle P_{2} =P(Δ2(Qπ(Gt1,t2+1,t31)),x+e(Gt1,t2+1,t31)+3).\displaystyle=P(\Delta_{2}(Q^{\pi}(G_{t_{1},t_{2}+1,t_{3}-1})),x+e(G_{t_{1},t_{2}+1,t_{3}-1})+3).

By direct computation, P1P2P_{1}-P_{2} is always positive (details are deferred in the Appendix), which indicates the largest root of P1P_{1} less than the largest root of P2P_{2}.

Since q2>2q_{2}>2, by Lemma 2.9 (if t1=1t_{1}=1 and t3=0t_{3}=0, the previous results (see Theorem 2.4 in [2]) verify that q2>32.5n>2q_{2}>3-\frac{2.5}{n}>2), P(Qπ(Gt1,t2,t3,x))P(Q^{\pi}(G_{t_{1},t_{2},t_{3}},x)) and P(Qπ(Gt1,t2+1,t31,x))P(Q^{\pi}(G_{t_{1},t_{2}+1,t_{3}-1},x)) contain the largest two roots of PQ(Gt1,t2,t3,x)P_{Q}(G_{t_{1},t_{2},t_{3}},x) and PQ(Gt1,t2+1,t31,x)P_{Q}(G_{t_{1},t_{2}+1,t_{3}-1},x). By Corollary 2.1, we have f(Gt1,t2,t3)>f(Gt1,t2+1,t31)f(G_{t_{1},t_{2},t_{3}})>f(G_{t_{1},t_{2}+1,t_{3}-1}). By recursive relationships, we have f(G1)>f(G11)f(G_{1})>f(G_{1}^{1}).

Let Gt1,t2G_{t_{1},t_{2}} be a graph of order nn (n=t1+t2)(n=t_{1}+t_{2}), which consists of two independent sets Kti¯\overline{K_{t_{i}}} (i=1,2)(i=1,2), where t1t_{1} triangles share one common edge and t2t_{2} pendant vertices at the end of the common edge. Let

P3\displaystyle P_{3} =P(Δ2(Qπ(Gt1,t2)),x+e(Gt1,t2)+3),\displaystyle=P(\Delta_{2}(Q^{\pi}(G_{t_{1},t_{2}})),x+e(G_{t_{1},t_{2}})+3),
P4\displaystyle P_{4} =P(Δ2(Qπ(Gt11,t2+1)),x+e(Gt11,t2+1)+3).\displaystyle=P(\Delta_{2}(Q^{\pi}(G_{t_{1}-1,t_{2}+1})),x+e(G_{t_{1}-1,t_{2}+1})+3).

By using computer computations, P3P4P_{3}-P_{4} is always positive (details are deferred in the Appendix), which indicates the largest root of P3P_{3} less than the largest root of P4P_{4}.

Since q2>2q_{2}>2, by Lemma 2.9 (if t1=1t_{1}=1, the previous results (see Theorem 2.4 in [2]) verify that q2>32.5n>2q_{2}>3-\frac{2.5}{n}>2), P(Qπ(Gt1,t2,x))P(Q^{\pi}(G_{t_{1},t_{2}},x)) and P(Qπ(Gt11,t2+1,x))P(Q^{\pi}(G_{t_{1}-1,t_{2}+1},x)) contains the largest two roots of PQ(Gt1,t2,x)P_{Q}(G_{t_{1},t_{2}},x) and PQ(Gt11,t2+1,x)P_{Q}(G_{t_{1}-1,t_{2}+1},x). By Corollary 2.1, we have f(Gt1,t2)>f(Gt11,t2+1)f(G_{t_{1},t_{2}})>f(G_{t_{1}-1,t_{2}+1}). By recursive relationships, we have f(G11)>f(G12)f(G_{1}^{1})>f(G_{1}^{2}). Then f(G1)>f(G11)>f(G12)=f(K1,n1+)f(G_{1})>f(G_{1}^{1})>f(G_{1}^{2})=f(K^{+}_{1,n-1}).

When GG2G\cong G_{2}, the graph GG has an obvious partition G=H{G1,G2,,Gk}(k=4)G=\bigvee_{H}\{G_{1},G_{2},...,G_{k}\}(k=4) where the corresponding quotient matrix is

Qπ=(1100tt+22001410022).\displaystyle Q^{\pi}=\left(\begin{array}[]{cccc}1&1&0&0\\ t&t+2&2&0\\ 0&1&4&1\\ 0&0&2&2\end{array}\right).

By Lemma 2.1 and direct calculation, the characteristic polynomial of Q(G)Q(G) is

PQ(G,x)=P(Qπ,x)(x2)(x1)t1.\displaystyle P_{Q}(G,x)=P(Q^{\pi},x)(x-2)(x-1)^{t-1}.

