数学物理学报, 2025, 45(5): 1392-1404

一类算子束拟谱的精细刻画与最小拟谱

薛瑞瑶,, 侯国林,*

内蒙古大学数学科学学院 呼和浩特 010021

The Fine Pseudo-Spectra of a Class of Operator Pencils with Minimal Pseudo-Spectra

Xue Ruiyao,, Hou Guolin,*

School of Mathematical Science, Inner Mongolia University, Huhhot 010021

通讯作者: * 侯国林,E-mail:smshgl@imu.edu.cn

收稿日期: 2024-11-28   修回日期: 2025-01-9  

基金资助: 国家自然科学基金(12261064)
国家自然科学基金(11861048)
内蒙古自然科学基金(2025MS01026)

Received: 2024-11-28   Revised: 2025-01-9  

Fund supported: NSFC(12261064)
NSFC(11861048)
Natural Science Foundation of Inner Mongolia(2025MS01026)

作者简介 About authors

薛瑞瑶,E-mail:ruiyaoxue@mail.imu.edu.cn

摘要

$T, S$ 分别是 Hilbert 空间 $\mathcal{X}$ 中的稠定闭线性算子和有界线性算子, $\lambda S-T$ 表示一次算子束. 该文刻画了算子束的拟谱、拟点谱、拟连续谱和拟剩余谱的结构, 并得出算子束的广义拟谱等于其最小拟谱的充要条件. 设 $H$ 是 Hamilton 算子, $J$ 是辛单位算子, 之后得到了算子束 $\lambda J-H$ 的拟谱及其细分的对称性, 并论证了拟谱与其广义拟谱相等, 且等于其最小拟谱. 最后用例子进行了辅证.

关键词: 算子束; 拟谱; 分块算子矩阵; Hamilton 算子

Abstract

Let $T$ and $S$ be densely defined closed linear operators and bounded linear operators in the Hilbert space $\mathcal{X}$, respectively. Denote by $\lambda S - T$ the operator pencils. This paper investigates the structure of the fine pseudo-spectra of $\lambda S - T$, including pseudo-point spectra, pseudo-continuous spectra, and pseudo-residual spectra. Furthermore, a necessary and sufficient condition is derived for the equality of the generalized pseudo-spectra of operator pencils to the minimal pseudo-spectra. Let $H$ denote Hamiltonian operators and $J$ denote symplectic identity operators; the symmetries of the pseudo-spectra and their subdivisions for the operator pencils $\lambda J - H$ are obtained. Additionally, it is proven that the pseudo-spectra of $\lambda J - H$ coincide with both the generalized pseudo-spectra and the minimal pseudo-spectra. Finally, an illustrative example is provided to substantiate the theoretical conclusions.

Keywords: operator pencils; pseudo-spectra; operator matrices; Hamiltonian operators

PDF (580KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

本文引用格式

薛瑞瑶, 侯国林. 一类算子束拟谱的精细刻画与最小拟谱[J]. 数学物理学报, 2025, 45(5): 1392-1404

Xue Ruiyao, Hou Guolin. The Fine Pseudo-Spectra of a Class of Operator Pencils with Minimal Pseudo-Spectra[J]. Acta Mathematica Scientia, 2025, 45(5): 1392-1404

1 引言

自伴算子的谱在量子力学中发挥着重要作用, 量子力学中可观测量和本征态是以自伴算子的谱为基础进行研究的. 但对于非正规算子 $A$, 基于谱的预测可能会导致错误的结果. 例如, 利用 GMRES 迭代方法求解线性方程 $Ax=b$ 时, 若 $A$ 非正规, 基于谱的 GMRES 方法的收敛速度会因为一个潜在的无界因子而变得缓慢, 运用 $A$ 的拟谱, GMRES 方法的收敛性便得到很好解决 [1]. 因此, 研究拟谱有助于理解涉及非正规算子的行为. Kato, Trefethen 等学者[24]给出了拟谱详尽的描述. 之后多位学者研究了各种算子的拟谱, 例如卷积算子、Toeplitz 算子、微分算子和幂零算子, 参见文献 [58]. 文献 [9]探讨了 Banach 代数情形的 $(n,\varepsilon)$-拟谱, 并给出了其拓扑性质.

拟谱的定义与特征值问题 $Tu = \lambda u$ 直接相关, 然而在一些应用中 [10,11], 需要处理广义特征值问题 $Tu = \lambda Su,$ 这本质上是矩阵束的拟谱问题. 文献 [12]基于范数给出了矩阵束拟谱的定义, 文献 [13]讨论了矩阵束在扰动下拟谱的定义、性质和计算问题, 并用于研究偏微分方程平稳解的渐近稳定性. 到目前为止, 一些学者致力于将这一概念推广到算子束的拟谱, 有时候也以 $S$-拟谱的名义进行研究. Ammar, Bouchekoua 和 Jeribi 等学者通过 $S$-谱的扰动刻画了 $S$-拟谱[14,15], 随后说明这些扰动算子在 Banach 空间中具有严格小于 $\varepsilon$ 的范数 [16].文献 [17]研究了 Banach 代数上算子束的 $(n,\varepsilon)$-拟谱的性质, 证明了算子束的拟谱映射定理和 $(n,\varepsilon)$-拟谱映射定理. 文献 [18]给出了算子束广义拟谱的定义并论证了在较严格条件下, 算子束的拟谱等于其最小拟谱的充分条件. 但是具有广泛应用的 Hamilton 算子和对应的辛单位算子构成的算子束的拟谱未见研究. 无穷维 Hamilton 算子是一类特殊的非自伴线性算子, 由无穷维 Hamilton 系统导出. 文献 [1921]研究了无穷维 Hamilton 算子的可逆性, 文献 [22]研究了无穷维 Hamilton 算子特征函数系的完备性并应用于弹性力学求解, 文献 [23,24]研究了无穷维 Hamilton 算子的辛自伴性, 文献 [25,26]研究了无穷维 Hamilton 算子的数值域的谱包含关系和对称性, 文献 [2730]分别研究了无穷维 Hamilton 算子的谱和拟谱的性质.

