1 Introduction

Let \(G=(V, E)\) be a simple undirected graph with vertex set \(V= \{v_1, v_2, \ldots , v_n\}\) and edge set \(E= \{e_1, e_2, \ldots , e_m\}\), i.e., \(|V|=n\) and \(|E|=m\). If \(m=n+c-1\), then G is called a c-cyclic graph. Particularly, if \(c=0, 1, 2, 3\), then G is called a tree, a unicyclic graph, a bicyclic graph, a tricyclic graph, respectively. For \(v\in V(G)\), N(v) denotes the neighborhood of v in G and \(d(v)=|N(v)|\) denotes the degree of vertex v. If \(d_i=d_G(v_i)\) for \(1\le i\le n\), then we call the sequence \(\pi =(d_1, d_2, \ldots , d_n)\) the degree sequence of G. Throughout this paper, we enumerate the degrees in non-increasing order, i.e., \(d_1\ge d_2\ge \cdots \ge d_n\). A non-increasing sequence \(\pi =(d_1, d_2, \ldots , d_n)\) is called graphic if there exists a graph G having \(\pi \) as its degree sequence. We use \(C_{\pi }\) to denote the class of connected graphs with degree sequence \(\pi \).

For a graph G, A(G) is its adjacency matrix and D(G) is the diagonal matrix of its degrees. The matrix \(Q(G)=D(G)+A(G)\) is called the signless Laplacian matrix of G. The largest eigenvalue of A(G) (resp., Q(G)) is called the spectral radius (resp., signless Laplacian spectral radius) of G and denoted by \(\rho (G)\) (resp., \(\mu (G)\)). If G is connected, then A(G) (resp., Q(G)) is irreducible and by the Perron–Frobenius theorem, \(\rho (G)\) (resp., \(\mu (G)\)) has multiplicity one and there exists a unique positive unit eigenvector corresponding to \(\rho (G)\) (resp., \(\mu (G)\)). In this paper, we use \(f = \big (f(v_{1}),f(v_{2}), \ldots , f(v_{n})\big )^{T}\) to indicate the unique positive unit eigenvector corresponding to \(\rho (G)\) (resp., \(\mu (G)\)), and call f the Perron vector of A(G) (resp., Q(G)). Furthermore, if \(\rho (G)\) (resp., \(\mu (G)\)) is greatest in \(C_{\pi }\), then G is called an extremal greatest graph of \(C_{\pi }\) for \(\rho (G)\) (resp., \(\mu (G)\)).

Suppose \(\pi =(d_1, d_2, \ldots , d_n)\) and \(\pi ^{\,\prime }=(d_1^{\,\prime }, d_2^{\,\prime }, \ldots , d_n^{\,\prime })\) are two non-increasing degree sequences. We say \(\pi \) is majorizated by \(\pi ^{\,\prime }\), denoted by \(\pi \lhd \pi ^{\,\prime }\), if and only if \(\pi \ne \pi ^{\,\prime }\), \(\sum \nolimits _{i=1}^{n} d_i=\sum \nolimits _{i=1}^{n} d_i^{\,\prime }\), and \(\sum \nolimits _{i=1}^j d_i\le \sum \nolimits _{i=1}^j d_i^{\,\prime }\) for all \(j=1, 2, \ldots , n-1\).

In 2008, Bıyıkoğlu and Leydold connected the majorization of degree sequences with ordering graphs by their spectral radius, and they obtained the majorization theorem of trees as follows.

Theorem 1.1

[1] Let \(\pi \) and \(\pi ^{\,\prime }\) be two different non-increasing degree sequences of trees with \(\pi \lhd \pi ^{\,\prime }\). Suppose T and \(T^{\,\prime }\) are the trees with the greatest spectral radius in \(C_{\pi }\) and \(C_{\pi ^{\,\prime }}\), respectively. Then, \(\rho (T)<\rho (T^{\,\prime })\).

Almost at the same time, Zhang [15] proved the majorization theorem for the Laplacian spectral radius of trees. In the sequel, similar problems have been studied extensively. Liu et al. [9] and Zhang [16] proved the majorization theorems for the spectral radius and signless Laplacian spectral radius of unicyclic graphs, respectively. Jiang et al. [6] and Huang et al. [5] proved the majorization theorems for the spectral radius and signless Laplacian spectral radius of bicyclic graphs, respectively, and Jiang et al. [6] provided a counterexample to show that the majorization theorem cannot hold for tricyclic graphs. Liu and Liu et al. [7, 9, 12, 13] proved the majorization theorems for the spectral radius and signless Laplacian spectral radius of c-cyclic graphs with additional restrictions, respectively. Recently, Liu et al. [10] proved the majorization theorems for the spectral radius and signless Laplacian spectral radius of pseudographs. For more results, one may refer to [8, 11].