By simple calculations we obtain

P(Qπ,t+3.251.5t+4)\displaystyle P(Q^{\pi},t+3.25-\frac{1.5}{t+4}) =0.25(4+t)4g(t)>0,\displaystyle=\frac{0.25}{(4+t)^{4}}g(t)>0,
P(Qπ,4.75)\displaystyle P(Q^{\pi},4.75) =0.3t19.98<0,\displaystyle=-0.3t-19.98<0,
P(Qπ,2)\displaystyle P(Q^{\pi},2) =4t>0,\displaystyle=4t>0,
P(Qπ,1)\displaystyle P(Q^{\pi},1) =t<0,\displaystyle=-t<0,
P(Qπ,0)\displaystyle P(Q^{\pi},0) =8>0,\displaystyle=8>0,

with g(t)=(t2+5.3t+8.9)(t+5.2)(t+4.6)(t+4.3)(t+1.3)(t6.9)g(t)=(t^{2}+5.3t+8.9)(t+5.2)(t+4.6)(t+4.3)(t+1.3)(t-6.9), which indicates q1(G)<t+3.251.5t+4q_{1}(G)<t+3.25-\frac{1.5}{t+4} and q2(G)<4.75q_{2}(G)<4.75, then S2(G)<t+81.5t+4=e(G)+31.5nS_{2}(G)<t+8-\frac{1.5}{t+4}=e(G)+3-\frac{1.5}{n}, we have f(G)>1.5nf(G)>\frac{1.5}{n}.

When GG3G\cong G_{3}, we use the quotient matrix approach, when GG4G\cong G_{4}, we use the 2-th additive compound matrix approach, details are deferred in the Appendix. This completes the proof.

For nonempty set XV(G)X\subset V(G), let N(X)={vvXandN(v)X}N(X)=\{v\mid v\not\in X\ \text{and}\ N(v)\cap X\neq\emptyset\}. Suppose that VV^{\prime}, EE^{\prime} are nonempty subsets of V(G)V(G) and E(G)E(G), the subgraphs induced by VV^{\prime} and EE^{\prime} are denoted by G[V]G[V^{\prime}] and G[E]G[E^{\prime}], respectively.

Lemma 3.4.

Let GG be a graph with m(G)=3m(G)=3. Then f(G)>1.5nf(G)>\frac{1.5}{n}.

Proof.

Without loss of generality, we may suppose that e1=a1b1e_{1}=a_{1}b_{1}, e2=a2b2e_{2}=a_{2}b_{2} and e3=a3b3e_{3}=a_{3}b_{3} are three independent edges in GG.

We first suppose that GG contains K32K2K_{3}\cup 2K_{2} as a subgraph and V(K3)={a1,b1,u}V(K_{3})=\{a_{1},b_{1},u\}. Let V1=V(K32K2)V_{1}=V(K_{3}\cup 2K_{2}), M=V(G)V1M^{\prime}=V(G)-V_{1}. Then |M|4|M^{\prime}|\geq 4. Since m(G)=3m(G)=3, MM^{\prime} is an independent set and e(V(K3),M)=0e(V(K_{3}),M^{\prime})=0. Since GG has neither 3K1,23K_{1,2} nor 4K24K_{2} as a subgraph and |M|4|M^{\prime}|\geq 4, there exists exactly one vertex a3a_{3} (w.l.o.g.) which is adjacent to vertices of MM^{\prime} and d(b3)=1d(b_{3})=1.

Let V2={a1,b1,u,a2,b2}V_{2}=\{a_{1},b_{1},u,a_{2},b_{2}\}, E1={a3w|wV2}E_{1}=\{a_{3}w|w\in V_{2}\}. Then G[E1]G[E_{1}] is a star with center a3a_{3}, and GE1G-E_{1} is a disjoint union of a star K1,|M|+1K_{1,|M^{\prime}|+1} with center a3a_{3} and a graph G[V2]G[V_{2}] containing K3+K2K_{3}+K_{2}.

If G[V2]G[V_{2}] is connected, then G[V2]G[V_{2}] is spanning subgraph of K5K_{5} which contains K3K2K_{3}\cup K_{2} as a subgraph, by direct calculation, we have q1(G[V2])<e(G[V2])0.15<e(G[V2])1.5n(n11)q_{1}(G[V_{2}])<e(G[V_{2}])-0.15<e(G[V_{2}])-\frac{1.5}{n}\ (n\geq 11). Since n(K1,|M|+1)6n(K_{1,|M^{\prime}|+1})\geq 6, so q2(G[V2])q2(K5)<q1(K1,|M|+1)q_{2}(G[V_{2}])\leq q_{2}(K_{5})<q_{1}(K_{1,|M^{\prime}|+1}). Hence S2(GE1)=q1(G[V2])+q1(K1,|M|+1)<e(G[V2])1.5n+e(K1,|M|+1)+1=e(GE1)+11.5nS_{2}(G-E_{1})=q_{1}(G[V_{2}])+q_{1}(K_{1,|M^{\prime}|+1})<e(G[V_{2}])-\frac{1.5}{n}+e(K_{1,|M^{\prime}|+1})+1=e(G-E_{1})+1-\frac{1.5}{n}. Thus by Corollary 2.2, S2(G)S2(G[E1])+S2(GE1)<e(G[E1])+2+e(GE1)+11.5n=e(G)+31.5nS_{2}(G)\leq S_{2}(G[E_{1}])+S_{2}(G-E_{1})<e(G[E_{1}])+2+e(G-E_{1})+1-\frac{1.5}{n}=e(G)+3-\frac{1.5}{n}, we have f(G)>1.5nf(G)>\frac{1.5}{n}.

