This paper characterizes the unique connected graph with the minimum least signless Laplacian eigenvalue among complements of bicyclic graphs using graft transformations.
Contribution
It introduces two graft transformations and applies them to identify the graph with the minimal eigenvalue in this class.
Findings
01
Identifies the unique connected graph with minimum least signless Laplacian eigenvalue among complements of bicyclic graphs.
02
Develops graft transformations as tools for spectral graph analysis.
03
Provides a characterization that can guide future spectral graph theory research.
Abstract
Suppose that G is a connected simple graph with the vertex set V(G)={v1,v2,⋯,vn}. Then the adjacency matrix of G is A(G)=(aij)n×n, where aij=1 if vi is adjacent to vj, and otherwise aij=0. The degree matrix D(G)=diag(dG(v1),dG(v2),…,dG(vn)), where dG(vi) denotes the degree of vi in the graph G (1≤i≤n). The matrix Q(G)=D(G)+A(G) is called the signless Laplacian matrix of G. The least eigenvalue of Q(G) is also called the least signless Laplacian eigenvalue of G. In this paper we give two graft transformations and then use them to characterize the unique connected graph whose least signless Laplacian eigenvalue is minimum among the complements of all bicyclic graphs.
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsGraph theory and applications · Synthesis and Properties of Aromatic Compounds · Metal-Organic Frameworks: Synthesis and Applications
Full text
The least signless Laplacian eigenvalue of the complements of bicyclic graphs ††thanks: This work is supported by NSFC (No. 11461071).
School of Mathematical Sciences, Xinjiang Normal University,
Urumqi, Xinjiang 830054, P.R.China
Abstract.
Suppose that G is a connected simple graph with the vertex set V(G)={v1,v2,⋯,vn}.
Then the adjacency matrix of G is A(G)=(aij)n×n,
where aij=1 if vi is adjacent to vj, and otherwise aij=0.
The degree matrix D(G)=diag(dG(v1),dG(v2),…,dG(vn)),
where dG(vi) denotes the degree of vi in the graph G (1≤i≤n).
The matrix Q(G)=D(G)+A(G) is called the signless Laplacian matrix of G.
The least eigenvalue of Q(G) is also called the least signless Laplacian eigenvalue of G.
In this paper we give two graft transformations and then use them to characterize the unique connected graph
whose least signless Laplacian eigenvalue is minimum among the complements of all bicyclic graphs.
Key words:
The graft transformation; The least signless Laplacian eigenvalue; Bicyclic graph; Complement.
CLC number:
O 157.5
1. Introduction
Suppose that G is a connected simple graph with the vertex set V(G)={v1,v2,⋯,vn}.
Then the adjacency matrix of G is A(G)=(aij)n×n,
where aij=1 if vi is adjacent to vj, and otherwise aij=0.
The degree matrix D(G)=diag(dG(v1),dG(v2),…,dG(vn)),
where dG(vi) denotes the degree of vi in the graph G (1≤i≤n).
The matrix Q(G)=D(G)+A(G) is called the signless Laplacian matrix of G.
Since Q(G) is positive semidefinite,
its eigenvalues can be arranged as λ1(G)≥λ2(G)≥⋯≥λn(G)≥0,
where λn(G) is also called the least signless Laplacian eigenvalue of G, denoted by λ(G).
The least signless Laplacian eigenvalues of connected graphs have been studied extensively.
Cardoso, Cvetković, Rowlinson and Simić [1] determined the unique graph
whose least signless Laplacian eigenvalue attains the minimum among all connected non-bipartite graphs.
Guo, Chen and Yu [4] obtained a lower bound on the least signless Laplacian eigenvalue of a graph.
Guo and Zhang [5] described the unique graph whose least signless Laplacian eigenvalue attains the minimum
among all non-bipartite connected graphs with fixed maximum degree.
Fan, Wang and Guo [2] determined the graph whose least signless Laplacian eigenvalue
attains the minimum or maximum among all connected non-bipartite graphs with fixed order and given number of pendant vertices.