Let \(G_1\) and \(G_2\) be graphs with disjoint vertex sets, and \(G_1\vee G_2\) denote the join of \(G_1\) and \(G_2\). If G is connected, then \(K_1\vee G\) is called a single-cone graph. In particular, if G is a tree (unicyclic graph) of order \(n-1\), then we call \(K_1\vee G\) a single-cone tree (unicyclic graph) of order n. For a non-increasing graphic sequence \(\pi =(d_1, d_2, \ldots , d_n)\), let

$$\begin{aligned} J_{\pi }=\{G: \,G\,\, \text{ is } \text{ a } \text{ single-cone } \text{ graph } \text{ with } \text{ degree } \text{ sequence }\,\, \pi \}. \end{aligned}$$

If \(G\in J_{\pi }\) and \(\rho (G)\ge \rho (G^{\,\prime })\) (resp., \(\mu (G)\ge \mu (G^{\,\prime })\)) for any other \(G^{\,\prime }\in J_{\pi }\), then we call G has the greatest spectral radius (resp., signless Laplacian spectral radius) in \(J_{\pi }\).

In this paper, we give the majorization theorems for the spectral radius and signless Laplacian spectral radius of single-cone trees and single-cone unicyclic graphs, and the main results can be stated as follows:

Theorem 1.2

Let \(\pi \) and \(\pi ^{\,\prime }\) be two different non-increasing degree sequences of single-cone trees with \(\pi \lhd \pi ^{\,\prime }\). Suppose G and \(G^{\,\prime }\) are the single-cone trees with the greatest spectral radius (resp., signless Laplacian spectral radius) in \(J_{\pi }\) and \(J_{\pi ^{\,\prime }}\), respectively. Then, \(\rho (G)<\rho (G^{\,\prime })\) (resp., \(\mu (G)<\mu (G^{\,\prime })\)).

Theorem 1.3

Let \(\pi \) and \(\pi ^{\,\prime }\) be two different non-increasing degree sequences of single-cone unicyclic graphs with \(\pi \lhd \pi ^{\,\prime }\). Suppose G and \(G^{\,\prime }\) are the single-cone unicyclic graphs with the greatest spectral radius (resp., signless Laplacian spectral radius) in \(J_{\pi }\) and \(J_{\pi ^{\,\prime }}\), respectively. Then, \(\rho (G)<\rho (G^{\,\prime })\) (resp., \(\mu (G)<\mu (G^{\,\prime })\)).

The rest of the paper is organized as follows. In Sect. 2, we recall some basic notions and lemmas used further, and prove a new lemma. In Sect. 3, we give the proof of Theorem 1.2. In Sect. 4, we give the proof of Theorem 1.3.

2 Preliminaries

Given a unit n-vector \(g=(g_1, g_2, \ldots , g_n)^T \in \mathbf{R}^n\), g can be considered as a function defined on V(G), that is, each vertex \(v_i\) is mapped to \(g_i =g(v_i)\). The Rayleigh quotients of the adjacency matrix A(G) and signless Laplacian matrix Q(G) are defined, respectively, as follows:

$$\begin{aligned} {\mathcal {R}}_{A(G)}(g)=\sum _{uv\in E}2g(u)g(v)\,\,\text { and }\,\,{\mathcal {R}}_{Q(G)}(g)=\sum _{uv\in E}(g(u)+g(v))^2. \end{aligned}$$

It follows from the Rayleigh–Ritz theorem that

Lemma 2.1

[3, 4] Let S denote the set of unit vectors on V. Then,

$$\begin{aligned} \rho (G)= & {} \max \limits _{g\in S} {\mathcal {R}}_{A(G)}(g)=2\max \limits _{g\in S} \sum \limits _{uv\in E} g(u)g(v),\\ \mu (G)= & {} \max \limits _{g\in S} {\mathcal {R}}_{Q(G)}(g)=\max \limits _{g\in S} \sum \limits _{uv\in E} (g(u)+g(v))^2. \end{aligned}$$

Moreover, if \({\mathcal {R}}_{A(G)}(g)=\rho (G)\) (resp., \({\mathcal {R}}_{Q(G)}(g)=\mu (G)\)) for a positive vector \(g\in S\), then g is an eigenvector corresponding to \(\rho (G)\) (resp., \(\mu (G)\)).

Let \(f=(f(v_1), f(v_2), \ldots , f(v_n))^T\) be the Perron vector of A(G) (resp., Q(G)). Then, \(\rho (G)f(v)=\sum \nolimits _{uv\in E} f(u)\) (resp., \(\mu (G)f(v)\)\(=d(v)f(v)+\sum _{uv\in E} f(u)\)) for \(v\in V(G)\). We will refer to such an equation as the eigenvalue equation of \(\rho (G)\) (resp., \(\mu (G)\)). Let \(G-u\) denote the graph that arises from G by deleting the vertex \(u\in V(G)\) and all the edges incident with u, and \(G-uv\) denote the graph that arises from G by deleting the edge \(uv\in E(G)\). Similarly, \(G+uv\) is the graph that arises from G by adding an edge \(uv\notin E(G)\), where \(u, v\in V(G)\).

In order to complete the proof of the theorems, we also need the following definition and lemmas.