If G[V2]G[V_{2}] is disconnected, then G[V2]=K3K2G[V_{2}]=K_{3}\cup K_{2}. Let t=d(a2)+d(b2)t=d(a_{2})+d(b_{2}). We have t=3,4t=3,4. It is not hard to see that m(Ge2)=2m(G-e_{2})=2 and Ge2G-e_{2} contains disjoint union of a star K1,|M|+1K_{1,|M^{\prime}|+1} with center a3a_{3} and K3K_{3}. Therefore, by Lemma 2.9, q2(Ge2)tq_{2}(G-e_{2})\geq t. Using Lemmas 3.3 and 2.10, we find that S2(G)<S2(Ge2)+1<(e(Ge2)+31.5n)+1=e(G)+31.5nS_{2}(G)<S_{2}(G-e_{2})+1<(e(G-e_{2})+3-\frac{1.5}{n})+1=e(G)+3-\frac{1.5}{n}, we have f(G)>1.5nf(G)>\frac{1.5}{n}.

Next suppose that GG has no subgraph K32K2K_{3}\cup 2K_{2}. Since m(G)=3m(G)=3, M=V(G)V({e1,e2,e3})M=V(G)-V(\{e_{1},e_{2},e_{3}\}) is an independent set and |M|5|M|\geq 5. Since GG has no 4K24K_{2} and K3+2K2K_{3}+2K_{2} as subgraph, either NM(ai)=N_{M}(a_{i})=\emptyset or NM(bi)=N_{M}(b_{i})=\emptyset for i=1,2,3i=1,2,3. Without loss of generality, we may assume that N(M){a1,a2,a3}N(M)\subseteq\{a_{1},a_{2},a_{3}\}. We consider the following three cases.

Case 1: |N(M)|=3|N(M)|=3.

We have N(M)={a1,a2,a3}N(M)=\{a_{1},a_{2},a_{3}\}. The bipartite subgraph G[E({a1,a2,a3},M)]G[E(\{a_{1},a_{2},a_{3}\},M)] has no perfect matching since GG has no 3K1,23K_{1,2}. By Hall’s Theorem, there exists a subset of {a1,a2,a3}\{a_{1},a_{2},a_{3}\} with 2 elements, say {a2,a3}\{a_{2},a_{3}\}, such that |N({a2,a3})M|=1|N(\{a_{2},a_{3}\})\cap M|=1 (we have |N({a2,a3})M|0|N(\{a_{2},a_{3}\})\cap M|\neq 0 since |N(M)|=3|N(M)|=3). This means that there exists exactly one vertex yMy\in M which is adjacent to both a2a_{2} and a3a_{3}. If d(b1)2d(b_{1})\geq 2, then we clearly find a subgraph isomorphic to 3K1,23K_{1,2} in GG, a contradiction. Therefore, d(b1)=1d(b_{1})=1. Let V1={a2,b2,a3,b3,y}V_{1}=\{a_{2},b_{2},a_{3},b_{3},y\}, E1={a1w|wV1}E_{1}=\{a_{1}w|w\in V_{1}\}. Then G[E1]G[E_{1}] is a star with center a1a_{1}, and GE1G-E_{1} is a disjoint union of a star K1,|M|K_{1,|M|} with center a1a_{1} and a graph G[V1]G[V_{1}] containing P5P_{5}. Since G has no K32K2K_{3}\cup 2K_{2}, two edges a2a3a_{2}a_{3} and b2b3b_{2}b_{3} can not exist at the same time, therefore G[V1]G[V_{1}] is isomorphic to graphs in Fig. 2.

Refer to caption
Figure 2: Possible forms of graphs containing a P5P_{5}

By direct calculation, we have q1(G[V1])<e(G[V1])1.5n(n11)q_{1}(G[V_{1}])<e(G[V_{1}])-\frac{1.5}{n}\ (n\geq 11). Since n(K1,|M|)6n(K_{1,|M|})\geq 6, so q2(G[V1])q1(K1,|M|)q_{2}(G[V_{1}])\leq q_{1}(K_{1,|M|}). Hence S2(GE1)=q1(G[V1])+q1(K1,|M|)<e(G[V1])1.5n+e(K1,|M|)+1=e(GE1)+11.5nS_{2}(G-E_{1})=q_{1}(G[V_{1}])+q_{1}(K_{1,|M|})<e(G[V_{1}])-\frac{1.5}{n}+e(K_{1,|M|})+1=e(G-E_{1})+1-\frac{1.5}{n}. Thus by Corollary 2.2, S2(G)S2(G[E1])+S2(GE1)<e(G[E1])+2+e(GE1)+11.5n=e(G)+31.5nS_{2}(G)\leq S_{2}(G[E_{1}])+S_{2}(G-E_{1})<e(G[E_{1}])+2+e(G-E_{1})+1-\frac{1.5}{n}=e(G)+3-\frac{1.5}{n}, we have f(G)>1.5nf(G)>\frac{1.5}{n}.

Case 2: |N(M)|=2|N(M)|=2.

Without loss of generality, suppose that N(M)={a1,a2}N(M)=\{a_{1},a_{2}\}. Since m(G)=3m(G)=3, b1b_{1} is not adjacent to b2b_{2}. If b1b_{1} is adjacent to a3a_{3} or b3b_{3}, then changing the role of e1e_{1}, e2e_{2}, e3e_{3} by three independent edges {a1,z}\{a_{1},z\}, e2e_{2}, e3e_{3} for some vertex zMN(a1)z\in M\cap N(a_{1}), we have Case 1. Therefore, we may assume that b1b_{1}, and similarly b2b_{2}, is adjacent to none of the vertices a3a_{3} and b3b_{3}. Hence we only need to consider |N({a3,b3})|=1|N(\{a_{3},b_{3}\})|=1 or 2, respectively. Consider the existence of the edge a1a2a_{1}a_{2}, possible forms of graph GG are in Fig. 3.

