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

Spectral radius and spanning trees of graphs111Supported by National Natural Science Foundation of China (Nos. 11971445 and 12261032) and Natural Science Foundation of Henan Province (No. 202300410377).

Guoyan Aoa,b, Ruifang Liua, Jinjiang Yuana
a
School of Mathematics and Statistics, Zhengzhou University, Zhengzhou, Henan 450001, China
b School of Mathematics and Statistics, Hulunbuir University, Hailar, Inner Mongolia 021008, China
Corresponding author. E-mail addresses: [email protected], [email protected], [email protected].

Abstract

For integer k2,k\geq 2, a spanning kk-ended-tree is a spanning tree with at most kk leaves. Motivated by the closure theorem of Broersma and Tuinstra [Independence trees and Hamilton cycles, J. Graph Theory 29 (1998) 227–237], we provide tight spectral conditions to guarantee the existence of a spanning kk-ended-tree in a connected graph of order nn with extremal graphs being characterized. Moreover, by adopting Kaneko’s theorem [Spanning trees with constraints on the leaf degree, Discrete Appl. Math. 115 (2001) 73–76], we also present tight spectral conditions for the existence of a spanning tree with leaf degree at most kk in a connected graph of order nn with extremal graphs being determined, where k1k\geq 1 is an integer.

Keywords: Spanning tree, Spectral radius, Closure, Leaf degree

AMS Classification: 05C50; 05C35

1 Introduction

In this paper, we consider simple, undirected and connected graphs. For undefined terms and notions, one can refer to [5] and [2]. Let GG be a graph with vertex set V(G)V(G) and edge set E(G)E(G). The order and size of GG are denoted by |V(G)|=n|V(G)|=n and |E(G)|=e(G)|E(G)|=e(G), respectively. We denote by dG(v)d_{G}(v), ω(G)\omega(G), i(G)i(G) and G¯\overline{G} the vertex vv of degree in GG, the clique number, the number of isolated vertices and the complement of G,G, respectively. We use PnP_{n}, CnC_{n}, KnK_{n} and Km,nK_{m,n} to denote the path of order nn, the cycle of order nn, the complete graph of order nn, and the complete bipartite graph with bipartition (X,Y)(X,Y), where |X|=m|X|=m and |Y|=n|Y|=n. Let G1G_{1} and G2G_{2} be two vertex-disjoint graphs. We denote by G1+G2G_{1}+G_{2} the disjoint union of G1G_{1} and G2.G_{2}. The join G1G2G_{1}\vee G_{2} is the graph obtained from G1+G2G_{1}+G_{2} by adding all possible edges between V(G1)V(G_{1}) and V(G2)V(G_{2}).

Let A(G)A(G) and D(G)D(G) be the adjacency matrix and diagonal degree matrix of GG. Let Q(G)=D(G)+A(G)Q(G)=D(G)+A(G) be the signless Laplacian matrix of GG. The largest eigenvalues of A(G)A(G) and Q(G)Q(G), denoted by ρ(G)\rho(G) and q(G)q(G), are called the spectral radius and the signless Laplacian spectral radius of GG, respectively. For convenience, Liu, Lai and Das [23] introduced the nonnegative matrix Aa(G)=aD(G)+A(G)(a0)A_{a}(G)=aD(G)+A(G)~{}(a\geq 0) of a graph GG. The largest eigenvalue of Aa(G)A_{a}(G) is called the AaA_{a}-spectral radius of GG, denoted by ρa(G)\rho_{a}(G). It is clear that ρ0(G)=ρ(G)\rho_{0}(G)=\rho(G) and ρ1(G)=q(G).\rho_{1}(G)=q(G).

The concept of closure of a graph was used implicitly by Ore [25], and formally introduced by Bondy and Chvatal [4]. Fix an integer l0l\geq 0, the ll-closure of a graph GG is the graph obtained from GG by successively joining pairs of nonadjacent vertices whose degree sum is at least ll until no such pair exists. Denote by Cl(G)C_{l}(G) the ll-closure of G.G.

For integer k2,k\geq 2, a spanning kk-ended-tree of a connected graph GG is a spanning tree with at most kk leaves. Note that a spanning 22-ended-tree is a Hamilton path. Hence a spanning kk-ended-tree of a graph is a natural generalization of a Hamilton path. Fiedler and Nikiforov [14] initially proved sufficient conditions for a graph to contain a Hamilton path in terms of the spectral radius of a graph or its complement. Ning and Ge [20], Li and Ning [21] improved and extended results in [14]. As a special case of spanning kk-ended-tree, there are many sufficient conditions to assure a graph to contain a Hamilton path (see for example [6, 22, 24, 28, 29]). In fact, a Hamilton path can also be generalized to a spanning kk-tree. A spanning kk-tree of a connected graph GG is a spanning tree in which every vertex has degree at most kk, where k2k\geq 2 is an integer. Fan et al. [11] presented spectral conditions for the existence of a spanning kk-tree in a connected graph.

Broersma and Tuinstra [3] showed that the problem of whether a given connected graph contains a spanning kk-ended tree is NP-complete. Furthermore, they presented a degree sum condition to guarantee that a connected graph has a spanning kk-ended tree.

Theorem 1.1 (Broersma and Tuinstra [3]).

Let k2k\geq 2 be an integer, and let GG be a connected graph of order nn. If d(u)+d(v)nk+1d(u)+d(v)\geq n-k+1 for every two nonadjacent vertices uu and vv, then GG has a spanning kk-ended tree.

Moreover, Broersma and Tuinstra [3] proved the following closure theorem for the existence of a spanning kk-ended-tree, which is crucial to our proof.

Theorem 1.2 (Broersma and Tuinstra [3]).

Let GG be a connected graph of order n,n, and kk be an integer with 2kn1.2\leq k\leq n-1. Then GG has a spanning kk-ended-tree if and only if the (n1)(n-1)-closure Cn1(G)C_{n-1}(G) of GG has a spanning kk-ended-tree.

Inspired by the work of Broersma and Tuinstra [3], we focus on the following interesting problem in this paper.

Problem 1.1.

Let GG be a connected graph of order nn which has no a spanning kk-ended-tree. What is the maximum spectral radius (or signless Laplacian spectral radius) of G?G? Moreover, characterize all the extremal graphs.

For more results on spanning kk-ended-tree, one can refer to [7, 8, 9, 12, 18, 19, 26, 27]. Inspired by the ideas on Hamilton path from Li and Ning [21] and using typical spectral techniques, we prove tight spectral conditions to guarantee the existence of a spanning kk-ended-tree in a connected graph. Let R(n,s)R(n,s) be an ss-regular graph with nn vertices.

Theorem 1.3.