Lemma 2.2

[4, 14] Let uv be two vertices of the connected graph G. Suppose \(v_1, v_2, \ldots , \)\(v_s\in N_G(v)\backslash N_G(u)~(1\le s\le d_G(v))\), and f is the Perron vector of A(G) (resp., Q(G)). Let \(G^{\,'}\) be the graph obtained from G by deleting the edges \(vv_i\) and adding the edges \(uv_i~(1\le i\le s)\). If \(f(u)\ge f(v)\), then \(\rho (G^{\,'})>\rho (G)\) (resp., \(\mu (G^{\,'})>\mu (G)\)).

Lemma 2.3

[1, 15] Let G be a connected graph of order n such that \(v_1v_3\), \(v_2v_4\in E(G)\), \(v_1v_2\), \(v_3v_4\notin E(G)\). Let \(G^{\,'}=G-v_1v_3-v_2v_4+v_1v_2+v_3v_4\). Suppose f is the Perron vector of A(G) (resp., Q(G)); if \(f(v_1)\ge f(v_4)\) and \(f(v_2)\ge f(v_3)\), then \(\rho (G^{\,'})\ge \rho (G)\) (resp., \(\mu (G^{\,'})\ge \mu (G)\)), where the equalities hold if and only if \(f(v_1)=f(v_4)\) and \(f(v_2)=f(v_3)\).

Lemma 2.4

[8] Let G be a connected graph and f be the Perron vector of A(G) (resp., Q(G)). Let \(G\,'\) be a connected graph obtained from G by deleting \(t \, (\ge 1)\) edges and adding another t new edges such that \(G\not \cong G\,'\). Suppose that there exists a vertex \(v\in V(G)\) such that \( N_{G}(v)\subset N_{G\,'}(v)\) or \( N_{G\,'}(v)\subset N_{G}(v)\). If \({\mathcal {R}}_{A(G\,')}(f)\ge {\mathcal {R}}_{A(G)}(f)\) (resp., \({\mathcal {R}}_{Q(G\,')}(f)\ge {\mathcal {R}}_{Q(G)}(f)\)), then \(\rho (G\,')>\rho (G)\) (resp., \(\mu (G\,')>\mu (G)\)).

Lemma 2.5

[2, 9] Let \(\pi =(d_1, d_2, \ldots , d_n)\) be a sequence with \(d_1\ge d_2\ge \cdots \ge d_n>0\). Then, \(\pi \) is graphic if and only if

$$\begin{aligned} \sum \limits _{i=1}^{n} d_i \text{ is } \text{ even } \text{ and } \sum \limits _{i=1}^{k} d_i\le k(k-1)+\sum \limits _{i=k+1}^{n} \min \{d_i, k\} \quad \text{ for } \text{ all } \quad k=1, 2, \ldots , n-1. \end{aligned}$$

Lemma 2.6

[12] Let \(\pi =(d_1, d_2, \ldots , d_n)\) and \(\pi ^{\,\prime }=(d_1^{\,\prime }, d_2^{\,\prime }, \ldots , d_n^{\,\prime })\) be two non-increasing degree sequences with \(\pi \lhd \pi ^{\,\prime }\). Then, \(d_n\ge d_n^{\,\prime }\).

Lemma 2.7

[2, 15] Let \(\pi \) and \(\pi ^{\,\prime }\) be two non-increasing graphic degree sequences. If \(\pi \lhd \pi ^{\,\prime }\), then there exists a series non-increasing graphic degree sequences \(\pi _1, \ldots , \pi _k\) such that \((\pi =)\)\(\pi _0\lhd \pi _1\lhd \cdots \lhd \pi _k \lhd \pi _{k+1}(=\pi ^{\,\prime })\), and \(\pi _i\) and \(\pi _{i+1}\) differ only at two positions, where the differences are 1 for \(0\le i\le k\).

Definition 2.8

[6, 10] Let \(\pi =(d_1, d_2, \ldots , d_n)\) and \(\pi ^{\,\prime }=(d_1^{\,\prime }, d_2^{\,\prime }, \ldots , d_n^{\,\prime })\) be two different non-increasing degree sequences. We say \(\pi \) is star majorizated by \(\pi ^{\,\prime }\), denoted by \(\pi \lhd ^* \pi ^{\,\prime }\), if and only if \(\pi \lhd \pi ^{\,\prime }\) and only two components of \(\pi \) and \(\pi ^{\,\prime }\) are different by 1, that is, \(d_i=d_i^{\, \prime }\) for \(i\ne k, l\), \(1\le k<l\le n\) and \(d_k^{\, \prime }=d_k+1\), \(d_l^{\, \prime }=d_l-1\).

Lemma 2.9

Suppose \(\pi =(d_1, d_2, \ldots , d_n)\) is a non-increasing degree sequence. If G is an extremal greatest single-cone graph for \(\rho (G)\) (resp., \(\mu (G)\)) in \(J_{\pi }\) with Perron vector f, then there exists an ordering of the vertices of G such that \(d(v_i)=d_i\) for \(1\le i\le n\) and \(f(v_1)\ge f(v_2)\ge \cdots \ge f(v_n)\).