Guo, Ren and Shi[3] determined the graph whose the least signless Laplacian eigenvalue
attains the maximum among all connected unicyclic graphs.
He and Zhou [6] gave a sharp upper bound on the least signless Laplacian eigenvalue of a graph using domination number.
Wang and Fan [8] determined the graph whose the least signless Laplacian eigenvalue is minimum.
Yu, Fan and Wang [12] determined the unique graph
whose least signless Laplacian eigenvalue attains the minimum among all connected non-bipartite bicyclic graphs.
Yu, Guo and Xu [10] determined the unique graph whose least signless Laplacian eigenvalue attains the minimum
among all connected non-bipartite graphs with given matching number and edge cover number, respectively.
Wen, Zhao and Liu [9] determined the graph which has the minimum the least signless Laplacian eigenvalue
among all non-bipartite graphs with given stability number and covering number, respectively.
Suppose that G=(V(G),E(G)) is a connected simple graph.
The complement of G is denoted by Gc=(V(Gc),E(Gc)),
where V(Gc)=V(G) and E(Gc)={xy:x,y∈V(G),xy∈/E(G)}.
If ∣E(G)∣=∣V(G)∣+1 then G is bicyclic graph.
Denote by K1,n−1 the star graph on n vertices,
and by k1,n−1+2e the graph which is obtained from k1,n−1 by connecting two pairs of different pendant vertices.
The complement (k1,n−1+2e)c of k1,n−1+2e contains an isolated vertex,
and so it is not a connected graph.
We also note that the complement of arbitrary bicyclic graph is non-bipartite graph if its order is greater or equal 12.
Therefore, we will only consider the complements of those connected bicyclic graphs on n≥12 vertices except k1,n−1+2e.
Li and Wang [7] determined the unique graph whose least signless Laplacian eigenvalue attains the minimum
in the set of the complements of all trees except K1,n−1.
Yu, Fan and Ye [11] obtained the unique graph whose least signless Laplacian eigenvalue
attains the minimum in the set of the complements of all unicyclic graphs except K1,n−1+e.
In this paper we will give two graft transformations and then use them to characterize the unique graph
whose least signless Laplacian eigenvalue is minimum
among the complements of all bicyclic graphs on n≥12 vertices except k1,n−1+2e.
2. Main results
Suppose that G is a graph with the vertex set V(G)={v1,v2,…,vn}.
Let X=(X1,X2,…,Xn)T be an unit vector such that Xi=X(vi) (1≤i≤n).
Then we have
[TABLE]
[TABLE]
The equality (2.2) holds if and only if X is the eigenvector of Q(G) corresponding to λ(G).
In what follows the vertex sets of all graphs on n vertices are write as
{v1,v2,…,vn}, and let X=(X1,X2,…,Xn)T be an unit vector such that Xi=X(vi) (1≤i≤n)
and X1≥X2≥⋯≥0≥⋯≥Xn.
Lemma 2.1. [7]. Let X be as above with X1>0 and Xn<0.
Then for any 1≤i,j≤n,
(Xi+Xj)2≤max{(Xi+X1)2,(Xi+Xn)2} and (Xi+Xj)2≤max{(Xj+X1)2,(Xj+Xn)2}.
Let Cnc(n≥12) be the set of all connected graphs
each of which is the complement of a connected bicyclic graph on n vertices.
Then, for any G∈Cnc, λ(G)>0.
Suppose that vi and vj are two vertices of a graph G.
Then the distancedG(vi,vj) between vi and vj is the length of the shortest path between vi and vj.
Define a b-graph to be a graph consisting of two vertex-disjoint
cycles and a path joining them having only its
end-vertices in common with the cycles. Define a
∞-graph to be a graph consisting of two cycles
with exactly one vertex in common. Define a
θ-graph to be a graph consisting of two basic cycles with at least two vertices in common.
Obviously, a bicyclic graph is one of b-graph, ∞-graph and θ-graph with trees attached.
Let the graphs Gk(p,q) (1≤k≤12) be shown in Fig. 1.