本文首先给出了算子束的拟点谱、拟连续谱和拟剩余谱的定义, 并且得到了算子束拟谱及其细分的分布, 研究了对角分块算子矩阵束与矩阵束内部算子元的拟谱的关系. 受文献 [18]的启发, 得到了算子束广义拟谱等于其最小拟谱的充分必要条件. 对于由 Hamilton 算子和辛单位算子构成的算子束, 得到了其拟谱的分布, 给出了拟谱、广义拟谱和最小拟谱之间的关系, 最后用由调和方程导出的 Hamilton 算子和对应的辛单位算子构成的算子束说明了结论的有效性.

2 预备知识

为叙述简洁, 文中始终用 $\mathcal{X}$ 表示 Hilbert 空间, $I$ 表示 $\mathcal{X}$ 上的单位算子. $\mathcal{B(X)}$, $\mathcal{C(X)}$ 分别表示 $\mathcal{X}$ 上所有有界线性算子, 所有稠定闭算子的集合. 对任意的 $T \in \mathcal{C(X)}$, $\mathcal{D}(T), \mathcal{N}(T)$ 分别表示 $T$ 的定义域和零空间. $\rm{B}(0,\varepsilon)$ 表示以 0 为圆心, 以 $\varepsilon$ 为半径的复开圆盘. 若 A 表示复平面的子集, $\partial$(A) 表示 A 的边界, $\overline{\rm{A}}$ 表示 A 的闭包, dist($\lambda,\rm{A})$ 表示 $\lambda \in \mathbb{C}$ 到 A 的欧氏距离.

下面介绍一些相关定义.

${\bf定义2.1}$$T \in \mathcal{C(X)}$, $S \in \mathcal{B(X)}$, 算子束 $\lambda S-T$ 的预解集定义为

$\rho(T,S)=\{\lambda \in \mathbb{C}:\lambda S-T \mbox{ 有有界逆} \}.$

算子束的谱集定义为 $\sigma(T,S)=\mathbb{C} \backslash \rho(T,S);$$\sup\{|\lambda|:\lambda \in \sigma(T,S)\}$ 为算子束的谱半径, 记为 $r(T,S)$, 当 $S=I$ 时, 简记为 $r(T)$.

$\lambda \in \sigma(T,S)$ 时, 有下面三种可能

${\bf定义2.2}$ (1) 算子束的点谱为: $\sigma_{p}(T,S)=\{\lambda \in \mathbb{C}: \lambda S-T \mbox{不是单射} \}$;

(2) 算子束的连续谱为: $\sigma_{c}(T,S)=\{\lambda \in \mathbb{C}: \lambda S-T \mbox{是单射},$ $\overline{(\lambda S-T)(\mathcal{D}(T))}=\mathcal{X} \mbox{ 且 } (\lambda S-T)(\mathcal{D}(T)) \neq \mathcal{X} \}$;

(3) 算子束的剩余谱为: $\sigma_{r}(T,S)=\{\lambda \in \mathbb{C}: \lambda S-T \mbox{是单射}, \overline{(\lambda S-T)(\mathcal{D}(T))} \neq \mathcal{X} \}$.

根据算子值域的闭性和稠定性, 可以将算子束的点谱, 剩余谱作进一步细分

$\begin{align*} & \sigma_{p,1}(T,S)=\{\lambda \in \sigma_{p}(T,S): (\lambda S-T)(\mathcal{D}(T)) =\mathcal{X}\}; \notag \\ & \sigma_{p,2}(T,S)=\{\lambda \in \sigma_{p}(T,S): \overline{(\lambda S-T)(\mathcal{D}(T))}=\mathcal{X}, (\lambda S-T)(\mathcal{D}(T)) \mbox{ 不闭}\};\notag \\ & \sigma_{p,3}(T,S)=\{\lambda \in \sigma_{p}(T,S): \overline{(\lambda S-T)(\mathcal{D}(T))} \neq \mathcal{X}, (\lambda S-T)(\mathcal{D}(T)) \mbox{ 闭的} \};\notag \\ & \sigma_{p,4}(T,S)=\{\lambda \in \sigma_{p}(T,S): \overline{(\lambda S-T)(\mathcal{D}(T))} \neq \mathcal{X}, (\lambda S-T)(\mathcal{D}(T)) \mbox{ 不闭} \};\notag \\ & \sigma_{r,1}(T,S)=\{\lambda \in \sigma_{r}(T,S): (\lambda S-T)(\mathcal{D}(T)) \mbox{ 闭的} \};\notag \\ & \sigma_{r,2}(T,S)=\{\lambda \in \sigma_{r}(T,S): (\lambda S-T)(\mathcal{D}(T)) \mbox{ 不闭} \}. \end{align*}$

${\bf定义2.3}$[18]$T \in \mathcal{C(X)}$, $S \in \mathcal{B(X)}$$\varepsilon >0.$ 算子束 $\lambda S-T$ 的拟谱为

$\begin{align*} \Sigma_{\varepsilon}(T,S) &=\sigma(T,S) \cup \bigg\{\lambda \in \rho(T,S):\|(\lambda S-T)^{-1}\| > \frac{1}{\varepsilon}\bigg\} =\bigcup_{\substack {E\in \mathcal{B(X)} \\ \|E\|< \varepsilon}} \sigma(T+E,S). \end{align*}$

${\bf定义2.4}$[18] 算子束 $\lambda S-T $ 的广义拟谱为

$\begin{align*} \Lambda_{\varepsilon}(T,S) =\sigma(T,S) \cup \bigg\{\lambda \in \rho(T,S):\|(\lambda S-T)^{-1}S\| > \frac{1}{\varepsilon}\bigg\}. \end{align*}$

${\bf注 2.1}$$S=I$ 时, 有

$\Sigma_{\varepsilon}(T,S)=\Lambda_{\varepsilon}(T,S)=\sigma_{\varepsilon}(T) = \sigma(T) \cup \bigg\{\lambda \in \rho(T):\|(\lambda I-T)^{-1}\| > \frac{1}{\varepsilon}\bigg\}.$