Proof

Since G is a single-cone graph, there exists a vertex \(v_1\) such that \(d_G(v_1)=n-1\) and \(G-v_1\) is a connected graph. Create an ordering of the vertices of G beginning with \(v_1\) and appending other vertices after it. We use the notation \(v_i\prec v_j\) to indicate that the vertex \(v_i\) precedes the vertex \(v_j\) in the ordering of vertices. Clearly, \(v_1\prec v_i\) for \(i=2, 3, \ldots , n\). The order of other vertices is defined as follows: if \(d_G(v_i)>d_G(v_j)\), or \(d_G(v_i)=d_G(v_j)\) and \(f(v_i)\ge f(v_j)\), then \(v_i\prec v_j\). It is easy to see that this ordering satisfies \(d_G(v_1)\ge d_G(v_2)\ge \cdots \ge d_G(v_n)\). We will prove that \(f(v_1)\ge f(v_2)\ge \cdots \ge f(v_n)\).

Firstly, we claim that \(f(v_1)\ge f(v_i)\) for \(2\le i\le n\). Otherwise, we suppose that there exists some vertex \(v_i\) such that \(f(v_1)<f(v_i)\). If \(d_G(v_i)=n-1\), then \(N_G(v_1){\setminus } \{v_i\}=N_G(v_i){\setminus } \{v_1\}\). By the eigenvalue equation of \(\rho (G)\) (resp., \(\mu (G)\)), we have \(f(v_1)=f(v_i)\), contradicting \(f(v_1)<f(v_i)\). If \(d_G(v_i)<n-1\), then \(N_G(v_1){\setminus } (N_G(v_i)\cup \{v_i\})\ne \varnothing \). Let

$$\begin{aligned} G^{\,\prime }=G-\sum _{u\in N_G(v_1){\setminus } (N_G(v_i)\cup \{v_i\})}v_1u+\sum _{u\in N_G(v_1){\setminus } (N_G(v_i)\cup \{v_i\})}v_iu. \end{aligned}$$

Then, \(d_{G^{\,\prime }}(v_i)=d_{G}(v_1)\), \(d_{G^{\,\prime }}(v_1)=d_G(v_i)\) and \(d_{G^{\,\prime }}(v)=d_G(v)\) for \(v\in V(G){\setminus } \{v_1, v_i\}\). Noting that \(G-v_1\) is a connected graph and the neighbors of \(v_i\) in \(G-v_1\) are adjacent to \(v_1\) in \(G^{\,\prime }-v_i\), we have \(G^{\,\prime }-v_i\) which is a connected graph. This implies that \(G^{\,\prime }\in J_{\pi }\). By Lemma 2.2, we have \(\rho (G^{\,\prime })>\rho (G)\) ( resp., \(\mu (G^{\,\prime })>\mu (G)\)), a contradiction because G is an extremal greatest single-cone graph for \(\rho (G)\) (resp., \(\mu (G)\)) in \(J_{\pi }\). Therefore, \(f(v_1)\ge f(v_i)\) for \(2\le i\le n\).

Secondly, we show that \(v_i\prec v_j\) implies \(f(v_i)\ge f(v_j)\) for all \(v_i, v_j\in V(G){\setminus } \{v_1\}\). Otherwise, we suppose that there exist two vertices such that \(v_i\prec v_j\) but \(f(v_j)>f(v_i)\). Then, \(d(v_i)\ge d(v_j)\). Noting that \(v_i, v_j\in V(G){\setminus } \{v_1\}\), there exists a shortest path \(P_{ij}\) from \(v_i\) to \(v_j\) in \(G-v_1\). If \(d_G(v_i)>d_G(v_j)\), let \(k=d_G(v_i)-d_G(v_j)\), \(v_l\in V(P_{ij})\) and \(v_lv_i\in E(G)\). Then, there exist k vertices \(u_1, \ldots , u_k\in N_G(v_i){\setminus } (N_G(v_j)\cup \{v_l\})\). Let

$$\begin{aligned} G^{\,\prime }=G-\sum _{s=1}^k v_iu_s+\sum _{s=1}^k v_ju_s. \end{aligned}$$

Then, \(G^{\,\prime }\in J_{\pi }\). By Lemma 2.2, we have \(\rho (G^{\,\prime })>\rho (G)\) (resp., \(\mu (G^{\,\prime })>\mu (G)\)), a contradiction because G is an extremal greatest single-cone graph for \(\rho (G)\) (resp., \(\mu (G)\)) in \(J_{\pi }\). If \(d_G(v_i)=d_G(v_j)\), noting that \(v_i\prec v_j\), we have \(f(v_i)\ge f(v_j)\), contradicting \(f(v_j)> f(v_i)\).