Let GG be a connected graph of order nn and k2k\geq 2 be an integer. Each of the following holds.
(i) If nmax{6k+5,k2+32k+2}n\geq{\rm max}\{6k+5,k^{2}+\frac{3}{2}k+2\} and ρ(G)ρ(K1(Knk1+kK1)),\rho(G)\geq\rho(K_{1}\vee(K_{n-k-1}+kK_{1})), then GG has a spanning kk-ended-tree unless GK1(Knk1+kK1)G\cong K_{1}\vee(K_{n-k-1}+kK_{1}).
(ii) If nmax{6k+5,32k2+32k+2}n\geq{\rm max}\{6k+5,\frac{3}{2}k^{2}+\frac{3}{2}k+2\} and q(G)q(K1(Knk1+kK1)),q(G)\geq q(K_{1}\vee(K_{n-k-1}+kK_{1})), then GG has a spanning kk-ended-tree unless GK1(Knk1+kK1)G\cong K_{1}\vee(K_{n-k-1}+kK_{1}).
(iii) If ρ(G¯)k(n2),\rho(\overline{G})\leq\sqrt{k(n-2)}, then GG has a spanning kk-ended-tree unless GK1,k+1G\cong K_{1,k+1}.
(iv) If q(G¯)n+k2q(\overline{G})\leq n+k-2 and GR(t,tn+k2)Knt(n+k2tn),G\notin R(t,t-\frac{n+k}{2})\vee K_{n-t}~{}(\frac{n+k}{2}\leq t\leq n), then GG has a spanning kk-ended-tree.

Kaneko [17] introduced the concept of leaf degree of a spanning tree. Let TT be a spanning tree of a connected graph GG. The leaf degree of a vertex vV(T)v\in V(T) is defined as the number of leaves adjacent to vv in TT. Furthermore, the leaf degree of TT is the maximum leaf degree among all the vertices of TT. They posed a necessary and sufficient condition for a connected graph to contain a spanning tree with leaf degree at most kk. Let i(GS)i(G-S) denote the number of isolated vertices of GS.G-S.

Theorem 1.4 (Kaneko [17]).

Let GG be a connected graph and k1k\geq 1 be an integer. Then GG has a spanning tree with leaf degree at most kk if and only if i(GS)<(k+1)|S|i(G-S)<(k+1)|S| for every nonempty subset SV(G)S\subseteq V(G).

In this paper, we consider the following spectral extremal problem.

Problem 1.2.

Let GG be a connected graph of order nn which has no a spanning tree with leaf degree at most kk. What is the maximum spectral radius (or signless Laplacian spectral radius) of G?G? Moreover, characterize all the extremal graphs.

Motivated by Kaneko’s theorem, we present tight spectral conditions for the existence of a spanning tree with leaf degree at most kk in a connected graph.

Theorem 1.5.

Let GG be a connected graph of order n2k+12n\geq 2k+12, where k1k\geq 1 is an integer and a{0,1}a\in\{0,1\}. If ρa(G)ρa(K1(Knk2+(k+1)K1)),\rho_{a}(G)\geq\rho_{a}(K_{1}\vee(K_{n-k-2}+(k+1)K_{1})), then GG has a spanning tree with leaf degree at most kk unless GK1(Knk2+(k+1)K1)G\cong K_{1}\vee(K_{n-k-2}+(k+1)K_{1}).

2 Proof of Theorem 1.3

Before presenting our main result, we first show that an (n1)(n-1)-closed connected graph GG must contain a large clique if its number of edges is large enough.

Lemma 2.1.

Let HH be an (n1)(n-1)-closed connected graph of order n6k+5n\geq 6k+5, where k2k\geq 2 is an integer. If

e(H)(nk12)+k2+k+1,e(H)\geq{n-k-1\choose 2}+k^{2}+k+1,

then ω(H)nk\omega(H)\geq n-k.

Proof.

For any two vertices u,vV(H)u,v\in V(H) with degree at least n12\frac{n-1}{2}, we have dH(u)+dH(v)n1.d_{H}(u)+d_{H}(v)\geq n-1. Note that HH is an (n1)(n-1)-closed graph. Then any two vertices of degree at least n12\frac{n-1}{2} must be adjacent in HH. Let CC be the vertex set of a maximum clique of HH which contains all vertices of degree at least n12,\frac{n-1}{2}, and FF be the subgraph of HH induced by V(H)C.V(H)\setminus C. Let |C|=r|C|=r. Then |V(F)|=nr|V(F)|=n-r. Next we will evaluate the value of r.r.

Case 1. 1rn3+k+1.1\leq r\leq\frac{n}{3}+k+1.

Note that

e(H[C])=(r2),e(C,V(F))=uV(F)dC(u)ande(F)=uV(F)dH(u)uV(F)dC(u)2.e(H[C])={r\choose 2},e(C,V(F))=\sum_{u\in V(F)}d_{C}(u)~{}~{}\mbox{and}~{}~{}e(F)=\frac{\sum_{u\in V(F)}d_{H}(u)-\sum_{u\in V(F)}d_{C}(u)}{2}.
Claim 1.

dH(u)n22d_{H}(u)\leq\frac{n-2}{2} and dC(u)r1d_{C}(u)\leq r-1 for each uV(F)u\in V(F).

Proof.

Suppose to the contrary that dH(u)n12d_{H}(u)\geq\frac{n-1}{2} or dC(u)rd_{C}(u)\geq r for each uV(F)u\in V(F). Assume first that dH(u)n12d_{H}(u)\geq\frac{n-1}{2}. Then we have dH(u)+dH(v)n1d_{H}(u)+d_{H}(v)\geq n-1 for each vCv\in C. Note that HH is an (n1)(n-1)-closed graph. Then uu is adjacent to every vertex of C,C, and hence C{u}C\cup\{u\} is a larger clique, which contradicts the maximality of CC. Note that |C|=r|C|=r. If dC(u)rd_{C}(u)\geq r, then dC(u)=r,d_{C}(u)=r, and hence uu is adjacent to every vertex of C.C. It follows that C{u}C\cup\{u\} is a larger clique, a contradiction. ∎

By Claim 1, we have

e(H)\displaystyle e(H) =\displaystyle= e(H[C])+e(C,V(F))+e(F)\displaystyle e(H[C])+e(C,V(F))+e(F)
=\displaystyle= (r2)+uV(F)dC(u)+uV(F)dH(u)2\displaystyle{r\choose 2}+\frac{\sum_{u\in V(F)}d_{C}(u)+\sum_{u\in V(F)}d_{H}(u)}{2}
\displaystyle\leq (r2)+(r1)(nr)2+(n2)(nr)4\displaystyle{r\choose 2}+\frac{(r-1)(n-r)}{2}+\frac{(n-2)(n-r)}{4}
=\displaystyle= (n4+12)r+n24n\displaystyle(\frac{n}{4}+\frac{1}{2})r+\frac{n^{2}}{4}-n
\displaystyle\leq (n4+12)(n3+k+1)+n24n\displaystyle(\frac{n}{4}+\frac{1}{2})(\frac{n}{3}+k+1)+\frac{n^{2}}{4}-n
=\displaystyle= n23+(k4712)n+k2+12\displaystyle\frac{n^{2}}{3}+(\frac{k}{4}-\frac{7}{12})n+\frac{k}{2}+\frac{1}{2}
<\displaystyle< (nk12)+k2+k+1,\displaystyle{n-k-1\choose 2}+k^{2}+k+1,

for n6k+5n\geq 6k+5. This contradicts e(H)(nk12)+k2+k+1.e(H)\geq{n-k-1\choose 2}+k^{2}+k+1.