Suppose that G is a graph on n vertices and
that v1vl1vl2⋯vltvn is the shortest path between v1 and vn.
Let X=(X1,X2,…,Xn)T satisfy X1>0 and Xn<0.
Delete vl1vl2, and add the edge v1vl2 if (X1+Xl2)2≥(Xn+Xl1)2,
and otherwise add the edge vnvl1.
The above process is called T1-transformation of G.
Next we always denote by GT1 the result graph which is obtained after some steps of T1-transformation for G.
Now we use T1-transformation to prove the below result.
Lemma 2.2.Let H′={Gk(p,q)∣1≤k≤12}.
If Gc∈Cnc then either GT1∈H′ or dGT1(v1,vn)≤2.
Proof. If G∈H′ or dG(v1,vn)≤2 then it is finished.
So we next assume G∈/H′ and dG(v1,vn)>2.
Now we distinguish four cases to discuss.
Case 1.v1 and vn are on the same cycle of G.
Making some steps of T1-transformation for G we can see that dGT1(v1,vn)≤dG(v1,vn)−1.
Hence we can determine that the case is true.
Case 2.v1 and vn are on the different cycles of G.
In this case G is the b-graph or ∞-graph with trees attached.
Making some steps of T1-transformation for G
we can observe that either GT1 is θ-graph with trees attached
or dGT1(v1,vn)≤dG(v1,vn)−1.
If the former rises then the Case 1 shows that the case is true,
and otherwise repeat this process.
Thus we can determine that the case is true.
Case 3. One of v1 and vn is on the cycle and another is on the tree of G.
Suppose without loss of generality that v1 is on the tree and that vn is on the cycle.
After making some steps of T1-transformation for G
we can see that either v1 and vn is on the same cycle of GT1
or dGT1(v1,vn)≤dG(v1,vn)−1.
If the former rises then the Case 1 shows that the case is true,
and otherwise repeat this process.
Thus we can determine that the case is true.
Case 4. Both v1 and vn are on the tree of G.
After making some steps of T1-transformation for G
we can see that either one of v1 and vn is on the cycle and another is on the tree of GT1,
or dGT1(v1,vn)≤dG(v1,vn)−1.
If the former rises then the Case 3 shows that the case is true,
and otherwise repeat this process.
Hence we can determine that the case is true. □
If X=(X1,X2,…,Xn)T satisfies XTQ(G)X=λ(G)
then X is the unit first signless Laplacian eigenvector of G.
Clearly, for each 1≤i≤n, we have
[TABLE]
where NG(vi) denotes the neighbour set of vi in the graph G.
The equation (2.3) is also called the signless Laplacian eigenvalue-equation of G.
If X=(X1,X2,…,Xn)T is an unit first signless Laplacian eigenvector of a graph G
such that X1≥X2≥⋯≥Xn,
then X1>0 and Xn<0
since the matrix Q(G)−λ(G)In is positive semidefinite,
where In is the identity matrix of order n.
We use θ4 and θ5 to denote respectively θ-graph on 4 and 5 vertices,
and ∞5 and ∞6 to denote respectively ∞-graph on 5 and 6 vertices,
and b6 and b7 to denote respectively b-graph on 6 and 7 vertices whose two cycles are all length three.
Let the graphs Hi(1≤i≤7) be shown in Fig. 2.
Suppose H is the graph on n vertices which is obtained from θ4, θ5, ∞5,
∞6, b6 or b7 by attaching some trees to it.
Then we can find a pendent vertex vs whose neighbour vt is neither v1 nor vn.
Let X=(X1,X2,…,Xn)T satisfy X1>0 and Xn<0.
Delete vtvs, and add the edge v1vs if (X1+Xs)2≥(Xn+Xs)2,
and otherwise add the edge vnvs.
The above process is called T2-transformation of H.
Next we always denote by HT2 the result graph which is obtained after some steps of T2-transformation for H.
Now we use T1 and T2-transformations to prove the below important result.
Lemma 2.3.Let H′′={Hi∣1≤i≤7}.