$\|S\| < 1$ 时, 有 $\|(\lambda S-T)^{-1}S\| < \|(\lambda S-T)^{-1}\|$, 因此 $\Lambda_{\varepsilon}(T,S) \subset \Sigma_{\varepsilon}(T,S).$ 即使当 $\|S\|= 1$ 时, 包含关系还可能成立.考虑矩阵

$T=\begin{pmatrix} z & 0 \\ 0 & 0 \end{pmatrix}, \mbox{这里} z \in \mathbb{C}, S=\begin{pmatrix} 0 & 0 \\ 0 & 1 \end{pmatrix}.$

经过简单计算, $\sigma(T,S)=\{0\},$$\|(\lambda S-T)^{-1}\|=\max\bigg\{\dfrac{1}{|z|}, \dfrac{1}{|\lambda|}\bigg\},$$\|(\lambda S-T)^{-1}S\|=\bigg\{\dfrac{1}{|\lambda|}\bigg\}.$

$|z|<|\lambda|$ 时, 有 $\Lambda_{\varepsilon}(T,S)\subsetneq \Sigma_{\varepsilon}(T,S).$ 这说明算子束的广义拟谱可以比拟谱更精细.

${\bf定义2.5}$$T \in \mathcal{C(X)}$, $S \in \mathcal{B(X)}$$\varepsilon >0.$ 算子束的拟点谱 $\Sigma_{\varepsilon,p}(T,S),$ 拟连续谱 $\Sigma_{\varepsilon,c}(T,S),$ 拟剩余谱 $\Sigma_{\varepsilon,r}(T,S)$ 定义为

$\Sigma_{\varepsilon,\alpha}(T,S)=\{\lambda \in \mathbb{C}: \mbox{存在} E \in \mathcal{B(X)}, \|E\| < \varepsilon, \mbox{使得} \lambda \in \sigma_{\alpha}(T+E,S)\},$

这里 $\alpha \in \{p, c,r\}.$

类似地, 可以将算子束的拟点谱, 拟剩余谱作进一步细分.

$\Sigma_{\varepsilon,p,i}(T,S)=\{\lambda \in \mathbb{C}: \mbox{存在} E \in \mathcal{B(X)}, \|E\| < \varepsilon, \mbox{使得} \lambda \in \sigma_{p,i}(T+E,S)\},$

这里 $i=1,2,3,4.$

$\Sigma_{\varepsilon,r,j}(T,S)=\{\lambda \in \mathbb{C}: \mbox{存在} E \in \mathcal{B(X)}, \|E\| < \varepsilon, \mbox{使得} \lambda \in \sigma_{r,j}(T+E,S)\},$

这里 $j=1,2.$

${\bf定义2.6}$[30]$H: \mathcal{D}(H) \subset \mathcal{X} \oplus \mathcal{X} \rightarrow \mathcal{X} \oplus \mathcal{X}$ 是稠定闭线性算子. 若 $H$ 满足 $(JH)^*=JH,$ 其中$J=\begin{pmatrix} 0 & I \\ -I & 0 \end{pmatrix}$ 是辛单位算子, 则称 $H$ 是一个无穷维 Hamilton 算子.

${\bf定义2.7}$[31]$T \in \mathcal{C(X)}$, $\mathcal{M}$$\mathcal{X}$ 的一个闭子空间, $T \mathcal{M}$ 表示 $T$ 作用在 $\mathcal{D}(T) \cap \mathcal{M}$ 上. 若 $T \mathcal{M} \subseteq \mathcal{M}$, 则称 $\mathcal{M}$$T$ 的一个不变子空间.

${\bf命题2.1}$[4]$T \in \mathcal{C(X)}$ 是自伴算子, $\lambda \notin \sigma(T),$

$\|(\lambda I-T)^{-1}\|=r((\lambda I-T)^{-1})=\frac{1}{\rm{dist} \it{(\lambda,\sigma(T))}}.$

${\bf命题2.2}$[32]$T \in \mathcal{C(X)}$, 若 $\|T\| < 1,$$I-T$$\mathcal{X}$ 上有有界逆, 且 $(I-T)^{-1}=\sum\limits_{n=0}^{+\infty}T^n.$

3 算子束 $\lambda S-T$ 拟谱的精细刻画

本节给出 Hilbert 空间中算子束 $\lambda S-T$ 的拟谱, 广义拟谱的一些性质. 首先由算子束的拟点谱, 拟连续谱和拟剩余谱的定义得到算子束 $\lambda S-T$ 的拟谱细分和算子束 $\overline{\lambda} S^{*}-T^{*}$ 的拟谱细分的关系.

${\bf定理3.1}$$T \in \mathcal{C(X)},$$S \in \mathcal{B(X)} $$\varepsilon >0$, 则下面的叙述成立

(1) $\lambda \in \Sigma_{\varepsilon,p}(T,S) \Rightarrow \overline{\lambda} \in \Sigma_{\varepsilon,p}(T^*,S^*) \cup \Sigma_{\varepsilon,r}(T^*,S^*)$;

(2) $\lambda \in \Sigma_{\varepsilon,r}(T,S) \Rightarrow \overline{\lambda} \in \Sigma_{\varepsilon,p}(T^*,S^*)$;

(3) $\lambda \in \Sigma_{\varepsilon,c}(T,S) \Leftrightarrow \overline{\lambda} \in \Sigma_{\varepsilon,c}(T^*,S^*)$;

(4) $\lambda \in \Sigma_{\varepsilon}(T,S) \Leftrightarrow \overline{\lambda} \in \Sigma_{\varepsilon}(T^*,S^*)$.