Refer to caption
Figure 3: Possible forms of graphs GG with m(G)=3m(G)=3 and |N(M)|=2|N(M)|=2

When GGiG\cong G_{i}, i=6,7,8,9,10,12,13i=6,7,8,9,10,12,13, let t=d(a3)+d(b3)t=d(a_{3})+d(b_{3}), we have t{3,4,5}t\in\{3,4,5\}.

We first present several special cases: t1=t2=1t_{1}=t_{2}=1 when GG6G\cong G_{6}; t1=0t_{1}=0 and t2=2t_{2}=2; t1=t2=1t_{1}=t_{2}=1; t1=2t_{1}=2 and t2=0t_{2}=0 when GG8,G10,G13G\cong G_{8},G_{10},G_{13} (were checked by computer computations).

Except these special cases, it is not hard to see that m(Ge3)=2m(G-e_{3})=2 and Ge3G-e_{3} contains two disjoint stars K1,t1K_{1,t-1} with centers a1a_{1} and a2a_{2} Therefore, by Lemma 2.9, q2(Ge3)tq_{2}(G-e_{3})\geq t. Using Lemmas 3.3 and 2.10, we find that S2(G)<S2(Ge3)+1<(e(Ge3)+31.5n)+1=e(G)+31.5nS_{2}(G)<S_{2}(G-e_{3})+1<(e(G-e_{3})+3-\frac{1.5}{n})+1=e(G)+3-\frac{1.5}{n}, we have f(G)>1.5nf(G)>\frac{1.5}{n}.

When GG11G\cong G_{11}, we use the 2-th additive compound matrix approach, by using computer computations and recursive relationships, we have f(G11)>f(G111)f(G_{11})>f(G_{11}^{1}) and f(G111)>f(G112)f(G_{11}^{1})>f(G_{11}^{2}). Then f(G11)>f(G111)>f(G112)=f(G2)f(G_{11})>f(G_{11}^{1})>f(G_{11}^{2})=f(G_{2}), we have f(G)>1.5nf(G)>\frac{1.5}{n}.

When GG14G\cong G_{14}, we use the 2-th additive compound matrix approach, by using computer computations and recursive relationships, we have f(G14)>f(G141)f(G_{14})>f(G_{14}^{1}) and f(G141)>f(G142)f(G_{14}^{1})>f(G_{14}^{2}). Then f(G14)>f(G141)>f(G142)=f(G4)f(G_{14})>f(G_{14}^{1})>f(G_{14}^{2})=f(G_{4}), we have f(G)>1.5nf(G)>\frac{1.5}{n}.

Case 3: |N(M)|=1|N(M)|=1.

Without loss of generality, suppose that N(M)={a1}N(M)=\{a_{1}\}. If d(b1)2d(b_{1})\geq 2, then we clearly find three independent edges e1e^{\prime}_{1}, e2e^{\prime}_{2}, e3e^{\prime}_{3} in GG such that the set M=V(G)V({e1,e2,e3}M^{\prime}=V(G)-V(\{e^{\prime}_{1},e^{\prime}_{2},e^{\prime}_{3}\} is an independent set and |N(M)|2|N(M^{\prime})|\geq 2 which is dealt with as the previous cases. Therefore, d(b1)=1d(b_{1})=1. Let V1={a2,b2,a3,b3}V_{1}=\{a_{2},b_{2},a_{3},b_{3}\}, E1={a1w|wV1}E_{1}=\{a_{1}w|w\in V_{1}\}. Then G[E1]G[E_{1}] is a star with center a1a_{1}, and GE1G-E_{1} is a disjoint union of a star K1,|M|+1K_{1,|M|+1} with center a1a_{1} and a graph G[V1]G[V_{1}] containing 2K22K_{2}. Therefore G[V1]G[V_{1}] is isomorphic to graphs in Fig. 4.

Refer to caption
Figure 4: Possible forms of graphs containing 2K22K_{2}

By direct calculation, we have q1(G[V1])<e(G[V1])+11.5n(n11)q_{1}(G[V_{1}])<e(G[V_{1}])+1-\frac{1.5}{n}\ (n\geq 11), hence S2(GE1)=q1(G[V1])+q1(K1,|M|+1)<e(G[V1])+11.5n+e(K1,|M|)+1=e(GE1)+21.5nS_{2}(G-E_{1})=q_{1}(G[V_{1}])+q_{1}(K_{1,|M|+1})<e(G[V_{1}])+1-\frac{1.5}{n}+e(K_{1,|M|})+1=e(G-E_{1})+2-\frac{1.5}{n}. If |N(a1)V1|=1|N(a_{1})\cap V_{1}|=1, which means |E1|=1|E_{1}|=1, by Corollary 2.2, S2(G)S2(G[E1])+S2(GE1)<e(G[E1])+1+e(GE1)+21.5n=e(G)+31.5nS_{2}(G)\leq S_{2}(G[E_{1}])+S_{2}(G-E_{1})<e(G[E_{1}])+1+e(G-E_{1})+2-\frac{1.5}{n}=e(G)+3-\frac{1.5}{n}, we have f(G)>1.5nf(G)>\frac{1.5}{n}. Hence we only need to consider |N(a1)|V1|=2|N(a_{1})|\cap V_{1}|=2, 3 and 4, respectively. Possible forms of graph GG are in Fig. 5 (If GG is a tree graph, the assertion holds by Corollary 2.4).