Combining the above arguments, we have \(f(v_1)\ge f(v_2)\ge \cdots \ge f(v_n)\). \(\square \)

3 The proof of Theorem 1.2

The proof of Theorem 1.2

Since \(\pi \lhd \pi ^{\,\prime }\), it follows from Lemma 2.7 and Definition 2.8 that there exists a series non-increasing graphic degree sequences \(\pi _1, \ldots , \pi _k\) such that \((\pi =)\)\(\pi _0\lhd ^* \pi _1\lhd ^* \cdots \lhd ^* \pi _k \lhd ^* \pi _{k+1}(=\pi ^{\,\prime })\). Let \(\pi _i=(d_1^{\,(i)}, d_2^{\,(i)}, \ldots , d_n^{\,(i)})\) for \(0\le i\le k\). Clearly, \(d_1=d_1^{\,(1)}=\cdots =d_1^{\,(k)}=d_1^{\,\prime }=n-1\). By Lemma 2.6, we have \(d_n\ge d_n^{\, (1)}\ge \cdots \ge d_n^{\,(k)}\ge d_n^{\,\prime }\). Noting that \(\pi \) and \(\pi '\) are two different non-increasing degree sequences of single-cone trees, we have \(d_n=d_n^{\,\prime }=2\). This implies that \(d_n=d_n^{\,(1)}=\cdots =d_n^{\, (k)}=d_n^{\,\prime }=2\).

Since \(\pi \lhd ^* \pi _1\), without loss of generality, we suppose that \(d_k+1=d_k^{\,(1)}\), \(d_l-1=d_l^{\,(1)}\), \(d_j=d_j^{\,(1)}\) for \(j\notin \{ k,l\}\), and \(1<k<l<n\). Let f be the Perron vector of A(G) (resp., Q(G)). Then, Lemma 2.9 implies that there exists an ordering of the vertices of G such that \(d(v_i)=d_i\) for \(1\le i\le n\) and \(f(v_1)\ge f(v_2)\ge \cdots \ge f(v_n)\). Particularly, \(f(v_k)\ge f(v_l)\).

Assume that \(G-v_1\) is a tree and \(P_{kl}\) is a shortest path from \(v_k\) to \(v_l\) in \(G-v_1\). Noting that \(d_l=d_l^{\,(1)}+1>2\), there must exist some \(w\in N_{G-v_1}(v_l){\setminus }{N_{G-v_1}(v_k)}\) such that \(w\notin V(P_{kl})\). Let \(G_1=G-v_lw+v_kw\). Then, \(G_1-v_1\) is a tree, \(d_{G_1}(v_k)=d_G(v_k)+1\), \(d_{G_1}(v_l)=d_G(v_l)-1\), and \(d_{G_1}(v)=d_G(v)\) for \(v\in V(G){\setminus } \{v_k, v_l\}\). This implies that \(G_1\) is a single-cone tree and \(G_1\in J_{\pi _1}\). Noting that \(f(v_k)\ge f(v_l)\), by Lemma 2.2, we have \(\rho (G)<\rho (G_1)\) (resp., \(\mu (G)<\mu (G_1)\)). Let \(G_1^*\) be the single-cone tree with the greatest spectral radius (resp., signless Laplacian spectral radius) in \(J_{\pi _1}\). Then, \(\rho (G)<\rho (G_1)\le \rho (G_1^*)\) (resp., \(\mu (G)<\mu (G_1)\le \mu (G_1^*)\)).

By a similar reasoning as the above, we can obtain that \(\pi _i\) is a non-increasing degree sequence of a single-cone tree for each \(2\le i\le k\). Let \(G_i^*\) be a single-cone tree with the greatest spectral radius (resp., signless Laplacian spectral radius) in \(J_{\pi _i}\). Then, we have \(\rho (G)<\rho (G_1^*)<\cdots<\rho (G_k^*)<\rho (G^{\,\prime })\) (resp., \(\mu (G)<\mu (G_1^*)<\cdots<\mu (G_k^*)<\mu (G^{\,\prime }))\). \(\square \)

4 The proof of Theorem 1.3

Lemma 4.1

Let \(\pi =(d_1, d_2, \ldots , d_n)\) be a non-increasing degree sequence of a single-cone unicyclic graph, and G be an extremal greatest single-cone unicyclic graph for \(\rho (G)\) (resp., \(\mu (G)\)) in \(J_{\pi }\). Suppose \(\pi ^{\,\prime }=(d_1^{\,\prime }, d_2^{\,\prime }, \ldots , d_n^{\,\prime })\) (\(d_n^{\,\prime }\ge 2\)) is a non-increasing graphic degree sequence such that \(\pi \lhd ^* \pi ^{\,\prime }\). Then, there exists a single-cone unicyclic graph \(G^{\,\prime }\in J_{\pi ^{\,\prime }}\) such that \(\rho (G)<\rho (G^{\,\prime })\) (resp., \(\mu (G)<\mu (G^{\,\prime })\)).

Proof

Since \(\pi \lhd ^* \pi ^{\,\prime }\), without loss of generality, we suppose that \(d_k+1=d_k^{\,\prime }\), \(d_l-1=d_l^{\,\prime }\), and \(d_i=d_i^{\,\prime }\) for \(i\ne k,l\). Since \(\pi \) is a non-increasing degree sequence of a single-cone unicyclic graph, then \(d_1=d_1^{\,\prime }=n-1\), \(d_i\ge 2\) for \(1\le i\le n\), and \(1<k<l\le n\). Assume that \(G-v_1\) is a unicyclic graph. Let \(P_{kl}\) be a shortest path from \(v_k\) to \(v_l\) in \(G-v_1\), \(u\in N_{G-v_1}(v_l)\cap V(P_{kl})\) and f be the Perron vector of A(G) (resp., Q(G)). By Lemma 2.9, there exists an ordering of the vertices of G such that \(d(v_i)=d_i\) for \(1\le i\le n\) and \(f(v_1)\ge f(v_2)\ge \cdots \ge f(v_n)\). Particularly, \(f(v_k)\ge f(v_l)\).