${\bf证}$ (1) 若 $\lambda \in \Sigma_{\varepsilon,p}(T,S),$ 那么存在 $E \in \mathcal{B(X)}$, $\|E\| < \varepsilon$ 和非零元 $x \in \mathcal{D}(T)$ 使得 $(\lambda S-T-E)x=0.$ 对任意的 $y \in \mathcal{D}(T^*),$$((\lambda S-T-E)x,y)=(x,(\lambda S-T-E)^{*}y)$ 成立, 这表明 $\overline{(\overline{\lambda} S^*-T^*-E^*)(\mathcal{D}(T^*)})\neq \mathcal{X}$.$\overline{\lambda} S^*-T^*-E^*$ 不是单射, 则 $\overline{\lambda} \in \sigma_{p}(T^*+E^*,S^*)$; 再结合 $\|E^*\|=\|E\| < \varepsilon$, 得 $\overline{\lambda} \in \Sigma_{\varepsilon,p}(T^*,S^*).$$\overline{\lambda} S^*-T^*-E^*$ 是单射, 类似地有 $\overline{\lambda} \in \Sigma_{\varepsilon,r}(T^*,S^*)$. 因此, $\overline{\lambda} \in \Sigma_{\varepsilon,p}(T^*,S^*) \cup \Sigma_{\varepsilon,r}(T^*,S^*);$

(2) 若 $\lambda \in \Sigma_{\varepsilon,r}(T,S),$ 那么存在 $E \in \mathcal{B(X)}$, $\|E\| < \varepsilon,$ 使得 $\lambda \in \sigma_{r}(T+E,S).$ 则有 $\overline{(\lambda S-T-E)(\mathcal{D}(T)})$$=\mathcal{N}(\overline{\lambda}S^{*}-T^{*}-E^{*})^{\bot} \neq \mathcal{X},$$\mathcal{N}(\overline{\lambda}S^{*}-T^{*}-E^{*})\neq \{0\},$ 注意到 $\|E^{*}\|=\|E\|< \varepsilon,$ 那么 $\lambda \in \Sigma_{\varepsilon,p}(T^*,S^*);$

(3) 若 $\lambda \in \Sigma_{\varepsilon,c}(T,S),$ 那么存在 $E \in \mathcal{B(X)}$$\|E\| < \varepsilon,$ 使得 $\lambda \in \sigma_{c}(T+E,S).$ 则有 $\overline{(\lambda S-T-E)(\mathcal{D}(T)})$$=\mathcal{X},$ 但是 $(\lambda S-T-E)(\mathcal{D}(T))\neq \mathcal{X},$$\overline{(\lambda S-T-E)(\mathcal{D}(T)})=\mathcal{N}(\overline{\lambda}S^{*}-T^{*}-E^{*})^{\bot} = \mathcal{X},$$\mathcal{N}(\overline{\lambda}S^{*}-T^{*}-E^{*})=\{0\}.$ 所以 $\overline{\lambda} \in \rho(T^*+E^*,S^{*}) \cup \sigma_{r}(T^*+E^*,S^{*}) \cup \sigma_{c}(T^*+E^*,S^{*}).$ 我们断言, $\overline{\lambda} \in \sigma_{c}(T^*+E^*,S^*).$$\overline{\lambda} \in \rho(T^*+E^*,S^{*}),$$\lambda \in \rho(T+E,S),$ 这与 $\lambda \in \sigma_{c}(T+E,S)$ 矛盾. 若 $\overline{\lambda} \in \sigma_{r}(T^*+E^*,S^{*}),$$\lambda \in \sigma_{p}(T+E,S),$ 也与 $\lambda \in \sigma_{c}(T+E,S)$ 矛盾. 故 $\overline{\lambda} \in \sigma_{c}(T^*+E^*,S^*),$ 注意到 $\|E^{*}\|=\|E\| <\varepsilon,$$\overline{\lambda} \in \Sigma_{\varepsilon,c}(T^*,S^*).$ 充分性的证明类似;

(4) 若 $\lambda \in \Sigma_{\varepsilon}(T,S),$$\lambda \in \sigma(T,S)$ 或者满足 $\|(\lambda S-T)^{-1}\| > \frac{1}{\varepsilon}.$$\lambda \in \sigma(T,S),$$\overline{\lambda} \in \sigma(T^{*},S^{*});$$\|(\lambda S-T)^{-1}\| > \frac{1}{\varepsilon},$$\|(\overline{\lambda} S^{*}-T^{*})^{-1}\| > \frac{1}{\varepsilon}.$ 结合上述两种情况有 $\overline{\lambda} \in \Sigma_{\varepsilon}(T^*,S^*).$ 充分性的证明类似.

类似于定理 3.1的证明, 可得到定理 3.2.

${\bf定理3.2}$$T \in \mathcal{C(X)}, S \in \mathcal{B(X)}$$\varepsilon >0$, 则下面的叙述成立:

(1) $\lambda \in \Sigma_{\varepsilon,p,1}(T,S) \Leftrightarrow \overline{\lambda} \in \Sigma_{\varepsilon,r,1}(T^*,S^*)$;

(2) $\lambda \in \Sigma_{\varepsilon,p,2}(T,S) \Leftrightarrow \overline{\lambda} \in \Sigma_{\varepsilon,r,2}(T^*,S^*)$;

(3) $\lambda \in \Sigma_{\varepsilon,p,3}(T,S) \Leftrightarrow \overline{\lambda} \in \Sigma_{\varepsilon,p,3}(T^*,S^*)$;

(4) $\lambda \in \Sigma_{\varepsilon,p,4}(T,S) \Leftrightarrow \overline{\lambda} \in \Sigma_{\varepsilon,p,4}(T^*,S^*)$.

下面我们从分块算子矩阵角度出发, 讨论对角分块算子矩阵束在一般扰动下的拟谱.

${\bf定理3.3}$$T=\left( \begin{array}{cccc} T_1 & 0 \\ 0 & T_2 \\ \end{array} \right), S=\left( \begin{array}{cccc} S_1 &0 \\ 0 & S_2 \\ \end{array} \right)$, $T_{i}\in \mathcal{C(X)}, S_{i}\in\mathcal{B(X)}, i=1,2$$\varepsilon >0,$$\Sigma_{\varepsilon}(T,S) =\Sigma_{\varepsilon}(T_1,S_1) \cup \Sigma_{\varepsilon}(T_2,S_2). $

${\bf证}$$\lambda \in \Sigma_{\varepsilon}(T_1,S_1),$ 对于 $\varepsilon > 0,$ 存在 $E_1$$\|E_1\| < \varepsilon$, 使得 $\lambda \in \sigma(T_1+E_1, S_1).$$E=\left(\begin{array}{cc} E_1 & 0 \\ 0 & 0 \\ \end{array} \right)$$\|E\|=\|E_1\| < \varepsilon,$ 则有 $\lambda \in \sigma(T+E,S)$, 即 $\Sigma_{\varepsilon}(T_1,S_1) \subset \Sigma_{\varepsilon}(T,S).$ 同理可证, $ \Sigma_{\varepsilon}(T_2,S_2) \subset \Sigma_{\varepsilon}(T,S).$ 因此 $\Sigma_{\varepsilon}(T_1,S_1) \cup \Sigma_{\varepsilon}(T_2,S_2) \subset \Sigma_{\varepsilon}(T,S).$