Case 2. n3+k+1<rnk1.\frac{n}{3}+k+1<r\leq n-k-1.

Note that

e(H[C])=(r2)ande(C,V(F))+e(F)uV(F)dH(u).e(H[C])={r\choose 2}~{}~{}\mbox{and}~{}~{}e(C,V(F))+e(F)\leq\sum_{u\in V(F)}d_{H}(u).
Claim 2.

dH(u)nr1d_{H}(u)\leq n-r-1 for each uV(F)u\in V(F).

Proof.

Suppose that dH(u)nrd_{H}(u)\geq n-r for each uV(F)u\in V(F). Then we have dH(u)+dH(v)(nr)+(r1)=n1d_{H}(u)+d_{H}(v)\geq(n-r)+(r-1)=n-1 for each vCv\in C. Note that HH is (n1)(n-1)-closed. Then uu is adjacent to every vertex of CC. This implies that C{u}C\cup\{u\} is a larger clique, a contradiction. ∎

By Claim 2, we have

e(H)\displaystyle e(H) =\displaystyle= e(H[C])+e(C,V(F))+e(F)\displaystyle e(H[C])+e(C,V(F))+e(F)
\displaystyle\leq (r2)+uV(F)dH(u)\displaystyle{r\choose 2}+\sum_{u\in V(F)}d_{H}(u)
\displaystyle\leq (r2)+(nr)(nr1)\displaystyle{r\choose 2}+(n-r)(n-r-1)
=\displaystyle= 32r2(2n12)r+n2n\displaystyle\frac{3}{2}r^{2}-(2n-\frac{1}{2})r+n^{2}-n
\displaystyle\triangleq f(r).\displaystyle f(r).

Note that f(r)f(r) is a concave function on r.r. Since n3+k+2tnk1,\lfloor\frac{n}{3}\rfloor+k+2\leq t\leq n-k-1, then we have

e(H)max{f(n3+k+2),f(nk1)}=(nk12)+k2+k<e(H),e(H)\leq\max\{f(\lfloor\frac{n}{3}\rfloor+k+2),f(n-k-1)\}={n-k-1\choose 2}+k^{2}+k<e(H),

for n3k+5n\geq 3k+5, a contradiction.

By Cases 1 and 2, we know that rnk,r\geq n-k, and hence ω(H)rnk\omega(H)\geq r\geq n-k. This completes the proof. ∎

With the help of Lemma 2.1, we prove a technical sufficient condition in terms of e(G)e(G) to assure that GG has a spanning kk-ended-tree.

Lemma 2.2.

Let GG be a connected graph of order nmax{6k+5,k2+k+2}n\geq{\rm max}\{6k+5,k^{2}+k+2\}, where k2k\geq 2 is an integer. If

e(G)(nk12)+k2+k+1,e(G)\geq{n-k-1\choose 2}+k^{2}+k+1,

then GG has a spanning kk-ended-tree unless Cn1(G)K1(Knk1+kK1)C_{n-1}(G)\cong K_{1}\vee(K_{n-k-1}+kK_{1}).

Proof.

Suppose that GG has no a spanning kk-ended-tree, where nmax{6k+5,k2+k+2}n\geq{\rm max}\{6k+5,k^{2}+k+2\} and k2k\geq 2. Let H=Cn1(G)H=C_{n-1}(G). It suffices to prove that HK1(Knk1+kK1).H\cong K_{1}\vee(K_{n-k-1}+kK_{1}).

By Theorem 1.2, HH has no a spanning kk-ended-tree either. Note that GHG\subseteq H. Since e(G)(nk12)+k2+k+1e(G)\geq{n-k-1\choose 2}+k^{2}+k+1, then e(H)(nk12)+k2+k+1e(H)\geq{n-k-1\choose 2}+k^{2}+k+1. According to Lemma 2.1, we have ω(H)nk\omega(H)\geq n-k.

Next we will characterize the structure of HH. Let CC be a maximum clique of HH and FF be a subgraph of HH induced by V(H)\C.V(H)\backslash C. First we claim that V(F)V(F)\neq\emptyset. If V(F)=V(F)=\emptyset, then HKnH\cong K_{n}. Obviously, HH has a spanning kk-ended-tree, a contradiction.

Claim 3.

ω(H)=nk\omega(H)=n-k.

Proof.

Recall that ω(H)nk\omega(H)\geq n-k. It suffices to prove that ω(H)nk\omega(H)\leq n-k. Assume to the contrary that ω(H)nk+1\omega(H)\geq n-k+1. Let CC^{\prime} be an (nk+1)(n-k+1)-clique of HH and FF^{\prime} be a subgraph of HH induced by V(H)\CV(H)\backslash C^{\prime}. Then |V(F)|=k1>0|V(F^{\prime})|=k-1>0 because of k2.k\geq 2. Note that GG is a connected graph. Then HH is also connected, and hence there exists a vertex vV(F)v\in V(F^{\prime}) which is adjacent to a vertex in V(C).V(C^{\prime}). We take a path PP such that V(P)=V(C){v}.V(P)=V(C^{\prime})\cup\{v\}. Note that |V(F)v|=k2|V(F^{\prime})-v|=k-2. Therefore, HH has a spanning tree with at most kk leaves, which means that HH has a spanning kk-ended-tree, a contradiction. So we have ω(H)=nk\omega(H)=n-k, as we claimed. ∎

By Claim 3, we know that CKnkC\cong K_{n-k}. Let V(C)={u1,u2,,unk}V(C)=\{u_{1},u_{2},\ldots,u_{n-k}\} and V(F)={v1,v2,,vk}V(F)=\{v_{1},v_{2},\ldots,v_{k}\}.

Claim 4.

V(F)V(F) is an independent set.

Proof.