Refer to caption
Figure 5: Possible forms of graphs GG with m(G)=3m(G)=3 and |N(M)|=1|N(M)|=1

When GGiG\cong G_{i}, i=17,24,25i=17,24,25, let t=d(a3)+d(b3)t=d(a_{3})+d(b_{3}) (pick a2a_{2} and b2b_{2} when GG25G\cong G_{25}), we have t{3,4}t\in\{3,4\}. It is not hard to see that m(Ge3)=2m(G-e_{3})=2 and Ge3G-e_{3} contains two disjoint stars K1,t1K_{1,t-1} with centers a1a_{1} and b2b_{2} (contains disjoint union of a star K1,t1K_{1,t-1} with center a1a_{1} and a K3K_{3} when GG25G\cong G_{25}). Therefore, by Lemma 2.9, q2(Ge3)tq_{2}(G-e_{3})\geq t. Using Lemmas 3.3 and 2.10, we find that S2(G)<S2(Ge3)+1<(e(Ge3)+31.5n)+1=e(G)+31.5nS_{2}(G)<S_{2}(G-e_{3})+1<(e(G-e_{3})+3-\frac{1.5}{n})+1=e(G)+3-\frac{1.5}{n}, we have f(G)>1.5nf(G)>\frac{1.5}{n}.

When GG is isomorphic to other graphs in Fig. 5, we use the quotient matrix approach, details are deferred in the Appendix. This completes the proof. ∎

Proof of Theorem 1.1: The result follows immediately from Corollary 2.3, Lemmas 3.1, 3.2, 3.3 and 3.4.

4 Declaration of competing interest

There is no competing interest.

Appendix

The details of polynomial P1P2P_{1}-P_{2} and P3P4P_{3}-P_{4} can be obtained from:

https://github.com/ZZM0329/Polynomial/tree/main/G1

Table 1: 2-th additive compound matrix approach
No. Corresponding quotient matrix QπQ^{\pi} Characteristic polynomial PQ(G,x)P_{Q}(G,x)
G4G_{4} (110tt+33015)\left(\begin{array}[]{ccc}1&1&0\\ t&t+3&3\\ 0&1&5\end{array}\right) P(Qπ,x)(x2)2(x1)t1P(Q^{\pi},x)(x-2)^{2}(x-1)^{t-1}
G5G_{5} (200110101000101t1t20t1+t20t10t30t1+t3)\left(\begin{array}[]{ccccc}2&0&0&1&1\\ 0&1&0&1&0\\ 0&0&1&0&1\\ t_{1}&t_{2}&0&t_{1}+t_{2}&0\\ t_{1}&0&t_{3}&0&t_{1}+t_{3}\end{array}\right) P(Qπ,x)(x2)t11(x1)t2+t32P(Q^{\pi},x)(x-2)^{t_{1}-1}(x-1)^{t_{2}+t_{3}-2}
G11G_{11} (200110010100001010t1t20t1+t2+202t10t30t1+t3+22000114)\left(\begin{array}[]{cccccc}2&0&0&1&1&0\\ 0&1&0&1&0&0\\ 0&0&1&0&1&0\\ t_{1}&t_{2}&0&t_{1}+t_{2}+2&0&2\\ t_{1}&0&t_{3}&0&t_{1}+t_{3}+2&2\\ 0&0&0&1&1&4\end{array}\right) P(Qπ,x)(x2)t1(x1)t2+t32P(Q^{\pi},x)(x-2)^{t_{1}}(x-1)^{t_{2}+t_{3}-2}
G14G_{14} (200110010100001010t1t20t1+t2+312t10t31t1+t3+32000114)\left(\begin{array}[]{cccccc}2&0&0&1&1&0\\ 0&1&0&1&0&0\\ 0&0&1&0&1&0\\ t_{1}&t_{2}&0&t_{1}+t_{2}+3&1&2\\ t_{1}&0&t_{3}&1&t_{1}+t_{3}+3&2\\ 0&0&0&1&1&4\end{array}\right) P(Qπ,x)(x2)t1(x1)t2+t32P(Q^{\pi},x)(x-2)^{t_{1}}(x-1)^{t_{2}+t_{3}-2}
Table 2: Quotient matrix approach
No. Corresponding quotient matrix QπQ^{\pi} Characteristic polynomial PQ(G,x)P_{Q}(G,x)
G3G_{3} (1100tt+11001320013)\left(\begin{array}[]{cccc}1&1&0&0\\ t&t+1&1&0\\ 0&1&3&2\\ 0&0&1&3\end{array}\right) [x4(t+8)x3+(6t+19)x2(7t+16)x+4](x1)t[x^{4}-(t+8)x^{3}+(6t+19)x^{2}-(7t+16)x+4](x-1)^{t}
G15G_{15} (11000tt+3201013000001101012)\left(\begin{array}[]{ccccc}1&1&0&0&0\\ t&t+3&2&0&1\\ 0&1&3&0&0\\ 0&0&0&1&1\\ 0&1&0&1&2\end{array}\right) [x5(t+10)+x4+(6t+34)x3(10t+48)x2+(3t+27)x4](x1)t[x^{5}-(t+10)+x^{4}+(6t+34)x^{3}-(10t+48)x^{2}+(3t+27)x-4](x-1)^{t}
G16G_{16} (1100tt+42201300103)\left(\begin{array}[]{cccc}1&1&0&0\\ t&t+4&2&2\\ 0&1&3&0\\ 0&1&0&3\end{array}\right) [x3(t+8)x2+(3t+15)x8](x3)(x1)t+1[x^{3}-(t+8)x^{2}+(3t+15)x-8](x-3)(x-1)^{t+1}
G18G_{18} (110000tt+21010012100001210010131000011)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+2&1&0&1&0\\ 0&1&2&1&0&0\\ 0&0&1&2&1&0\\ 0&1&0&1&3&1\\ 0&0&0&0&1&1\end{array}\right) [x4(t+10)x3+(7t+34)x2(13t+46)x+4t+20](x1)tx[x^{4}-(t+10)x^{3}+(7t+34)x^{2}-(13t+46)x+4t+20](x-1)^{t}x
G19G_{19} (110000tt+21001012100001210000121010012)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+2&1&0&0&1\\ 0&1&2&1&0&0\\ 0&0&1&2&1&0\\ 0&0&0&1&2&1\\ 0&1&0&0&1&2\end{array}\right) [x4(t+8)x3+(5t+20)x2(5t+17)x+4](x23x+1)(x1)t1[x^{4}-(t+8)x^{3}+(5t+20)x^{2}-(5t+17)x+4](x^{2}-3x+1)(x-1)^{t-1}
G20G_{20} (110000tt+20110001100011310010131000011)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+2&0&1&1&0\\ 0&0&1&1&0&0\\ 0&1&1&3&1&0\\ 0&1&0&1&3&1\\ 0&0&0&0&1&1\end{array}\right) [x4(t+8)x3+(5t+18)x2(3t+15)x+4](x23x+1)(x1)t1[x^{4}-(t+8)x^{3}+(5t+18)x^{2}-(3t+15)x+4](x^{2}-3x+1)(x-1)^{t-1}
G21G_{21} (110000tt+31110012100011310010131000011)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+3&1&1&1&0\\ 0&1&2&1&0&0\\ 0&1&1&3&1&0\\ 0&1&0&1&3&1\\ 0&0&0&0&1&1\end{array}\right) [x6(t+13)x5+(9t+62)x4(26t+140)x3+(27t+158)x2(8t+84)x+16](x1)t1[x^{6}-(t+13)x^{5}+(9t+62)x^{4}-(26t+140)x^{3}+(27t+158)x^{2}-(8t+84)x+16](x-1)^{t-1}
G22G_{22} (110000tt+31101012100011310000121010012)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+3&1&1&0&1\\ 0&1&2&1&0&0\\ 0&1&1&3&1&0\\ 0&0&0&1&2&1\\ 0&1&0&0&1&2\end{array}\right) [x6(t+13)x5+(9t+63)x4(27t+145)x3+(31t+165)x2(11t+87)x+16](x1)t1[x^{6}-(t+13)x^{5}+(9t+63)x^{4}-(27t+145)x^{3}+(31t+165)x^{2}-(11t+87)x+16](x-1)^{t-1}
G23G_{23} (110000tt+41111012100011310010131010012)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+4&1&1&1&1\\ 0&1&2&1&0&0\\ 0&1&1&3&1&0\\ 0&1&0&1&3&1\\ 0&1&0&0&1&2\end{array}\right) [x4(t+11)x3+(6t+37)x2(7t+47)x+20](x3)(x1)t[x^{4}-(t+11)x^{3}+(6t+37)x^{2}-(7t+47)x+20](x-3)(x-1)^{t}
G26G_{26} (110000tt+21001012100001311000121010113)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+2&1&0&0&1\\ 0&1&2&1&0&0\\ 0&0&1&3&1&1\\ 0&0&0&1&2&1\\ 0&1&0&1&1&3\end{array}\right) [x5(t+12)x4+(9t+51)x3(24t+94)x2+(19t+71)x16](x1)t[x^{5}-(t+12)x^{4}+(9t+51)x^{3}-(24t+94)x^{2}+(19t+71)x-16](x-1)^{t}
G27G_{27} (110000tt+20101001100011411000121010113)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+2&0&1&0&1\\ 0&0&1&1&0&0\\ 0&1&1&4&1&1\\ 0&0&0&1&2&1\\ 0&1&0&1&1&3\end{array}\right) [x6(t+13)x5+(10t+61)x4(31t+135)x3+(35t+150)x2(12t+80)x+16](x1)t1[x^{6}-(t+13)x^{5}+(10t+61)x^{4}-(31t+135)x^{3}+(35t+150)x^{2}-(12t+80)x+16](x-1)^{t-1}