相反地, 已知 $\sigma(T,S)=\sigma(T_1,S_1) \cup \sigma(T_2,S_2)$, 我们只要证 $\Sigma_{\varepsilon}(T,S) \backslash \sigma(T,S) \subset (\Sigma_{\varepsilon}(T_1,S_1) \cup \Sigma_{\varepsilon}(T_2,S_2)) \backslash (\sigma(T_1,S_1) \cup \sigma(T_2,S_2).$$\lambda \in \Sigma_{\varepsilon}(T,S) \backslash \sigma(T,S),$ 存在 $E=\left( \begin{array}{cc} E_{11} & E_{12} \\ E_{21} & E_{22} \\ \end{array} \right)$ 且满足 $ \|E\| < \varepsilon,$$\lambda \in \sigma(T+E,S)$. 已知

$\lambda S-T-E=\left( \begin{array}{cc} \lambda S_{1}-T_{1}-E_{11} & -E_{12} \\ -E_{21} & \lambda S_{2}-T_2-E_{22} \\ \end{array} \right)= \left( \begin{array}{cc} \lambda S_{1}-T_1 & 0 \\ 0 & \lambda S_2-T_2 \\ \end{array} \right)(I-M), $

其中

$M=\left( \begin{array}{cc} (\lambda S_{1}-T_{1})^{-1}E_{11} & (\lambda S_{1}-T_{1})^{-1}E_{12} \\ (\lambda S_{2}-T_2)^{-1}E_{21} & (\lambda S_{2}-T_2)^{-1}E_{22} \\ \end{array} \right).$

$I-M$ 不可逆, 根据命题 2.2, 有 $1 \leq \|M\|.$ 显然地, $\max\limits_{j \in \{1,2\}} \{\|(\lambda S_{1}-T_{1})^{-1}E_{1j}\|, \|(\lambda S_{2}-T_2)^{-1}E_{2j}\|\} \leq \|M\|$$ \max\limits_{j \in \{1,2\}} \{\|E_{1j}\|, \|E_{2j}\|\} \leq \|E\|<\varepsilon.$ 分两种情形讨论.

${\bf 情形 1}$$1 \leq \max\limits_{j \in \{1,2\}} \{\|(\lambda S_{1}-T_{1})^{-1}E_{1j}\|, \|(\lambda S_{2}-T_2)^{-1}E_{2j}\|\},$ 则对 $j \in \{1,2\},$$1 \leq \|(\lambda S_{1}-T_{1})^{-1}E_{1j}\| <\|(\lambda S_{1}-T_{1})^{-1}\|\varepsilon$ 或者 $1 \leq \|(\lambda S_{2}-T_{2})^{-1}E_{2j}\| <\|(\lambda S_{2}-T_{2})^{-1}\|\varepsilon,$ 即得 $\|(\lambda S_{1}-T_{1})^{-1}\|> \frac{1}{\varepsilon}$ 或者 $\|(\lambda S_{2}-T_{2})^{-1}\|> \frac{1}{\varepsilon}.$ 因此, $\lambda \in (\Sigma_{\varepsilon}(T_{1},S_1) \cup \Sigma_{\varepsilon}(T_2,S_2) \backslash (\sigma(T_1,S_1) \cup \sigma(T_1,S_1)).$

${\bf 情形 2}$$ \max\limits_{j \in \{1,2\}} \{\|(\lambda S_{1}-T_{1})^{-1}E_{1j}\|, \|(\lambda S_{2}-T_2)^{-1}E_{2j}\|\} < 1,$ 则对 $j \in \{1,2\},$$\|(\lambda S_{1}-T_{1})^{-1}E_{1j}\|< 1$$\|(\lambda S_{2}-T_2)^{-1}E_{2j}\|<1.$ 考虑 $\|(\lambda S_{1}-T_{1})^{-1}E_{11}\|< 1,$ 由命题 2.2,

$I-(\lambda S_{1}-T_1)^{-1}E_{11}=(\lambda S_{1}-T_{1})^{-1}(\lambda S_{1}-T_{1}-E_{11})$

是可逆的, 进而 $\lambda S_{1}-T_{1}-E_{11}$ 是可逆的. 类似地, 有 $\lambda S_{2}-T_{2}-E_{22}$ 是可逆的. 这表明 $\lambda \notin \Sigma_{\varepsilon}(T_1,S_1) \cap \Sigma_{\varepsilon}(T_2,S_2),$$ \lambda \in \Sigma_{\varepsilon}(T,S)$ 矛盾. 因此, $\lambda \in (\Sigma_{\varepsilon}(T_{1},S_1) \cup \Sigma_{\varepsilon}(T_2,S_2) \backslash (\sigma(T_1,S_1) \cup \sigma(T_1,S_1)).$

${\bf推论3.1}$$T \in \mathcal{C(X)}, S \in \mathcal{B(X)}$$\varepsilon>0.$$T$ 的值域是闭的, $T=\left(\begin{array}{cc} T_1 & 0 \\ 0 & 0 \\ \end{array} \right)$$:\mathcal{N}(T)^\bot $$\oplus \mathcal{N}(T) \longrightarrow \mathcal{R}(T) \oplus \mathcal{R}(T)^{\bot}.$$\mathcal{N}(T)^\bot$ 约化 $S$, $S$ 有算子矩阵表示 $S=\left(\begin{array}{cc} S_1 & 0 \\ 0 & S_2 \\ \end{array} \right)$$:\mathcal{N}(T)^\bot$$\oplus \mathcal{N}(T)\longrightarrow \mathcal{N}(T)^\bot \oplus \mathcal{N}(T).$ 那么有 $\Sigma_{\varepsilon}(T,S)=\Sigma_{\varepsilon}(T_1,S_1) \cup \Sigma_{\varepsilon}(0,S_2)$ 成立, 其中 $S_1=P_{N(T)^{\bot}}T|_{N(T)^{\bot}}, S_2=P_{N(T)}T|_{N(T)}, T_{1}=P_{R(T)}T|_{N(T)^{\bot}}.$

${\bf证}$ 由定理 3.3可得.