Given a graph Gc∈Cnc (n≥12),
there is a graph H∈H′∪H′′
such that λ(Gc)≥λ(Hc).
Proof. Set H=H′∪H′′
and X=(X1,X2,…,Xn)T to be the unit first signless Laplacian eigenvector of Gc
such that X1≥X2≥⋯≥Xn.
Then X1>0 and Xn<0.
If G∈H then it is finished.
So we assume G∈/H.
If dG(v1,vn)>2
then making some steps of T1-transformation for G,
by Lemma 2.2 we know that either GT1∈H
or GT1∈/H but dGT1(v1,vn)≤2.
If the former rises, then by the equation (2.1) and Lemma 2.1, we have
[TABLE]
Now we assume the latter rises and distinguish two cases to discuss GT1.
Case 1.dGT1(v1,vn)=1.
Case 1.1.v1 and vn are on the same cycle Ct+2 of order t+2.
If t≥2 then set Ct+2=v1vnvl1vl2⋯vltv1.
Delete vl1vl2, and add the edge v1vl1 if (X1+Xl1)2≥(Xn+Xl2)2, and otherwise add the edge vnvl2.
Repeat this process until the result graph G11 contains the cycle C3=v1vnvl1v1.
Set vs1vs2⋯vskvs1 to be another cycle which is contained in G11.
Delete vs1vs2, and add the edge v1vs2 if (X1+Xs2)2≥(Xn+Xs1)2, and otherwise add the edge vnvs1.
Repeating this process we can determine the result graph G is θ4 or ∞5 with trees attached.
If G∈/H,
then making some steps of T2-transformation for G
we can get the graph GT2 which is isomorphic to one of the graphs H1,H2 and H3 as in Fig. 2.
Case 1.2.v1 and vn are on the different cycle.
In this case GT1 is the b-graph with trees attached.
Suppose GT1 contains the cycle Ct+1=v1vl1vl2⋯vltv1.
When t≥3 we delete vl1vl2, and add the edge v1vl2 if (X1+Xl2)2≥(Xn+Xl1)2,
and otherwise add the edge vnvl1.
Denote by G21 the result graph.
If v1 and vn are on the same cycle of G21 then we enter Case 1.1,
and otherwise repeat this process until the result graph G22 contains the cycle C3=v1vl1vl2v1.
Next set vnvs1vs2⋯vskvn to be another cycle which is contained in G22.
Delete vs1vs2,
and add the edge vnvs2 if (Xn+Xs2)2≥(X1+Xs1)2, and otherwise add the edge v1vs1.
Denote by G23 the result graph.
If v1 and vn are on the same cycle of G23, then we enter Case 1.1,
and otherwise repeat this process until the result graph G is b6 with trees attached.
If G∈/H,
then making some steps of T2-transformation for G.
We can determine that the graph GT2 is isomorphic to the graph H4 as in Fig. 2.
Case 1.3. One of v1 and vn is on the cycle and another is on the tree.
Suppose without loss of generality that v1 is on the cycle and that vn is on the tree
for otherwise we replace X with −X.
Then GT1 contains the cycle Ct+1=v1vl1vl2⋯vltv1.
When t≥3 we delete vl1vl2, and add the edge v1vl2 if (X1+Xl2)2≥(Xn+Xl1)2, and otherwise add the edge vnvl1.
We denote by G31 the result graph.
If v1 and vn are on the same cycle of G31 then we enter Case 1.1,
and otherwise repeat this process until the result graph G32 contains the cycle C3=v1vl1vl2v1.
Set vs1vs2⋯vskvs1 to be another cycle of G32.
Delete vs1vs2, and add the edge v1vs2 if (X1+Xs2)2≥(Xn+Xs1)2, and otherwise add the edge vnvs1.
We denote by G33 the result graph.
If v1 and vn are on the cycles of G33 then we enter Case 1.1 or Case 1.2,
and otherwise repeating this process we can determine that the result graph G is θ4 or ∞5 with trees attached.