G28G_{28} (110000tt+31101012100011411000121010113)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+3&1&1&0&1\\ 0&1&2&1&0&0\\ 0&1&1&4&1&1\\ 0&0&0&1&2&1\\ 0&1&0&1&1&3\end{array}\right) [x6(t+15)x5+(11t+84)x4(40t+228)x3+(58t+319)x2(29t+221)x+60](x1)t1[x^{6}-(t+15)x^{5}+(11t+84)x^{4}-(40t+228)x^{3}+(58t+319)x^{2}-(29t+221)x+60](x-1)^{t-1}
G29G_{29} (11000tt+3102012100013201014)\left(\begin{array}[]{ccccc}1&1&0&0&0\\ t&t+3&1&0&2\\ 0&1&2&1&0\\ 0&0&1&3&2\\ 0&1&0&1&4\end{array}\right) [x5(t+13)x4+(9t+59)x3(23t+115)x2+(16t+92)x24](x2)(x1)t1[x^{5}-(t+13)x^{4}+(9t+59)x^{3}-(23t+115)x^{2}+(16t+92)x-24](x-2)(x-1)^{t-1}
G30G_{30} (11000tt+3012001100114201014)\left(\begin{array}[]{ccccc}1&1&0&0&0\\ t&t+3&0&1&2\\ 0&0&1&1&0\\ 0&1&1&4&2\\ 0&1&0&1&4\end{array}\right) [x5(t+13)x4+(9t+57)x3(21t+107)x2+(10t+86)x24](x2)(x1)t1[x^{5}-(t+13)x^{4}+(9t+57)x^{3}-(21t+107)x^{2}+(10t+86)x-24](x-2)(x-1)^{t-1}
G31G_{31} (11000tt+4112012100114201014)\left(\begin{array}[]{ccccc}1&1&0&0&0\\ t&t+4&1&1&2\\ 0&1&2&1&0\\ 0&1&1&4&2\\ 0&1&0&1&4\end{array}\right) [x4(t+12)x3+(7t+43)x2(8t+56)x+24](x3)(x2)(x1)t1[x^{4}-(t+12)x^{3}+(7t+43)x^{2}-(8t+56)x+24](x-3)(x-2)(x-1)^{t-1}
G32G_{32} (110000tt+21100013101011310000121001012)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+2&1&1&0&0\\ 0&1&3&1&0&1\\ 0&1&1&3&1&0\\ 0&0&0&1&2&1\\ 0&0&1&0&1&2\end{array}\right) [x4(t+10)x3+(7t+32)x2(11t+39)x+16](x23x+1)(x1)t1[x^{4}-(t+10)x^{3}+(7t+32)x^{2}-(11t+39)x+16](x^{2}-3x+1)(x-1)^{t-1}
G33G_{33} (1100tt+20200220123)\left(\begin{array}[]{cccc}1&1&0&0\\ t&t+2&0&2\\ 0&0&2&2\\ 0&1&2&3\end{array}\right) [x3(t+8)x2+(5t+17)x2t10](x3)(x2)(x1)t1x[x^{3}-(t+8)x^{2}+(5t+17)x-2t-10](x-3)(x-2)(x-1)^{t-1}x
G34G_{34} (11000tt+3120013200113100022)\left(\begin{array}[]{ccccc}1&1&0&0&0\\ t&t+3&1&2&0\\ 0&1&3&2&0\\ 0&1&1&3&1\\ 0&0&0&2&2\end{array}\right) [x5(t+12)x4+(8t+49)x3(17t+86)x2+(8t+64)x16](x3)(x1)t1[x^{5}-(t+12)x^{4}+(8t+49)x^{3}-(17t+86)x^{2}+(8t+64)x-16](x-3)(x-1)^{t-1}
G35G_{35} (110tt+44015)\left(\begin{array}[]{ccc}1&1&0\\ t&t+4&4\\ 0&1&5\end{array}\right) [x3(t+10)x2+(5t+25)x16](x3)2(x1)t[x^{3}-(t+10)x^{2}+(5t+25)x-16](x-3)^{2}(x-1)^{t}
G36G_{36} (110000tt+21100014111011310001131001012)\left(\begin{array}[]{cccccc}1&1&0&0&0&0\\ t&t+2&1&1&0&0\\ 0&1&4&1&1&1\\ 0&1&1&3&1&0\\ 0&0&1&1&3&1\\ 0&0&1&0&1&2\end{array}\right) [x6(t+15)x5+(12t+84)x4(48t+228)x3+(77t+319)x2(41t+221)x+60](x1)t1[x^{6}-(t+15)x^{5}+(12t+84)x^{4}-(48t+228)x^{3}+(77t+319)x^{2}-(41t+221)x+60](x-1)^{t-1}
G37G_{37} (1100tt+20200420123)\left(\begin{array}[]{cccc}1&1&0&0\\ t&t+2&0&2\\ 0&0&4&2\\ 0&1&2&3\end{array}\right) [x4(t+10)x3+(7t+29)x2(8t+28)x+8](x3)(x2)(x1)t1[x^{4}-(t+10)x^{3}+(7t+29)x^{2}-(8t+28)x+8](x-3)(x-2)(x-1)^{t-1}
G38G_{38} (1100tt+22001520022)\left(\begin{array}[]{cccc}1&1&0&0\\ t&t+2&2&0\\ 0&1&5&2\\ 0&0&2&2\end{array}\right) [x3(t+9)x2+(6t+18)x8](x3)(x2)(x1)t[x^{3}-(t+9)x^{2}+(6t+18)x-8](x-3)(x-2)(x-1)^{t}
G39G_{39} (11000tt+3120014210113100123)\left(\begin{array}[]{ccccc}1&1&0&0&0\\ t&t+3&1&2&0\\ 0&1&4&2&1\\ 0&1&1&3&1\\ 0&0&1&2&3\end{array}\right) [x5(t+14)x4+(10t+68)x3(28t+146)x2+(23t+139)x48](x3)(x1)t1[x^{5}-(t+14)x^{4}+(10t+68)x^{3}-(28t+146)x^{2}+(23t+139)x-48](x-3)(x-1)^{t-1}
G40G_{40} (11000tt+3210015110123000202)\left(\begin{array}[]{ccccc}1&1&0&0&0\\ t&t+3&2&1&0\\ 0&1&5&1&1\\ 0&1&2&3&0\\ 0&0&2&0&2\end{array}\right) [x5(t+14)x4+(10t+67)x3(27t+142)x2+(20t+136)x48](x3)(x1)t1[x^{5}-(t+14)x^{4}+(10t+67)x^{3}-(27t+142)x^{2}+(20t+136)x-48](x-3)(x-1)^{t-1}
G41G_{41} (1100tt+42201520123)\left(\begin{array}[]{cccc}1&1&0&0\\ t&t+4&2&2\\ 0&1&5&2\\ 0&1&2&3\end{array}\right) [x4(t+13)x3+(8t+51)x2(11t+75)x+36](x3)2(x1)t1[x^{4}-(t+13)x^{3}+(8t+51)x^{2}-(11t+75)x+36](x-3)^{2}(x-1)^{t-1}
G42G_{42} (1100tt+22001520024)\left(\begin{array}[]{cccc}1&1&0&0\\ t&t+2&2&0\\ 0&1&5&2\\ 0&0&2&4\end{array}\right) [x4(t+12)x3+(9t+43)x2(16t+56)x+24](x3)(x2)(x1)t1[x^{4}-(t+12)x^{3}+(9t+43)x^{2}-(16t+56)x+24](x-3)(x-2)(x-1)^{t-1}
G43G_{43} (11000tt+3210015110124100213)\left(\begin{array}[]{ccccc}1&1&0&0&0\\ t&t+3&2&1&0\\ 0&1&5&1&1\\ 0&1&2&4&1\\ 0&0&2&1&3\end{array}\right) [x4(t+13)x3+(9t+51)x2(15t+75)x+36](x3)2(x1)t1[x^{4}-(t+13)x^{3}+(9t+51)x^{2}-(15t+75)x+36](x-3)^{2}(x-1)^{t-1}
G44G_{44} (110tt+44017)\left(\begin{array}[]{ccc}1&1&0\\ t&t+4&4\\ 0&1&7\end{array}\right) [x3(t+12)x2+(7t+35)x24](x3)3(x1)t1[x^{3}-(t+12)x^{2}+(7t+35)x-24](x-3)^{3}(x-1)^{t-1}