Case 1:

\(N_{G-v_1}(v_l){\setminus } (N_{G-v_1}(v_k)\cup \{u\})\ne \emptyset \). Assume \(w\in N_{G-v_1}(v_l){\setminus } (N_{G-v_1}(v_k)\cup \{u\})\). Let \(G^{\,\prime }=G-v_lw+v_kw\). Then, \(G^{\,\prime }-v_1\) is a unicyclic graph, \(d_{G^{\,\prime }}(v_k)=d_G(v_k)+1\), \(d_{G^{\,\prime }}(v_l)=d_G(v_l)-1\), and \(d_{G^{\,\prime }}(v)=d_G(v)\) for \(v\in V(G){\setminus } \{v_k, v_l\}\). This implies that \(G^{\,\prime }\) is a single-cone unicyclic graph and \(G^{\,\prime }\in J_{\pi '}\). Noting that \(f(v_k)\ge f(v_l)\), by Lemma 2.2, we have \(\rho (G)<\rho (G^{\,\prime })\) (resp., \(\mu (G)<\mu (G^{\,\prime })\)).

Case 2:

\(N_{G-v_1}(v_l){\setminus } (N_{G-v_1}(v_k)\cup \{u\})=\emptyset \). Noting that G is a single-cone unicyclic graph, we have \(d_l\le 3\), and so \(d_l^{\,\prime }\le 2\). Since \(d_l^{\,\prime }\ge d_n^{\,\prime } \ge 2\), it follows that \(d_l^{\,\prime }=2\), \(d_l=3\), \(d_i=d_i^{\,\prime }=2\) for \(l+1\le i\le n\). Let \(w\in N_{G-v_1}(v_l){\setminus } \{u\}\). Then, \(w\in N_{G-v_1}(v_k)\). This implies that \(|V(P_{kl})|\le 3\).

Subcase 2.1:

\(|V(P_{kl})|=2\). In this case, \(C_3=v_kwv_lv_k\) is the unique cycle of \(G-v_1\). We claim that \(l\ge 5\). Otherwise, we suppose \(l\le 4\). Noting that \(\pi ^{\,\prime }\) is a non-increasing degree sequence and \(d_l^{\,\prime }=2\), we have \(d_i^{\,\prime }=2\) for \(4\le i\le n\). By \(\pi \lhd \pi ^{\,\prime }\), we have \(\sum \nolimits _{i=1}^{n} d_i^{\,\prime }=\sum \nolimits _{i=1}^{n} d_i=2(2n-2)\). It follows that

$$\begin{aligned} \sum \limits _{i=1}^{3} d_i^{\,\prime }=\sum \limits _{i=1}^{n} d_i^{\,\prime }-\sum \limits _{i=4}^{n} d_i^{\,\prime }=2n+2>2n=3(3-1)+\sum \limits _{i=4}^{n} \min \{d_i^{\,\prime }, 3\}, \end{aligned}$$

a contradiction to Lemma 2.5. Therefore, \(l\ge 5\). This implies that there must exist vertices ab such that \(a\notin \{v_1, v_k, v_l, w\}\), \(d_G(a)\ge 3\), \(d_G(b)=2\), and \(ab\in E(G)\).

If \(av_k\notin E(G)\), noting that \(d_G(v_l)=3\) and \(d_G(b)=2\), we have \(f(v_l)\ge f(b)\), \(av_l\notin E(G)\), and \(bv_k\notin E(G)\). We claim that \(f(v_k)\ge f(a)\). Otherwise, we suppose \(f(a)>f(v_k)\). Let \(G^*=G-ab-v_lv_k+av_l+bv_k\). By Lemma 2.3, we have \(\rho (G)<\rho (G^*)\) (resp., \(\mu (G)<\mu (G^*)\)). It is easy to see that \(G^*\in J_{\pi }\), which is a contradiction because G has the greatest spectral radius (resp., signless Laplacian spectral radius) in \(J_{\pi }\). Therefore, \(f(v_k)\ge f(a)\). Noting that \(d_i=d_i^{\,\prime }=2\) for \(l+1\le i\le n\), we have \(a\prec v_l\). It follows that \(f(a)\ge f(v_l)\). Let \(G^{\,\prime }=G-v_kv_l-ab+v_ka+v_kb\). Then, \(d_{G^{\,\prime }}(v_k)=d_G(v_k)+1\), \(d_{G^{\,\prime }}(v_l)=d_G(v_l)-1\), and \(d_{G^{\,\prime }}(v)=d_G(v)\) for \(v\in V(G){\setminus } \{v_k, v_l\}\). It is not difficult to see that \(G^{\,\prime }-v_1\) is a unicyclic graph, \(G^{\,\prime }\in J_{\pi '}\),