If G∈/H,
then making some steps of T2-transformation for G we can get the graph GT2
which is isomorphic to one of the graphs H5, H6 and H7 as in Fig. 2.
Case 1.4. Both v1 and vn are on the same tree T.
Suppose that the tree T is attached at the vertex vlm of the cycle Ck.
Now we observe the path vnv1vl1vl2⋯vlm.
Delete vl1vl2, and add the edge v1vl2 if (X1+Xl2)2≥(Xn+Xl2)2,
and otherwise add the edge vnvl2.
We make some steps of the above transformation until vl1 is on the cycle vl1vs2⋯vskvl1.
Next delete vl1vs2, and add the edge v1vs2 if (X1+Xs2)2≥(Xn+Xs2)2,
and otherwise add the edge vnvs2.
Then we can get the result graph G41 such that
v1 and vn are on the same cycle of G41 or one of v1 and vn is on the cycle and another is on the tree.
If the former rises then we enter Case 1.1, and otherwise we enter Case 1.3.
Case 2.dGT1(v1,vn)=2.
Case 2.1.v1 and vn are on the same cycle Ct+3.
If t≥2 then set Ct+3=v1vtvnvl1vl2⋯vltv1.
Delete vl1vl2, and add the edge v1vl1 if (X1+Xl1)2≥(Xn+Xl2)2,
and otherwise add the edge vnvl2.
Repeat this process until the result graph G11 contains the cycle C4=v1vtvnvl1v1.
Set vs1vs2⋯vskvs1 to be another cycle which is contained in G11.
Delete vs1vs2, and add the edge v1vs2 if (X1+Xs2)2≥(Xn+Xs1)2, and otherwise add the edge vnvs1.
Repeating this process we can determine that the result graph G is θ5 or ∞6 with trees attached.
If G∈/H,
then we make some steps of T2-transformation for G until the result graph GT2
is isomorphic to one of the graphs G1(p,q), G2(p,q) and G3(p,q) as in Fig. 1.
Case 2.2.v1 and vn are on the different cycle.
In this case GT1 is the b-graph or ∞-graph with trees attached.
Suppose GT1 contains the cycle Ct+1=v1vl1vl2⋯vltv1.
If t≥3 then delete vl1vl2, and add the edge v1vl2 if (X1+Xl2)2≥(Xn+Xl1)2,
and otherwise add the edge vnvl1.
We denote by G21 the result graph.
If v1 and vn are on the same cycle of G21 then we enter Case 2.1,
and otherwise repeat this process until the result graph G22 contains the cycle C3.
Let vnvs1vs2⋯vskvn be another cycle of G22.
Delete vs1vs2, and add the edge v1vs1 if (X1+Xs1)2≥(Xn+Xs2)2, and otherwise add the edge vnvs2.
We denote by G23 the result graph.
If v1 and vn are on the same cycle of G23 then we enter Case 2.1,
and otherwise repeat this process until result graph
G is ∞5, b7 or b6 with trees attached.
If G∈/H,
then we make some steps of T2-transformation for G
until the result graph GT2 is isomorphic to one of the graphs G4(p,q), G5(p,q) and G6(p,q) as in Fig. 1.
Case 2.3. One of v1 and vn is on the cycle and another is on the tree.
Suppose without loss of generality that v1 is on the cycle and that vn is on the tree.
Then GT1 contains the cycle Ct+1=v1vl1vl2⋯vltv1.
If t≥3 then delete vl1vl2, and add the edge v1vl2 if (X1+Xl2)2≥(Xn+Xl1)2,
and otherwise add the edge vnvl1.
We denote by G31 the result graph.
If v1 and vn are on the same cycle of G31 then we enter Case 2.1,
and otherwise repeat this process until the result graph G32 contains the cycle C3=v1vl1vl2v1.
Set vs1vs2⋯vskvs1 to be another cycle of G32.
Delete vs1vs2, and add the edge v1vs2 if (X1+Xs2)2≥(Xn+Xs1)2, and otherwise add the edge vnvs1.