在一般扰动下, 拟谱细分较难刻画, 我们用对角扰动来刻画. $\Sigma^{D}_{\varepsilon,p}(T,S), \Sigma^{D}_{\varepsilon,c}(T,S), \Sigma^{D}_{\varepsilon,r}(T,S)$ 分别表示算子矩阵束在对角扰动下的拟点谱, 拟连续谱和拟剩余谱.

${\bf定理3.4}$ 设 $T=\left(\begin{array}{cccc} T_1 & 0 \\ 0 & T_2 \\ \end{array} \right), S=\left( \begin{array}{cccc} S_1 &0 \\ 0 & S_2 \\ \end{array} \right)$, $T_{i}\in \mathcal{C(X)}, S_{i}\in\mathcal{B(X)}, i=1,2$$\varepsilon >0,$

(1) $\Sigma^{D}_{\varepsilon,p}(T,S) =\Sigma_{\varepsilon,p}(T_1,S_1) \cup \Sigma_{\varepsilon,p}(T_2,S_2) $;

(2) $\Sigma^{D}_{\varepsilon,c}(T,S) \cap (\rho(T_1,S_1)\cap \rho(T_2,S_2)) = ((\Sigma_{\varepsilon,c}(T_1,S_1) \cap \Sigma_{\varepsilon,c}(T_2,S_2))\cap (\rho(T_1,S_1)\cap \rho(T_2,S_2)))\\ \cup (\Sigma_{\varepsilon,c}(T_1,S_1)\cap \rho(T_2,S_2)) \cup (\Sigma_{\varepsilon,c}(T_2,S_2)\cap \rho(T_1,S_1))$;

(3) $\Sigma^{D}_{\varepsilon,r}(T,S) \backslash \Sigma_{p}(T,S) = (\Sigma_{\varepsilon,r}(T_1,S_1) \cup \Sigma_{\varepsilon,r}(T_2,S_2)) \backslash (\Sigma_{p}(T_1,S_1) \cup \Sigma_{p}(T_2,S_2)) $.

${\bf证}$ (1) 先证 $\Sigma_{\varepsilon,p}(T_1,S_1) \cup \Sigma_{\varepsilon,p}(T_2,S_2) \subseteq \Sigma^{D}_{\varepsilon,p}(T,S).$ 假设 $\lambda \in \Sigma_{\varepsilon,p}(T_1,S_1), $ 则存在 $E_{1} \in \mathcal{B(X)}$$\|E_1\|< \varepsilon$ 和非零元 $ x \in \mathcal{D}(T)$ 使得 $(\lambda S_1-T_1-E_1)x=0.$$x= \begin{pmatrix} x_1 \\ 0 \end{pmatrix}$$E=\left( \begin{array}{cccc} E_1 & 0 \\ 0 & 0 \\ \end{array} \right).$ 故有 $\lambda S-T-E=\left( \begin{array}{cccc} \lambda S_{1}-T_{1}-E_1 & 0 \\ 0 & \lambda S_2-T_{2} \\ \end{array} \right) \begin{pmatrix} x_1 \\ 0 \end{pmatrix} =\bf{0}$ 成立. 因此 $\lambda \in \Sigma^{D}_{\varepsilon,p}(T,S).$ 同理可证 $\Sigma_{\varepsilon,p}(T_2,S_2) \subseteq \Sigma^{D}_{\varepsilon,p}(T,S).$

再证 $\Sigma^{D}_{\varepsilon,p}(T,S) \subseteq \Sigma_{\varepsilon,p}(T_1,S_1) \cup \Sigma_{\varepsilon,p}(T_2,S_2).$ 假设 $\lambda \in \Sigma^{D}_{\varepsilon,p}(T,S), $ 则存在 $E=\left( \begin{array}{cccc} E_{11} & 0 \\ 0 & E_{22} \\ \end{array} \right) \\\in \mathcal{B(X \oplus X)},$$ \|E\|<\varepsilon,$ 且存在不全为零的元素 $f,g,$ 使得

$(\lambda S-T-E) \begin{pmatrix} f \\ g \end{pmatrix} = \left( \begin{array}{cccc} \lambda B_{1}-A_{1}-E_{11}& 0 \\ 0 & \lambda B_{2}-A_{2}-E_{22} \\ \end{array} \right) \begin{pmatrix} f \\ g \end{pmatrix}=\bf{0}.$

如果 $f=0,g\neq0,$ 那么 $\lambda \subseteq \Sigma_{\varepsilon,p}(T_1,S_1);$ 如果 $g=0,f\neq0,$ 那么 $\lambda \subseteq \Sigma_{\varepsilon,p}(T_2,S_2).$ 因此, 有 $\lambda \in \Sigma_{\varepsilon,p}(T_1,S_1) \cup \Sigma_{\varepsilon,p}(T_2,S_2);$

(2) 如果 $\lambda \in \Sigma^{D}_{\varepsilon,c}(T,S) \cap (\rho(T_1,S_1)\cap \rho(T_2,S_2)), $ 那么存在算子 $E=\left( \begin{array}{cccc} E_1 & 0 \\ 0 & E_2 \\ \end{array} \right)\in \mathcal{B(X \oplus X)},$$ \|E\|<\varepsilon,$ 使得

$\mathcal{N}(\lambda S-T-E)=\{0\}, \overline{(\lambda S-T-E)(\mathcal{D}(T))} =\mathcal{X} \oplus \mathcal{X} \mbox{且} (\lambda S-T-E)(\mathcal{D}(T)) \neq \mathcal{X} \oplus \mathcal{X}.$

从而存在相应的 $E_i, i=1,2$ 使得

$\mathcal{N}(\lambda S_i-T_i-E_i)=\{0\}, \overline{(\lambda S_i-T_i-E_i)(\mathcal{D}(T))} =\mathcal{X},(\lambda S_i-T_i-E_i)(\mathcal{D}(T)) \neq \mathcal{X},$