$$\begin{aligned} {\mathcal {R}}_{A(G^{\,\prime })}(f)-{\mathcal {R}}_{A(G)}(f)= & {} 2\sum \limits _{xy\in E(G^{\,\prime })} f(x)f(y)-2\sum \limits _{xy\in E(G)} f(x)f(y)\\= & {} 2f(v_k)f(a)+2f(v_k)f(b)-2f(a)f(b)-2f(v_k)f(v_l)\\= & {} 2f(v_k)(f(a)-f(v_l))+2f(b)(f(v_k)-f(a))\ge 0, \end{aligned}$$

and

$$\begin{aligned} {\mathcal {R}}_{Q(G^{\,\prime })}(f)-{\mathcal {R}}_{Q(G)}(f)= & {} \sum \limits _{xy\in E(G^{\,\prime })} (f(x)+f(y))^2-\sum \limits _{xy\in E(G)} (f(x)+f(y))^2\\= & {} f^2(v_k)-f^2(v_l)+2f(v_k)(f(a)-f(v_l))\\&+2f(b)(f(v_k)-f(a))\ge 0. \end{aligned}$$

By Lemma 2.1, we have \(\rho (G^{\,\prime })\ge {\mathcal {R}}_{A(G^{\,\prime })}(f)\ge {\mathcal {R}}_{A(G)}(f)=\rho (G)\) (resp., \(\mu (G^{\,\prime })\ge {\mathcal {R}}_{Q(G^{\,\prime })}(f)\ge {\mathcal {R}}_{Q(G)}(f)=\mu (G)\)). Noting that \(N_{G^{\,\prime }}(v_l)\subset N_G(v_l)\), by Lemma 2.4, we have \(\rho (G^{\,\prime })>\rho (G)\) (resp., \(\mu (G^{\,\prime })>\mu (G)\)).

If \(av_k\in E(G)\), we can show \(f(v_k)\ge f(a)\) similarly. Let \(G^{\,\prime }=G-wv_l-ab+wa+v_kb\). Then, \(d_{G^{\,\prime }}(v_k)=d_G(v_k)+1\), \(d_{G^{\,\prime }}(v_l)=d_G(v_l)-1\), and \(d_{G^{\,\prime }}(v)=d_G(v)\) for \(v\in V(G){\setminus } \{v_k, v_l\}\). It is not difficult to see that \(G^{\,\prime }-v_1\) is a unicyclic graph, \(G^{\,\prime }\in J_{\pi '}\),

$$\begin{aligned} {\mathcal {R}}_{A(G^{\,\prime })}(f)-{\mathcal {R}}_{A(G)}(f)= & {} 2\sum \limits _{xy\in E(G^{\,\prime })} f(x)f(y)-2\sum \limits _{xy\in E(G)} f(x)f(y)\\= & {} 2f(w)f(a)+2f(v_k)f(b)-2f(a)f(b)-2f(w)f(v_l)\\= & {} 2f(w)(f(a)-f(v_l))+2f(b)(f(v_k)-f(a))\ge 0, \end{aligned}$$

and

$$\begin{aligned} {\mathcal {R}}_{Q(G^{\,\prime })}(f)-{\mathcal {R}}_{Q(G)}(f)= & {} \sum \limits _{xy\in E(G^{\,\prime })} (f(x)+f(y))^2-\sum \limits _{xy\in E(G)} (f(x)+f(y))^2\\= & {} f^2(v_k)-f^2(v_l)+2f(w)(f(a)-f(v_l))\\&+\,2f(b)(f(v_k)-f(a))\ge 0. \end{aligned}$$

By Lemma 2.1, we have \(\rho (G^{\,\prime })\ge {\mathcal {R}}_{A(G^{\,\prime })}(f)\ge {\mathcal {R}}_{A(G)}(f)=\rho (G)\) (resp., \(\mu (G^{\,\prime })\ge {\mathcal {R}}_{Q(G^{\,\prime })}(f)\ge {\mathcal {R}}_{Q(G)}(f)=\mu (G)\)). Noting that \(N_{G^{\,\prime }}(v_l)\subset N_G(v_l)\), by Lemma 2.4, we have \(\rho (G^{\,\prime })>\rho (G)\) (resp., \(\mu (G^{\,\prime })>\mu (G)\)).

Subcase 2.2:

\(|V(P_{kl})|=3\). In this case, \(P_{kl}=v_kuv_l\) and \(C_4=v_kwv_luv_k\) is the unique cycle of \(G-v_1\). This implies that \(d_G(v_k)\ge 3\), \(d_G(w)\ge 3\), \(d_G(u)\ge 3\). By \(d_i=d^{\,\prime }_i=2\) for \(l+1\le i\le n\), we have \(w\prec v_l\) and \(u\prec v_l\). It follows that \(f(w)\ge f(v_l)\) and \(f(u)\ge f(v_l)\).