We denote by G33 the result graph.
If v1 and vn are on the cycles of G33 then we enter Case 2.1 or Case 2.2,
and otherwise repeat this process until the result graph G is θ4 or ∞5 with trees attached.
If G∈/H,
then we make some steps of T2-transformation for G until the result graph GT2
is isomorphic to one of graphs G7(p,q), G8(p,q), G9(p,q), G10(p,q), G11(p,q) and G12(p,q) as in Fig. 1.
Case 2.4. Both v1 and vn are on the same tree T.
Suppose that the tree T is attached at the vertex vlm of the cycle Ck.
Let vt be the common neighbour of v1 and vn.
Now we distinguish two cases to discuss.
Case 2.4.1. Suppose that there exists the path vnvtv1vl1vl2⋯vlm.
Delete vl1vl2, and add the edge v1vl2 if (X1+Xl2)2≥(Xn+Xl2)2, and otherwise add the edge vnvl2.
We make some steps of the above transformation
until we obtain the cycle vl1vs2vs3⋯vskvl1.
Delete vl1vs2, and add the edge v1vs2 if (X1+Xs2)2≥(Xn+Xs2)2,
and otherwise add the edge vnvs2.
We denote by G41 the result graph.
It can be observed that v1 is on the cycle and vn is on the tree of G41 or v1 and vn are on the same cycle of G41.
If the former rises then we enter Case 2.1, and otherwise we enter Case 2.3.
Case 2.4.2. Suppose that there exists the path v1vtvl1vl2⋯vlm.
Delete vtvl1, and add the edge v1vl1
if (X1+Xl1)2≥(Xn+Xl1)2,
and otherwise add the edge vnvl1.
We denote by G42 the result graph.
If we obtain the path vnvtv1vl1vl2⋯vlm,
then we enter Case 2.4.1,
and otherwise repeat this process until we obtain the cycle vtvs2vs3⋯vskvt.
Delete vtvl1, and add the edge v1vl1 if (X1+Xl1)2≥(Xn+Xl1)2,
and otherwise add the edge vnvl1.
We denote by G43 the result graph.
It can be observed that one of v1 and vn is on the cycle and another is on the tree of G43,
and then we enter Case 2.3.
From the above argument we observe that
for any Gc∈Cnc (n≥12),
there is a graph H∈H′∪H′′
such that
[TABLE]
Note that Q(Gc)=(n−2)In+Jn−Q(G),
where Jn denotes the all ones square matrix of order n.
Therefore, using the equations (2.2) and (2.4) we obtain
[TABLE]
Lemma 2.4.If H∈H′′,
then there is a graph H∗∈H′
such that λ(Hc)≥λ(H∗c).
Proof. Let X=(X1,X2,…,Xn)T be the unit first signless Laplacian eigenvector of Hc
such that X1≥X2≥⋯≥Xn. Then X1>0 and Xn<0.
We suppose without loss of generality that
∣Xn∣≥∣X1∣
for otherwise we can replace X with −X.
For the graphs Hi∈H′′ (i=1,2),
deleting the edge v1vn, and adding the edge vnvt,
we get the result graph H∗ which is isomorphic to G1(p,q) or G2(p,q) in H′.
Thus we obtain
[TABLE]
For the graph Hi∈H′′(3≤i≤7),
deleting the edge v1vn, and adding the edge vnvk,
we get the result graph H∗ which is isomorphic to Gℓ(p,q) (ℓ=2,6,7,8 or 11).
Thus we obtain
[TABLE]
Therefore, combining the equations (2.2), (2.5) and (2.6) we have
[TABLE]
Lemma 2.5 [7]. Let G be a simple graph.
Then λ(G)≤δ(G), where δ(G)=min{dG(v),v∈V(G)}.
Lemma 2.6.Let p and q be two positive integers such that p+q=n−5 (n≥12).
Then λ(G1c(p,q))>λ(G1c(n−5,0)).
Proof. We without loss of generality assume p≥q≥1.