其中 $(\lambda S_i-T_i-E_i)(\mathcal{D}(T)) \neq \mathcal{X}$ 分三种情况. 若 $i=1,2$ 都满足, 则

$\lambda \in ((\Sigma_{\varepsilon,c}(T_1,S_1) \cap \Sigma_{\varepsilon,c}(T_2,S_2))\cap (\rho(T_1,S_1)\cap \rho(T_2,S_2)));$

若只满足 $i=1$ 或只满足 $i=2$ 时, 则

$\lambda \in (\Sigma_{\varepsilon,c}(T_1,S_1)\cap \rho(T_2,S_2)) \cup (\Sigma_{\varepsilon,c}(T_2,S_2)\cap \rho(T_1,S_1)).$

因此, 有 $\lambda \in ((\Sigma_{\varepsilon,c}(T_1,S_1) \cap \Sigma_{\varepsilon,c}(T_2,S_2))\cap (\rho(T_1,S_1)\cap \rho(T_2,S_2))) \cup (\Sigma_{\varepsilon,c}(T_1,S_1)\cap \rho(T_2,S_2)) \\ \cup (\Sigma_{\varepsilon,c}(T_2,S_2)\cap \rho(T_1,S_1)).$

相反地, 若 $\lambda \in (\Sigma_{\varepsilon,c}(T_1,S_1) \cap \Sigma_{\varepsilon,c}(T_2,S_2))\cap (\rho(T_1,S_1)\cap \rho(T_2,S_2)),$ 那么存在 $E_j,j=1,2,$$ \|E_j\| <\varepsilon, $ 使得

$\mathcal{N}(\lambda S_j-T_j-E_j)=\{0\}, \overline{(\lambda S_j-T_j-E_j)(\mathcal{D}(T))} =\mathcal{X}, (\lambda S_j-T_j-E_j)(\mathcal{D}(T)) \neq \mathcal{X}.$

$F=\left(\begin{array}{cccc} F_1 & 0 \\ 0 & F_2 \\ \end{array} \right)\in \mathcal{B(X \oplus X)}, \|F\| <\varepsilon,$ 则有

$\mathcal{N}(\lambda S-T-F)=\{0\}, \overline{(\lambda S-T-F)(\mathcal{D}(T))} =\mathcal{X} \oplus \mathcal{X} \mbox{且} (\lambda S-T-F)(\mathcal{D}(T)) \neq \mathcal{X} \oplus \mathcal{X},$

$\lambda \in \Sigma^{D}_{\varepsilon,c}(T,S) \cap (\rho(T_1,S_1)\cap \rho(T_2,S_2). $

$\lambda \in \Sigma_{\varepsilon,c}(T_1,S_1) \cap \rho(T_2,S_2),$ 则存在 $F_3 \in \mathcal{B(X)}, \|F_3\|<\varepsilon,$ 使得 $\mathcal{N}(\lambda S_1-T_1-F_3)=\{0\}, \overline{(\lambda S_1-T_1-F_3)(\mathcal{D}(T_1))} =\mathcal{X} \mbox{且} (\lambda S_1-T_1-F_3)(\mathcal{D}(T_1)) \neq \mathcal{X},(\lambda S_2-T_2)(\mathcal{D}(T_2)) =\mathcal{X}.$

$\hat{F}=\left( \begin{array}{cccc} F_3 & 0 \\ 0 & 0 \\ \end{array} \right) \in \mathcal{B(X \oplus X)}, \|\hat{F}\|=\|F_3\|<\varepsilon,$ 这表明

$\mathcal{N}(\lambda S-T-\hat{F})=\{0\}, \overline{(\lambda S-T-\hat{F})(\mathcal{D}(T))} =\mathcal{X} \oplus \mathcal{X} \mbox{且} (\lambda S-T-\hat{F})(\mathcal{D}(T)) \neq \mathcal{X} \oplus \mathcal{X},$

因此 $\lambda \in \Sigma^{D}_{\varepsilon,c}(T,S) \cap (\rho(T_1,S_1)\cap \rho(T_2,S_2).$

同理有 $ \Sigma_{\varepsilon,c}(T_2,S_2) \cap \rho(T_1,S_1) \subseteq \Sigma^{D}_{\varepsilon,c}(T,S) \cap (\rho(T_1,S_1)\cap \rho(T_2,S_2).$

综上, $\Sigma^{D}_{\varepsilon,c}(T,S) \cap (\rho(T_1,S_1)\cap \rho(T_2,S_2)) = ((\Sigma_{\varepsilon,c}(T_1,S_1) \cap \Sigma_{\varepsilon,c}(T_2,S_2))\cap (\rho(T_1,S_1)\cap \rho(T_2,S_2))) \cup (\Sigma_{\varepsilon,c}(T_1,S_1)\cap \rho(T_2,S_2)) \cup (\Sigma_{\varepsilon,c}(T_2,S_2)\cap \rho(T_1,S_1));$

(3) 由类似的证法可得.

类似于算子束拟谱的等价定义, 我们得出算子束广义拟谱的等价定义, 并讨论其与$ \sigma(T,S)+B(0,\varepsilon)$ 之间的关系.

${\bf定理3.5}$[14]$\varepsilon >0$, $T \in \mathcal{C(X)}$, $S \in \mathcal{B(X)}$ 且满足 $\|S\|\leq 1$, 则

$\begin{equation*} \Lambda_{\varepsilon}(T,S) =\bigcup_{\substack {E\in \mathcal{B(X)} \\ \|E\|< \varepsilon}} \sigma(T+SE,S). \end{equation*}$

${\bf证}$ 先证明 "$\supset$". 设 $\lambda \in \mathbb{C} \backslash \Lambda_{\varepsilon}(T,S),$$\lambda \in \rho(T,S)$$\|(\lambda S-T)^{-1}S\| \leq \frac{1}{\varepsilon}.$

因为

$\lambda S-T-SE=(\lambda S-T)(I-(\lambda S-T)^{-1}SE), \|E\|<\varepsilon.$

进而

$\|(\lambda S-T)^{-1}SE\| \leq \|(\lambda S-T)^{-1}S\|\cdot \|E\|< \frac{1}{\varepsilon} \cdot \varepsilon=1.$

结合命题2.2, 知 $I-(\lambda S-T)^{-1}SE$ 可逆, $\lambda \in \rho(T+SE,S),$ 从而 $\lambda \notin \sigma(T+SE,S), \|E\|< \varepsilon$.