References

  • [1] B. Amaro, L.S. Lima, C.S. Oliveira, C. Lavor, N. Abreu, A note on the sum of the largest signless Laplacian eigenvalues, Electronic Notes in Discrete Mathematics., 54(2016), 175–180.
  • [2] M. Aouchiche, P. Hansen, C. Lucas, On the extremal values of the second largest Q- eigenvalue, Linear Algebra Appl., 435(2011), 2591–2606.
  • [3] F. Ashraf, G.R. Omidi, B. Tayfeh-Rezaibe, On the sum of signless Laplacian eigenvalues of graph, Linear Algebra Appl., 438(2013), 4539–4546.
  • [4] A.E. Brouwer, W.H. Haemers, Spectra of graphs, Springer, New York, 2012.
  • [5] J.N. Cooper, Constraints on Brouwer’s Laplacian spectrum conjecture, Linear Algebra Appl., 615 (2021) 11–27.
  • [6] D. Cvetković, P. Rowlinson, S. Simić, An Introduction to the Theory of Graph Spectra, Cambridge University Press, Cambridge, 2010.
  • [7] Z. Du, B. Zhou, Upper bounds for the sum of Laplacian eigenvalues of graphs, Linear Algebra Appl., 436 (2012) 3672–3683.
  • [8] K. Fan, On a theorem of wely concerning eigenvalues of linear transformations, Proc. Natl. Acad. Sci., 35(1949), 652–655.
  • [9] M. Fiedler, Special Matrices and their Applications in Numerical Mathematics, Martinus Nijhoff, Dordrecht, 1986.
  • [10] E. Fritscher, C. Hoppen, I. Rocha, V. Trevisan, On the sum of the Laplacian eigenvalues of a tree, Linear Algebra and its Applications., 435(2011), 371–399.
  • [11] W.H. Haemers, A. Mohammadian, B. Tayfeh-Rezaie, On the sum of Laplacian eigenvalues of graphs, Linear Algebra Appl., 432(2010), 2214–2221.
  • [12] W. Li, J. Guo, On the full Brouwer’s Laplacian spectrum conjecture, Discrete Mathematics., 345(2022)113078.
  • [13] Mayank, On Variants of the Grone-Merris Conjecture, Master’s thesis, Eindhoven University of Technology, 2010.
  • [14] C.S. Oliveira, L.S. Lima, P. Rama, P. Carvalho, Extremal graphs for the sum of the two largest signless Laplacian eigenvalues, Electronic Journal of Linear Algebra., 30(2015), 605–612.
  • [15] S. Wang, Y. Huang, B. Liu, On a conjecture for the sum of Laplacian eigenvalues, Math. Comput. Model., 56 (2012) 60–68.
  • [16] B. Wu, Y. Lou, C. He, Signless Laplacian and normalized Laplacian on the h-join operation of graphs, Discrete Mathematics Algorithms Appl., 6(2014)1450046.
  • [17] Y. Zheng, A note on the sum of the two largest signless Laplacian eigenvalues, Ars Combin., 148(2020), 183–191.