Next we will prove λ(G1c(p,q))>λ(G1c(p+1,q−1)).
Suppose X=(X1,X2,…,Xn)T is the unit first signless Laplacian eigenvector of G1c(p,q),
and let k1=λ(G1c(p,q)).
Then, by the symmetry of G1c(p,q) and the signless Laplacian eigen-equation (2.3),
we have
Note that G1c(p,q) is connected and δ(G1c(p,q))=q+1.
By Lemma 2.5, we have 0<k1≤q+1.
Since f1(k1;p,q)=0 and p≥q≥1,
it follows that f1(k1;p+1,q−1)>0 from the equation (2.7).
This shows that λ(G1c(p,q))>λ(G1c(p+1,q−1)),
which implies λ(G1c(p,q))>λ(G1c(n−5,0)).□
Lemma 2.7.Let p and q be two positive integers such that p+q=n−5 (n≥12).
Then λ(G2c(p,q))>λ(G2c(n−5,0)).
Proof. Suppose X=(X1,X2,⋯,Xn)T is the unit first signless Laplacian eigenvector of G2c(p,q),
and let k2=λ(G2c(p,q)).
Then, by the symmetry of G2c(p,q) and the signless Laplacian eigen-equation (2.3),
we have
where g1(x)=(n−q−4)x4+(−5n2+5nq+41n−21q−84)x3+(9n3−9n2q−110n2+74nq+447n−151q−609)x2+(−7n4+7n3q+113n3−85n2q−679n2+339nq+1799n−436q−1760)x+2n5−2n4q−40n4+32n3q+316n3−188n2q−1228n2+474nq+2330n−428q−1708.
Note that G2c(p,q) is connected and δ(G2c(p,q))=min{q+1,p+2}.
By Lemma 2.5, we have 0<k2≤min{q+1,p+2}.
Claim A in the Appendix shows that g1(x)>0 when 0<x≤min{q+1,p+2}.
Since f2(k2;n−5−q,q)=0,
it follows that f2(k2;n−5,0)>0 from the equation (2.8).
This shows that λ(G2c(p,q))>λ(G2c(n−5,0)). □
For Gk(p,q)∈H′(k=3,6,7,8,9,10,11,12),
we can prove as the proof of Lemma 2.7 that the following Lemma is true.
Lemma 2.8.Let p and q be two positive integers and n≥12.
Then we have
(i)
when p+q=n−5, λ(Gsc(p,q))>λ(Gsc(n−5,0))(s=8,9,11),
2. (ii)
when p+q=n−6, λ(Gtc(p,q))>λ(Gtc(n−6,0))(t=3,7,10), and
λ(G6c(p,q))>λ(G6c(0,n−6)),
3. (iii)
when p+q=n−7, λ(G12c(p,q))>λ(G12c(n−7,0)).
Lemma 2.9. *Let p and q be two positive integers such that p+q=n−7 (n≥12).
Then λ(G4c(p,q))>λ(G4c(n−7,0)). *
Proof. Without loss of generality assume p≥q≥1.
We will first prove λ(G4c(p,q))>λ(G4c(p+1,q−1)).
Suppose X=(X1,X2,…,Xn)T is the unit first signless Laplacian eigenvector of G4c(p,q),
and let k4=λ(G4c(p,q)).
Then, by the symmetry of G4c(p,q) and the signless Laplacian eigen-equation (2.3),
we have
where g2(x)=−x3+(4p+4q+11)x2+(−5p2−10pq−5q2−29p−29q−46)x+2p3+6p2q+6pq2+2q3+18p2+36pq+18q2+56p+56q+72.
Note that G4c(p,q) is connected and δ(G4c(p,q))=q+3.
By Lemma 2.5, we have 0<k4≤q+3.
Claim B in the Appendix shows that g2(x)>0 when 0<x≤q+3.
Since f4(k4;p,q)=0,
it follows f4(k4;p+1,q−1)>0 from the equation (2.9).