If \(f(v_k)\ge f(w)\), let \(G^{\,\prime }=G-wv_l-uv_l+v_kv_l+wu\). Then, \(d_{G^{\,\prime }}(v_k)=d_G(v_k)+1\), \(d_{G^{\,\prime }}(v_l)=d_G(v_l)-1\), \(d_{G^{\,\prime }}(v)=d_G(v)\) for \(v\in V(G){\setminus } \{v_k, v_l\}\). This implies that \(G^{\,\prime }-v_1\) is a unicyclic graph, \(G^{\,\prime }\in J_{\pi '}\),

$$\begin{aligned} {\mathcal {R}}_{A(G^{\,\prime })}(f)-{\mathcal {R}}_{A(G)}(f)= & {} 2\sum \limits _{xy\in E(G^{\,\prime })} f(x)f(y)-2\sum \limits _{xy\in E(G)} f(x)f(y)\\= & {} 2f(v_k)f(v_l)+2f(w)f(u)-2f(w)f(v_l)-2f(u)f(v_l)\\= & {} 2f(v_l)(f(v_k)-f(w))+2f(u)(f(w)-f(v_l))\ge 0, \end{aligned}$$

and

$$\begin{aligned} {\mathcal {R}}_{Q(G^{\,\prime })}(f)-{\mathcal {R}}_{Q(G)}(f)= & {} \sum \limits _{xy\in E(G^{\,\prime })} (f(x)+f(y))^2-\sum \limits _{xy\in E(G)} (f(x)+f(y))^2\\= & {} f^2(v_k)-f^2(v_l)+2f(v_l)(f(v_k)-f(w))\\&+\,2f(u)(f(w)-f(v_l))\ge 0. \end{aligned}$$

By Lemma 2.1, we have \(\rho (G^{\,\prime })\ge {\mathcal {R}}_{A(G^{\,\prime })}(f)\ge {\mathcal {R}}_{A(G)}(f)=\rho (G)\) (resp., \(\mu (G^{\,\prime })\ge {\mathcal {R}}_{Q(G^{\,\prime })}(f)\ge {\mathcal {R}}_{Q(G)}(f)=\mu (G)\). Noting that \(N_G(v_k)\subset N_{G^{\,\prime }}(v_k)\), by Lemma 2.4, we have \(\rho (G^{\,\prime })>\rho (G)\) (resp., \(\mu (G^{\,\prime })>\mu (G)\)).

If \(f(w)>f(v_k)\), let \(G^{\,\prime }=G-wv_l-uv_k+wu+v_kv_l\). It is easy to see that \(G^{\,\prime }-v_1\) is a unicyclic graph and \(d_{G^{\,\prime }}(v)=d_G(v)\) for \(v\in V(G)\). This implies that \(G^{\,\prime }\in J_{\pi }\). By Lemma 2.3, we have \(\rho (G^{\,\prime })>\rho (G)\) (resp., \(\mu (G^{\,\prime })>\mu (G)\)), a contradiction because G has the greatest spectral radius (resp., signless Laplacian spectral radius) in \(J_{\pi }\). \(\square \)

The proof of Theorem 1.3

Since \(\pi \lhd \pi ^{\,\prime }\), it follows from Lemma 2.7 and Definition 2.8 that there exists a series non-increasing graphic degree sequences \(\pi _1, \ldots , \pi _k\) such that \((\pi =)\)\(\pi _0\lhd ^* \pi _1\lhd ^* \cdots \lhd ^* \pi _k \lhd ^* \pi _{k+1}(=\pi ^{\,\prime })\). Let \(\pi _i=(d_1^{\,(i)}, d_2^{\,(i)}, \ldots , d_n^{\,(i)})\) for \(0\le i\le k\). By Lemma 2.6, we have \(d_n\ge d_n^{\,(1)}\ge \cdots \ge d_n^{\,(k)}\ge d_n^{\,\prime }\ge 2\).

For \(\pi \) and \(\pi _1\), Lemma 4.1 implies that there exists a single-cone unicyclic graph \(G_1\in J_{\pi _1}\) such that \(\rho (G)<\rho (G_1)\) (resp., \(\mu (G)<\mu (G_1)\)). It follows that \(\pi _1\) is a non-increasing degree sequence of a single-cone unicyclic graph. Let \(G_1^*\) be a single-cone unicyclic graph with the greatest spectral radius (resp., signless Laplacian spectral radius) in \(J_{\pi _1}\). Then, \(\rho (G)<\rho (G_1)\le \rho (G_1^*)\) (resp., \(\mu (G)<\mu (G_1)\le \mu (G_1^*)\)).

By a similar reasoning as the above, we can obtain that \(\pi _j\) is a non-increasing degree sequence of a single-cone unicyclic graph for each \(2\le j\le k\). Let \(G_j^*\) be a single-cone unicyclic graph with the greatest spectral radius (resp., signless Laplacian spectral radius) in \(J_{\pi _j}\) for \(2\le j\le k\). By Lemma 4.1, we have \(\rho (G)<\rho (G_1^*)<\cdots<\rho (G_k^*)<\rho (G^{\,\prime })\) (resp., \(\mu (G)<\mu (G_1^*)<\cdots<\mu (G_k^*)<\mu (G^{\,\prime })\)). \(\square \)