Assume that viv_{i} is adjacent to vjv_{j}, where vi,vjV(F)v_{i},v_{j}\in V(F). By the connectedness of HH, there exists a path from viv_{i} to CC. Thus, we can take the path PP such that V(C){vi,vj}V(P)V(C)\cup\{v_{i},v_{j}\}\subseteq V(P). Note that |V(H)V(P)|k2|V(H)\setminus V(P)|\leq k-2. Then HH has a spanning tree with at most kk leaves. That is, HH has a spanning kk-ended-tree, a contradiction. ∎

Claim 5.

dH(vi)=1d_{H}(v_{i})=1 for every viV(F).v_{i}\in V(F).

Proof.

Suppose there exists a vertex viV(F)v_{i}\in V(F) with dH(vi)2.d_{H}(v_{i})\geq 2. According to Claim 4, without loss of generality, we can assume that viv_{i} is adjacent to u1u_{1} and u2u_{2}, where u1,u2V(C).u_{1},u_{2}\in V(C). We take the path P=u1viu2u3unkP=u_{1}v_{i}u_{2}u_{3}\cdots u_{n-k} of length nkn-k.

Furthermore, we claim that vj(ji)v_{j}~{}(j\neq i) must be adjacent to u2u_{2} for each vjV(F)v_{j}\in V(F). In fact, if vjv_{j} is adjacent to u1u_{1} or unku_{n-k}, then there exists a path P=P+vju1P^{\prime}=P+v_{j}u_{1} or P+unkvjP+u_{n-k}v_{j} of length nk+1n-k+1. If vj(ji)v_{j}~{}(j\neq i) is adjacent to usu_{s}, where 3snk13\leq s\leq n-k-1. Then there exists a path P=Pus1us+us1unk+usvjP^{\prime}=P-u_{s-1}u_{s}+u_{s-1}u_{n-k}+u_{s}v_{j} of length nk+1n-k+1. By the above proof, we can always find a path PP^{\prime} with V(P)=V(C){vi,vj}.V(P^{\prime})=V(C)\cup\{v_{i},v_{j}\}. Note that |V(F)vivj|=k2|V(F)-v_{i}-v_{j}|=k-2. Hence HH has a spanning tree with at most kk leaves. This implies that HH has a spanning kk-ended-tree, a contradiction. Hence vj(ji)v_{j}~{}(j\neq i) must be adjacent to u2u_{2} for every vjV(F)v_{j}\in V(F).

At this time, we can observe that HH has a spanning tree T=Pu2u3+u1unk+jiu2vjT=P-u_{2}u_{3}+u_{1}u_{n-k}+\sum_{j\neq i}u_{2}v_{j} such that L(T)=V(F){vi}{u3}L(T)=V(F)\setminus\{v_{i}\}\cup\{u_{3}\}, where L(T)L(T) denotes the set of leaves of T.T. Note that |L(T)|=k|L(T)|=k. Hence HH has a spanning kk-ended-tree, a contradiction. ∎

Claim 6.

NH(vi)C=NH(vj)C,N_{H}(v_{i})\cap C=N_{H}(v_{j})\cap C, where iji\neq j.

Proof.

By Claims 4 and 5, we assume that NH(vi)C=uiN_{H}(v_{i})\cap C=u_{i} and NH(vj)C=uj,N_{H}(v_{j})\cap C=u_{j}, where iji\neq j. Then there exists a path PP such that V(P)=V(C){vi,vj}V(P)=V(C)\cup\{v_{i},v_{j}\} and L(P)={vi,vj}L(P)=\{v_{i},v_{j}\}, where L(P)L(P) is the set of leaves of P.P. Note that |V(F)vivj|=k2|V(F)-v_{i}-v_{j}|=k-2. Note that HH is a connected graph. Then HH has a spanning kk-ended-tree, a contradiction. Hence NH(vi)C=NH(vj)C,N_{H}(v_{i})\cap C=N_{H}(v_{j})\cap C, where iji\neq j. ∎

Refer to caption
Figure 1: Graph K1(Knk1+kK1).K_{1}\vee(K_{n-k-1}+kK_{1}).

By the above claims, we know that HK1(Knk1+kK1)H\cong K_{1}\vee(K_{n-k-1}+kK_{1}) (see Fig. 1). Note that the vertices of kK1kK_{1} are only adjacent to a vertex of KnkK_{n-k}. Then the number of leaves of a spanning tree of K1(Knk1+kK1)K_{1}\vee(K_{n-k-1}+kK_{1}) is at least k+1k+1. This implies that K1(Knk1+kK1)K_{1}\vee(K_{n-k-1}+kK_{1}) has no a spanning kk-ended-tree. Note that e(H)=(nk2)+ke(H)={n-k\choose 2}+k and nk2+k+2n\geq k^{2}+k+2. Then we have e(H)(nk12)+k2+k+1.e(H)\geq{n-k-1\choose 2}+k^{2}+k+1. Therefore, H=Cn1(G)K1(Knk1+kK1),H=C_{n-1}(G)\cong K_{1}\vee(K_{n-k-1}+kK_{1}), as desired. ∎

Next we will present important upper and lower bounds of ρ(G)\rho(G) and q(G)q(G) that will be used in our subsequent arguments.

Lemma 2.3 (Hong [16]).

Let GG be a connected graph with nn vertices. Then

ρ(G)2e(G)n+1.\rho(G)\leq\sqrt{2e(G)-n+1}.
Lemma 2.4 (Li and Ning [21]).

Let GG be a graph with non-empty edge set. Then

ρ(G)min{d(u)d(v):uvE(G)}.\rho(G)\geq{\rm min}\{\sqrt{d(u)d(v)}:uv\in E(G)\}.

Moreover, if GG is connected, then equality holds if and only if GG is regular or semi-regular bipartite.

Lemma 2.5 (Das [10], Feng and Yu [13]).

Let GG be a connected graph on nn vertices and e(G)e(G) edges. Then

q(G)2e(G)n1+n2.q(G)\leq\frac{2e(G)}{n-1}+n-2.
Lemma 2.6 (Li and Ning [21]).

Let GG be a graph with non-empty edge set. Then

q(G)min{d(u)+d(v):uvE(G)}.q(G)\geq{\rm min}\{d(u)+d(v):uv\in E(G)\}.

Moreover, if GG is connected, then equality holds if and only if GG is regular or semi-regular bipartite.

Let A=(aij)A=(a_{ij}) and B=(bij)B=(b_{ij}) be two n×nn\times n matrices. Define ABA\leq B if aijbija_{ij}\leq b_{ij} for all ii and jj, and define A<BA<B if ABA\leq B and ABA\neq B.

Lemma 2.7 (Berman and Plemmons [1], Horn and Johnson [15]).

Let A=(aij)A=(a_{ij}) and B=(bij)B=(b_{ij}) be two n×nn\times n matrices with the spectral radii λ(A)\lambda(A) and λ(B)\lambda(B). If 0AB0\leq A\leq B, then λ(A)λ(B)\lambda(A)\leq\lambda(B). Furthermore, if BB is irreducible and 0A<B0\leq A<B, then λ(A)<λ(B)\lambda(A)<\lambda(B).