This shows that λ(G4c(p,q))>λ(G4c(p+1,q−1)),
from which we obtain λ(G4c(p,q))>λ(G4c(n−7,0)).□
For G5(p,q)∈H′,
we can prove as the proof of Lemma 2.9 that the following result is true.
Lemma 2.10.Let p and q be two positive integers such that p+q=n−5(n≥12).
Then λ(G5c(p,q))>λ(G5c(n−5,0)).
Theorem 2.11.For any graph Gc∈Cnc(n≥12),
we have λ(Gc)≥λ(G1c(n−5,0))
with equality if and only if G≅G1(n−5,0).
Proof. Let f1(x;p,q), f2(x;p,q) and f4(x;p,q)
be as in the proof of Lemmas 2.6, 2.7 and 2.9, respectively.
Let g12(x)=f1(x;n−5,0)−f2(x;n−5,0).
Then g12(x)=(n−x−3)2(−2n2+(−x−18)n−2x2+8x+40).
Since 0<x≤1 and n≥12,
we have g12(x)>0.
Lemma 2.5 shows 0<λ(G1c(n−5,0)),λ(G2c(n−5,0))≤1.
Note that f1(λ(G1c(n−5,0));n−5,0)=0.
Thus, we obtain f2(λ(G1c(n−5,0));n−7,0)<0.
This implies λ(G1c(n−5,0))<λ(G2c(n−7,0)).
Let g14(x)=f1(x;n−5,0)−f4(x;n−7,0).
Then g14(x)=(n−3−x)g4(x),
where g4(x)=(2n−9)x4+(−10n2+92n−208)x3+(18n3−249n2+1149n−1814)x2+(−14n4+258n3−1791n2+5607n−6672)x+4n5−92n4+850n3−3956n2+9242n−8592.
Claim C in the Appendix shows that g4(x)>0, and so g14(x)>0 when 0<x≤1 and n≥12.
Lemma 2.5 shows 0<λ(G1c(n−5,0)),λ(G4c(n−7,0))≤1.
Note that f1(λ(G1c(n−5,0));n−5,0)=0.
Thus, we obtain f4(λ(G1c(n−5,0));n−7,0)<0.
This implies λ(G1c(n−5,0))<λ(G4c(n−7,0)).
Similarly, we can verify that the following results are true.
[TABLE]
Combining these inequations with Lemmas 2.3, 2.4, 2.6, 2.7, 2.8, 2.9 and 2.10,
we determine that the result is true. □
Bibliography12
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] D. M. Cardoso, D. Cvetković, P. Rowlinson, S. K. Simić, A sharp lower bound for the least engivenvalue of the signless Laplacian of a non-bipartite graph, Linear Algebra Appl. 429 (2008) 2770-2780.
2[2] Y. Z. Fan, Y. Wang, H. Guo, The least eigenvalues of the signless Laplacian of non-bipartite graphs with pendant vertices, Discrete Mathematics 313 (2013) 903-909.
3[3] J. M. Guo, J. Y. Ren, J. S. Shi, Maximizing the least signless Laplacian eigenvalue of unicyclic graphs, Linear Algebra Appl. 519 (2017) 136-145.
4[4] S. G. Guo, Y. G. Chen, G. Yu, A lower bound on the least signless Laplacian eigenvalue of a graph, Linear Algebra Appl. 448 (2014) 217-221.
5[5] S. G. Guo, R. Zhang, On the least signless Laplacian eigenvalue of a non-bipartite connected graph with fixed maximum degree, Journal of Inequalities and Applications (2017) 1-11.
6[6] C. X. He, M. Zhou, A Sharp Upper Bound on the Least Signless Laplacian Eigenvalue Using Domination Number, Graphs and Combinatorics 30 (2014) 1183-1192.
7[7] S. C. Li, S. J. Wang, The least eigenvalue of the signless Laplacian of the complements of trees, Linear Algebra Appl. 436 (2012) 2398-2405.
8[8] Y. Wang, Y. Z. Fan, The least eigenvalue of signless Laplacian of graphs under perturbation, Linear Algebra Appl. 436 (2012) 2084-2092.