Now, we are in a position to present the proof of Theorem 1.3.

Proof of Theorem 1.3. (i) Let GG be a connected graph of order n.n. Suppose to the contrary that GG has no a spanning kk-ended-tree, where nmax{6k+5,k2+32k+2}n\geq{\rm max}\{6k+5,k^{2}+\frac{3}{2}k+2\} and k2.k\geq 2. By Lemma 2.7, we know that ρ(K1(Knk1+kK1))>ρ(Knk)=nk1\rho(K_{1}\vee(K_{n-k-1}+kK_{1}))>\rho(K_{n-k})=n-k-1. By Lemma 2.3 and the assumption of Theorem 1.3, we have

nk1<ρ(K1(Knk1+kK1))ρ(G)2e(G)n+1.n-k-1<\rho(K_{1}\vee(K_{n-k-1}+kK_{1}))\leq\rho(G)\leq\sqrt{2e(G)-n+1}.

Hence e(G)>(nk1)2+n12(nk12)+k2+k+1e(G)>\frac{(n-k-1)^{2}+n-1}{2}\geq{n-k-1\choose 2}+k^{2}+k+1 for nk2+32k+2n\geq k^{2}+\frac{3}{2}k+2. Let H=Cn1(G)H=C_{n-1}(G). By Lemma 2.2, we have HK1(Knk1+kK1).H\cong K_{1}\vee(K_{n-k-1}+kK_{1}). Note that GH.G\subseteq H. Then we have

ρ(G)ρ(H)=ρ(K1(Knk1+kK1)).\rho(G)\leq\rho(H)=\rho(K_{1}\vee(K_{n-k-1}+kK_{1})).

Combining the assumption ρ(G)ρ(K1(Knk1+kK1))\rho(G)\geq\rho(K_{1}\vee(K_{n-k-1}+kK_{1})), we have GH.G\cong H. From the end of the proof in Lemma 2.2, we know that K1(Knk1+kK1)K_{1}\vee(K_{n-k-1}+kK_{1}) has no a spanning kk-ended-tree. Hence GK1(Knk1+kK1),G\cong K_{1}\vee(K_{n-k-1}+kK_{1}), as desired.

(ii) Let GG be a connected graph of order nn with nmax{6k+5,32k2+32k+2}.n\geq{\rm max}\{6k+5,\frac{3}{2}k^{2}+\frac{3}{2}k+2\}. Assume that GG has no a spanning kk-ended-tree, where k2k\geq 2 is an integer. By Lemma 2.7, we have q(K1(Knk1+kK1))>q(Knk)=2(nk1).q(K_{1}\vee(K_{n-k-1}+kK_{1}))>q(K_{n-k})=2(n-k-1). By Lemma 2.5 and the assumption, we have

2(nk1)<q(K1(Knk1+kK1))q(G)2e(G)n1+n2.2(n-k-1)<q(K_{1}\vee(K_{n-k-1}+kK_{1}))\leq q(G)\leq\frac{2e(G)}{n-1}+n-2.

We can deduce that e(G)>[2(nk1)n+2](n1)2(nk12)+k2+k+1e(G)>\frac{[2(n-k-1)-n+2](n-1)}{2}\geq{n-k-1\choose 2}+k^{2}+k+1 for n32k2+32k+2n\geq\frac{3}{2}k^{2}+\frac{3}{2}k+2. Let H=Cn1(G)H=C_{n-1}(G). By Lemma 2.2, we have HK1(Knk1+kK1)H\cong K_{1}\vee(K_{n-k-1}+kK_{1}). Since GHG\subseteq H, then

q(G)q(H)=q(K1(Knk1+kK1)).q(G)\leq q(H)=q(K_{1}\vee(K_{n-k-1}+kK_{1})).

By the assumption, q(G)q(K1(Knk1+kK1))q(G)\geq q(K_{1}\vee(K_{n-k-1}+kK_{1})), and hence GHG\cong H. Recall that K1(Knk1+kK1)K_{1}\vee(K_{n-k-1}+kK_{1}) has no a spanning kk-ended-tree. Hence GK1(Knk1+kK1).G\cong K_{1}\vee(K_{n-k-1}+kK_{1}).

(iii) Let GG be a connected graph of order nn and k2k\geq 2 be an integer. Let H=Cn1(G).H=C_{n-1}(G). If HKn,H\cong K_{n}, then HH has a spanning kk-ended-tree. By Theorem 1.2, GG also has a spanning kk-ended-tree, and the result follows. Next we always assume that HKn.H\ncong K_{n}.

Suppose to the contrary that GG has no a spanning kk-ended-tree. By Theorem 1.2, HH has no a spanning kk-ended-tree either. By Theorem 1.1, we have

dH(u)+dH(v)nkd_{H}(u)+d_{H}(v)\leq n-k

for every pair of nonadjacent vertices uu and vv (always exists) of HH. Then we have

dH¯(u)+dH¯(v)=n1dH(u)+n1dH(v)n+k2d_{\overline{H}}(u)+d_{\overline{H}}(v)=n-1-d_{H}(u)+n-1-d_{H}(v)\geq n+k-2

for every edge uvE(H¯)uv\in E(\overline{H}). Notice that every non-trivial component of H¯\overline{H} has a vertex with the degree at least n+k22.\frac{n+k-2}{2}. Then its order is at least n+k2\frac{n+k}{2}, which implies that H¯\overline{H} has exactly one non-trivial component FF and V(H¯\F)V(\overline{H}\backslash F) is the set of isolated vertices. Hence

dF(u)+dF(v)=dH¯(u)+dH¯(v)n+k2d_{F}(u)+d_{F}(v)=d_{\overline{H}}(u)+d_{\overline{H}}(v)\geq n+k-2

for uvE(F).uv\in E(F). Since dH(u)dG(u)1d_{H}(u)\geq d_{G}(u)\geq 1 and dH(v)dG(v)1d_{H}(v)\geq d_{G}(v)\geq 1, then we have dF(u)=dH¯(u)n2d_{F}(u)=d_{\overline{H}}(u)\leq n-2 and dF(v)=dH¯(v)n2d_{F}(v)=d_{\overline{H}}(v)\leq n-2. Hence dF(u)n+k2dF(v)kd_{F}(u)\geq n+k-2-d_{F}(v)\geq k and dF(v)n+k2dF(u)k.d_{F}(v)\geq n+k-2-d_{F}(u)\geq k. That is, kdF(u)n2k\leq d_{F}(u)\leq n-2 and kdF(v)n2k\leq d_{F}(v)\leq n-2. It is easy to see that

dF(u)dF(v)dF(u)(n+k2dF(u))f(dF(u)).d_{F}(u)d_{F}(v)\geq d_{F}(u)(n+k-2-d_{F}(u))\triangleq f(d_{F}(u)).

Note that f(dF(u))f(d_{F}(u)) is a convex function on dF(u)[k,n2]d_{F}(u)\in[k,n-2]. Then

dH¯(u)dH¯(v)=dF(u)dF(v)min{f(k),f(n2)}=k(n2).d_{\overline{H}}(u)d_{\overline{H}}(v)=d_{F}(u)d_{F}(v)\geq\mbox{min}\{f(k),f(n-2)\}=k(n-2).

for every edge uvE(H¯)uv\in E(\overline{H}). By the assumption and Lemma 2.4, we have

k(n2)ρ(G¯)ρ(H¯)minuvE(H¯)dH¯(u)dH¯(v)k(n2).\sqrt{k(n-2)}\geq\rho(\overline{G})\geq\rho(\overline{H})\geq\mbox{min}_{uv\in E(\overline{H})}\sqrt{d_{\overline{H}}(u)d_{\overline{H}}(v)}\geq\sqrt{k(n-2)}.

Then all the above inequalities must be equalities. This implies that G¯H¯\overline{G}\cong\overline{H}, ρ(H¯)=k(n2),\rho(\overline{H})=\sqrt{k(n-2)}, and there exists an edge uvE(H¯)uv\in E(\overline{H}) such that dH¯(u)=kd_{\overline{H}}(u)=k and dH¯(v)=n2d_{\overline{H}}(v)=n-2. Note that FF is a unique non-trivial component of H¯.\overline{H}. Then

ρ(F)=ρ(H¯)=minuvE(H¯)dH¯(u)dH¯(v)=minuvE(F)dF(u)dF(v).\rho(F)=\rho(\overline{H})=\mbox{min}_{uv\in E(\overline{H})}\sqrt{d_{\overline{H}}(u)d_{\overline{H}}(v)}=\mbox{min}_{uv\in E(F)}\sqrt{d_{F}(u)d_{F}(v)}.

By Lemma 2.4, FF is regular or semi-regular bipartite.

Assume first that FF is semi-regular bipartite. By symmetry, we can assume that dF(u)=kd_{F}(u)=k and dF(v)=n2d_{F}(v)=n-2 for every edge uvE(F)uv\in E(F). Hence FKk,n2.F\cong K_{k,n-2}. It is easy to see that nk+(n2),n\geq k+(n-2), and thus k=2k=2 since k2k\geq 2. Then we have H¯=FK2,n2\overline{H}=F\cong K_{2,n-2}, and hence GHK2+Kn2G\cong H\cong K_{2}+K_{n-2}, which contradicts that GG is connected.

Hence FF is regular. Then dF(v)=k=n2d_{F}(v)=k=n-2 for each vV(F),v\in V(F), and hence n=k+2n=k+2. If V(H¯\F)=ϕV(\overline{H}\backslash F)=\phi, then H¯=F\overline{H}=F is an (n2)(n-2)-regular graph with nn vertices. Obviously, H¯\overline{H} is obtained from KnK_{n} by deleting a perfect matching of H¯\overline{H}. Then GHG\cong H is the perfect matching of G,G, which contradicts the connectedness of G.G. Next we consider V(H¯\F)ϕV(\overline{H}\backslash F)\neq\phi. Then H¯=F+K1,\overline{H}=F+K_{1}, where FKk+1F\cong K_{k+1}. So we have GHK1,k+1G\cong H\cong K_{1,k+1}. It is obvious that GG has no a spanning kk-ended-tree. Moveover, ρ(G¯)=ρ(Kk+1)=k=k(n2).\rho(\overline{G})=\rho(K_{k+1})=k=\sqrt{k(n-2)}. Hence GK1,k+1G\cong K_{1,k+1}.

(iv) Let H=Cn1(G).H=C_{n-1}(G). By the beginning of the proof in (iii), we still assume that HKn.H\ncong K_{n}. Suppose that GG has no a spanning kk-ended-tree. By Theorem 1.2, HH has no a spanning kk-ended-tree either. From the proof of (iii), we can obtain that H¯\overline{H} has exactly one non-trivial component FF and

dF(u)+dF(v)n+k2d_{F}(u)+d_{F}(v)\geq n+k-2

for each uvE(F)uv\in E(F), where kdF(u)n2k\leq d_{F}(u)\leq n-2 and kdF(v)n2k\leq d_{F}(v)\leq n-2. By the assumption and Lemma 2.6, we have

n+k2q(G¯)q(H¯)=q(F)minuvE(F)(dF(u)+dF(v))n+k2.n+k-2\geq q(\overline{G})\geq q(\overline{H})=q(F)\geq\mbox{min}_{uv\in E(F)}(d_{F}(u)+d_{F}(v))\geq n+k-2.

Then all the above inequalities must be equalities, which implies that G¯H¯\overline{G}\cong\overline{H}, q(F)=n+k2,q(F)=n+k-2, and there exists an edge uvE(F)uv\in E(F) such that dF(u)+dF(v)=n+k2d_{F}(u)+d_{F}(v)=n+k-2. Furthermore, we have q(F)=minuvE(F)(dF(u)+dF(v))q(F)=\mbox{min}_{uv\in E(F)}(d_{F}(u)+d_{F}(v)). By Lemma 2.6, FF is regular or semi-regular bipartite.

If FF is semi-regular bipartite, then FKt,n+k2tF\cong K_{t,n+k-2-t} with ktn2.k\leq t\leq n-2. Since nt+(n+k2t)n\geq t+(n+k-2-t) and k2k\geq 2, then we have k=2k=2. Hence H¯=FKt,nt,\overline{H}=F\cong K_{t,n-t}, where ktn2k\leq t\leq n-2. Then GHKt+KntG\cong H\cong K_{t}+K_{n-t} with ktn2k\leq t\leq n-2, which contradicts the connectedness of GG.

Hence FF is regular. Then FF is an n+k22\frac{n+k-2}{2}-regular graph with tt vertices, where n+k2tn\frac{n+k}{2}\leq t\leq n. It is easy to see that H¯=F+(nt)K1\overline{H}=F+(n-t)K_{1}, and hence GHF¯KntG\cong H\cong\overline{F}\vee K_{n-t}. Note that F¯\overline{F} is an (tn+k2)(t-\frac{n+k}{2})-regular graph with tt vertices. For convenience, let F¯=R(t,tn+k2)\overline{F}=R(t,t-\frac{n+k}{2}). Then we can write

GR(t,tn+k2)Knt,G\cong R(t,t-\frac{n+k}{2})\vee K_{n-t},

where n+k2tn,\frac{n+k}{2}\leq t\leq n, a contradiction. The proof is completed. \Box

3 Proof of Theorem 1.5

Recall the nonnegative matrix Aa(G)=aD(G)+A(G)(a0)A_{a}(G)=aD(G)+A(G)~{}(a\geq 0) of a graph GG. Let ρa(G)\rho_{a}(G) be the AaA_{a}-spectral radius of G.G. Obviously, ρ0(G)=ρ(G)\rho_{0}(G)=\rho(G) and ρ1(G)=q(G).\rho_{1}(G)=q(G). Now, we are ready to present the proof of Theorem 1.5.

Proof of Theorem 1.5. Suppose to the contrary that GG has no a spanning tree with leaf degree at most kk for n2k+12n\geq 2k+12 and k1k\geq 1. By Theorem 1.4, there exists some nonempty subset SV(G)S\subseteq V(G) such that i(GS)(k+1)|S|.i(G-S)\geq(k+1)|S|. Let |S|=s|S|=s. Then GG is a spanning subgraph of G=Ks(Kn(k+2)s+(k+1)sK1)G^{\prime}=K_{s}\vee(K_{n-(k+2)s}+(k+1)sK_{1}) (see Fig. 2). By Lemma 2.7, we have ρa(G)ρa(G)\rho_{a}(G)\leq\rho_{a}(G^{\prime}) for a{0,1}a\in\{0,1\}. We distinguish the proof into the following two cases.

Case 1. s=1.s=1.

Then GK1(Knk2+(k+1)K1).G^{\prime}\cong K_{1}\vee(K_{n-k-2}+(k+1)K_{1}). From the above, we know that

ρa(G)ρa(K1(Knk2+(k+1)K1)).\rho_{a}(G)\leq\rho_{a}(K_{1}\vee(K_{n-k-2}+(k+1)K_{1})).

By the assumption ρa(G)ρa(K1(Knk2+(k+1)K1))\rho_{a}(G)\geq\rho_{a}(K_{1}\vee(K_{n-k-2}+(k+1)K_{1})), then we have GK1(Knk2+(k+1)K1).G\cong K_{1}\vee(K_{n-k-2}+(k+1)K_{1}). Note that the vertices of (k+1)K1(k+1)K_{1} are only adjacent to a vertex of Knk1K_{n-k-1}. Then the leaf degree of any spanning tree of K1(Knk2+(k+1)K1)K_{1}\vee(K_{n-k-2}+(k+1)K_{1}) is at least k+1k+1. This implies that K1(Knk2+(k+1)K1)K_{1}\vee(K_{n-k-2}+(k+1)K_{1}) has no a spanning tree with leaf degree at most kk. Hence GK1(Knk2+(k+1)K1)G\cong K_{1}\vee(K_{n-k-2}+(k+1)K_{1}).

Case 2. s2.s\geq 2.

Note that G=Ks(Kn(k+2)s+(k+1)sK1)G^{\prime}=K_{s}\vee(K_{n-(k+2)s}+(k+1)sK_{1}) and e(G)=(n(k+1)s2)+(k+1)s2e(G^{\prime})={n-(k+1)s\choose 2}+(k+1)s^{2}. Next we will discuss the different values of aa.

Subcase 2.1. a=0.a=0.

By Lemma 2.3, we have

ρ0(G)\displaystyle\rho_{0}(G^{\prime}) \displaystyle\leq 2e(G)n+1\displaystyle\sqrt{2e(G^{\prime})-n+1}
=\displaystyle= (nkss)(nkss1)+2(k+1)s2n+1\displaystyle\sqrt{(n-ks-s)(n-ks-s-1)+2(k+1)s^{2}-n+1}
=\displaystyle= (k2+4k+3)s2(2kn+2nk1)s+n22n+1\displaystyle\sqrt{(k^{2}+4k+3)s^{2}-(2kn+2n-k-1)s+n^{2}-2n+1}
\displaystyle\triangleq f1(s).\displaystyle\sqrt{f_{1}(s)}.

Note that ns+(k+1)s=(k+2)sn\geq s+(k+1)s=(k+2)s. Then 2snk+2.2\leq s\leq\frac{n}{k+2}. Moreover, we claim that

max2snk+2f1(s)=f1(2).\mbox{max}_{2\leq s\leq\frac{n}{k+2}}f_{1}(s)=f_{1}(2).

In fact, let g(n)=f1(2)f1(nk+2)g(n)=f_{1}(2)-f_{1}(\frac{n}{k+2}), by a direct calculation, we have

g(n)=(k+1)2n2(4k3+21k2+35k+18)n+4k4+34k3+102k2+128k+56(k+2)2.g(n)=\frac{(k+1)^{2}n^{2}-(4k^{3}+21k^{2}+35k+18)n+4k^{4}+34k^{3}+102k^{2}+128k+56}{(k+2)^{2}}.

Note that g(n)g(n) is a monotonically increasing function on n[2k+12,+)n\in[2k+12,+\infty). Then

f1(2)f1(nk+2)=g(n)g(2k+12)=24k2+8k16(k+2)2>0.f_{1}(2)-f_{1}(\frac{n}{k+2})=g(n)\geq g(2k+12)=\frac{24k^{2}+8k-16}{(k+2)^{2}}>0.

This implies that max2snk+2f1(s)=f1(2).\mbox{max}_{2\leq s\leq\frac{n}{k+2}}f_{1}(s)=f_{1}(2). By the above proof, we have

ρ0(G)ρ0(G)\displaystyle\rho_{0}(G)\leq\rho_{0}(G^{\prime}) \displaystyle\leq f1(2)\displaystyle\sqrt{f_{1}(2)}
=\displaystyle= (nk2)2(2kn+2n3k214k11)\displaystyle\sqrt{(n-k-2)^{2}-(2kn+2n-3k^{2}-14k-11)}
\displaystyle\leq (nk2)2(k2+14k+13)\displaystyle\sqrt{(n-k-2)^{2}-(k^{2}+14k+13)}
<\displaystyle< nk2,\displaystyle n-k-2,

since n2k+12n\geq 2k+12. On the other hand, by the assumption and Lemma 2.7, we have

ρ0(G)ρ0(K1(Knk2+(k+1)K1))>ρ0(Knk1)=nk2,\rho_{0}(G)\geq\rho_{0}(K_{1}\vee(K_{n-k-2}+(k+1)K_{1}))>\rho_{0}(K_{n-k-1})=n-k-2,

a contradiction.

Refer to caption
Figure 2: Graph Ks(Kn(k+2)s+(k+1)sK1)K_{s}\vee(K_{n-(k+2)s}+(k+1)sK_{1}).

Subcase 2.2. a=1.a=1.

By Lemma 2.5, we have

ρ1(G)\displaystyle\rho_{1}(G^{\prime}) \displaystyle\leq 2e(G)n1+n2\displaystyle\frac{2e(G^{\prime})}{n-1}+n-2
=\displaystyle= (nkss)(nkss1)+2(k+1)s2n1+n2\displaystyle\frac{(n-ks-s)(n-ks-s-1)+2(k+1)s^{2}}{n-1}+n-2
=\displaystyle= (k2+4k+3)s2(2kn+2nk1)s+2n24n+2n1\displaystyle\frac{(k^{2}+4k+3)s^{2}-(2kn+2n-k-1)s+2n^{2}-4n+2}{n-1}
\displaystyle\triangleq f2(s)n1.\displaystyle\frac{f_{2}(s)}{n-1}.

Note that 2snk+22\leq s\leq\frac{n}{k+2}. Similar to the proof of Subcase 2.1, we can obtain that

max2snk+2f2(s)=f2(2).\mbox{max}_{2\leq s\leq\frac{n}{k+2}}f_{2}(s)=f_{2}(2).

Then

ρ1(G)ρ1(G)\displaystyle\rho_{1}(G)\leq\rho_{1}(G^{\prime}) \displaystyle\leq f2(2)n1\displaystyle\frac{f_{2}(2)}{n-1}
=\displaystyle= 2(nk2)(2k+2)n4k216k12n1\displaystyle 2(n-k-2)-\frac{(2k+2)n-4k^{2}-16k-12}{n-1}
\displaystyle\leq 2(nk2)12k+12n1\displaystyle 2(n-k-2)-\frac{12k+12}{n-1}
<\displaystyle< 2(nk2),\displaystyle 2(n-k-2),

since n2k+12n\geq 2k+12. However, by the assumption, we have

ρ1(G)ρ1(K1(Knk2+(k+1)K1))>ρ1(Knk1)=2(nk2),\rho_{1}(G)\geq\rho_{1}(K_{1}\vee(K_{n-k-2}+(k+1)K_{1}))>\rho_{1}(K_{n-k-1})=2(n-k-2),

a contradiction. We complete the proof. \Box

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References

  • [1] A. Berman, R.J. Plemmons, Nonnegative matrices in the mathematical sciences, Academic Press, New York, 1979.
  • [2] A.E. Brouwer, W.H. Haemers, Spectra of Graphs, Springer, Berlin, 2011.
  • [3] H. Broersma, H. Tuinstra, Independence trees and Hamilton cycles, J. Graph Theory. 29 (1998) 227–237.
  • [4] J.A. Bondy, V. Chvatal, A method in graph theory, Discrete Math. 15 (1976) 111–135.
  • [5] J.A. Bondy, U.S.R. Murty, Graph Theory, Grad. Texts in Math. vol. 244, Springer, New York, 2008.
  • [6] V. Chvátal, P. Erdős, A note on hamiltonian circuits, Discrete Math. 2 (1972) 111–113.
  • [7] X.D. Chen, H.Q. Liu, F.L. Lu, M.D. Liu, Spanning kk-ended trees in quasi-claw-free graphs, Bull. Malays. Math. Sci. Soc. 43 (2020) 241–252.
  • [8] Y. Chen, G.T. Chen, Z.Q. Hu, Spanning 3-ended trees in kk-connected K1,4K_{1,4}-free graphs, Sci. China Math. 57 (2014) 1579–1586.
  • [9] Y. Chen, P.H. Ha, D.D. Hanh, Spanning trees with at most 4 leaves in K1,5K_{1,5}-free graphs, Discrete Math. 342 (2019) 2342–2349.
  • [10] K.C. Das, Maximizing the sum of the squares of the degrees of a graph, Discrete Math. 285 (2004) 57–66.
  • [11] D.D. Fan, S. Goryainov, X.Y. Huang, H.Q. Lin, The spanning kk-trees, perfect matchings and spectral radius of graphs, Linear Multilinear Algebra, 2021, doi.org/10.1080/03081087.2021.1985055.
  • [12] E. Flandrin, T. Kaiser, R. Kužel, H. Li, Z. Ryjáček, Neighborhood unions and extremal spanning trees, Discrete Math. 308 (2008) 2343–2350.
  • [13] L.H. Feng, G.H. Yu, On three conjectures involving the signless Laplacian spectral radius of graphs, Publ. Inst. Math. (Beograd) 85 (2009) 35–38.
  • [14] M. Fiedler, V. Nikiforov, Spectral radius and Hamiltonicity of graphs, Linear Algebra Appl. 432 (2010) 2170–2173.
  • [15] R.A. Horn, C.R. Johnson, Matrix analysis, Cambridge University Press, New York, 1986.
  • [16] Y. Hong, A bound on the spectral raoius of graphs, Linear Algebra Appl. 108 (1988) 135–139.
  • [17] A. Kaneko, Spanning trees with constraints on the leaf degree, Discrete Appl. Math. 115 (2001) 73–76.
  • [18] A. Kyaw, Spanning trees with at most 3 leaves in K1,4K_{1,4}-free graphs, Discrete Math. 309 (2009) 6146–6148.
  • [19] A. Kyaw, Spanning trees with at most kk leaves in K1,4K_{1,4}-free graphs, Discrete Math. 311 (2011) 2135–2142.
  • [20] B. Ning, J. Ge, Spectral radius and Hamiltonian properties of graphs, Linear Multilinear Algebra 63 (2015) 1520–1530.
  • [21] B.L. Li, B. Ning, Spectral analogues of Erdős’ and Moon-Moser’s theorems on Hamilton cycles, Linear Algebra Appl. 64 (2016) 2252–2269.
  • [22] M. Lu, H.Q. Liu, F. Tian, Spectral radius and Hamiltonian graphs, Linear Algebra Appl. 437 (2012) 1670–1674.
  • [23] M.H. Liu, H.-J. Lai, K.C. Das, Spectral results on Hamiltonian problem, Discrete Math. 342 (2019) 1718–1730.
  • [24] R.F. Liu, W.C. Shiu, J. Xue, Sufficient spectral conditions on Hamiltonian and traceable graphs, Linear Algebra Appl. 467 (2015) 254–266.
  • [25] O. Ore, Note on hamilton circuits, Amer. Math. Monthly 67 (1960) 55.
  • [26] M. Tsugaki, T. Yamashita, Spanning trees with few leaves, Graphs Combin. 23 (2007) 585–598.
  • [27] S. Win, On a conjecture of Las Vergnas concerning certain spanning trees in graphs, Resultate Math. 2 (1979) 215–224.
  • [28] B. Zhou, Signless Laplacian spectral radius and Hamiltonicity, Linear Algebra Appl. 432 (2010) 566–570.
  • [29] Q.N. Zhou, L.G. Wang, Some sufficient spectral conditions on Hamilton-connected and traceable graphs, Linear Multilinear Algebra 65 (2017) 224–234.