再证明 "$\subset$". 若 $\lambda \in \Lambda_{\varepsilon}(T,S)$, 假设 $\lambda \in \rho(T+SE,S),$ 对于所有的 $E \in \mathcal{B(X)}, \|E\|<\varepsilon.$$E=0,$$\lambda \in \rho(T,S),$ 可推得 $\overline{\lambda} \in \rho(T^*,S^*).$ 另取算子 $E,$$SE = \mu(\overline{\lambda}S^*-T^*)^{-1},$ 这里 $\mu \in \mathbb{C}$ 满足

$\begin{equation} \label{eqn:3.1} 0< |\mu|<\frac{\varepsilon}{\|(\overline{\lambda}S^*-T^*)^{-1}\|}. \end{equation}$

$\mu$ 满足 (3.1)式, 有

$\lambda S-T-SE=\lambda S-T-\mu(\overline{\lambda}S^*-T^*)^{-1}=\mu(\lambda S-T)(\frac{1}{\mu}-(\lambda S-T)^{-1}(\overline{\lambda}S^*-T^*)^{-1}),$ 这表明 $\frac{1}{\mu} \in \rho((\lambda S-T)^{-1}(\overline{\lambda}S^*-T^*)^{-1}).$ 那么 $\frac{1}{\mu} \in \sigma((\lambda S-T)^{-1}(\overline{\lambda}S^*-T^*)^{-1})$ 时, $|\mu|\geq \frac{\varepsilon}{\|(\overline{\lambda}S^*-T^*)^{-1}\|}.$ 又因为 $(\lambda S-T)^{-1}(\overline{\lambda}S^*-T^*)^{-1}$ 是自伴算子,$r((\lambda S-T)^{-1}(\overline{\lambda}S^*-T^*)^{-1})=\|(\lambda S-T)^{-1}(\overline{\lambda}S^*-T^*)^{-1}\|=\|(\lambda S-T)^{-1}\|^2 \leq \frac{\|(\overline{\lambda}S^*-T^*)^{-1}\|}{\varepsilon},$$\|(\lambda S-T)^{-1}\| \leq \frac{1}{\varepsilon},$ 推得矛盾.

${\bf定理3.6}$[15]$\varepsilon >0$, $T \in \mathcal{C(X)}$, $S \in \mathcal{B(X)}$ 且满足 $\|S\|\leq 1$, 则

(1) $\sigma(T,S)+\rm{B}(0,\varepsilon) \subset \Lambda_{\varepsilon}\it{(T,S)}$;

(2) $\Lambda_{\varepsilon}(T,S) + \rm{B} (0,\delta) \subset \Lambda_{\varepsilon+\delta}\it{(T,S)}$.

${\bf证}$ (1) 设 $\lambda \in \sigma(T,S)+\rm{B}(0,\varepsilon),$ 则存在 $\lambda_{1} \in \sigma(T,S), \lambda_{2} \in \rm{B}(0,\varepsilon),$ 使得 $\lambda=\lambda_{1}+\lambda_{2} \in \sigma(T+\lambda_{2}S,S).$$\|E\|=\|\lambda_{2}I\| < \varepsilon,$$SE \in \mathcal{B(X)} $$\|SE\| < \varepsilon.$ 因此由定理3.5, $\sigma(T,S)+\rm{B}(0,\varepsilon) \subset \Lambda_{\varepsilon}\it{(T,S)}.$

(2) 设 $\lambda \in \Lambda_{\varepsilon}(T,S) + \rm{B}(0,\delta)$, 则存在 $\lambda_{1} \in \Lambda_{\varepsilon}(T,S), \lambda_{2} \in \rm{B}(0,\delta).$$\lambda_{1} \in \Lambda_{\varepsilon}(T,S),$ 存在 $E \in \mathcal{B(X)}, \|E\|< \varepsilon,$$ \lambda_{1} \in \sigma(T+SE,S)$, 这样 $\lambda=\lambda_{1}+\lambda_{2} \in \sigma(T+SE+\lambda_{2}S,S)$, 又 $\|SE+\lambda_{2}S\| < \varepsilon + \delta,$$\lambda \in \Lambda_{\varepsilon+\delta}(T,S).$

我们称 $\sigma(T,S)+\rm{B}(0,\varepsilon)$ 为算子束的最小拟谱, 那么自然思考下述问题.

${\bf问题 1}$ 算子 $T$$S$ 满足什么条件时, 有 $\Lambda_{\varepsilon}(T,S)=\sigma(T,S)+\rm{B}(0,\varepsilon)$ 成立?

我们得到 $\Lambda_{\varepsilon}(T,S)=\sigma(T,S)+\rm{B}(0,\varepsilon)$ 的一个充分必要条件.

${\bf定理3.7}$$\varepsilon >0,$$T \in \mathcal{C(X)}$, $S \in \mathcal{B(X)}$ 且满足 $\|S\|\leq 1$. $T$$S$ 满足条件

$\begin{equation} \label{eqn:3.2} (G1):\|(\lambda S-T)^{-1}S\|=\frac{1}{\rm{dist} \it{(\lambda,\sigma(T,S))}}, \mbox{对于所有} \lambda \notin \sigma(T,S), \end{equation}$

当且仅当 $\Lambda_{\varepsilon}(T,S)=\sigma(T,S)+\rm{B}(0,\varepsilon).$

${\bf证}$ 一方面, 由定理3.6知$ \sigma(T,S)+\rm{B}(0,\varepsilon) \subset \Lambda_{\varepsilon}(T,S).$ 另一方面, 由条件 ($G1$) 知 dist$(\lambda,\sigma(T,S)) < \varepsilon,$ 这表明 $\Lambda_{\varepsilon}(T,S) \subset \sigma(T,S)+\rm{B}(0,\varepsilon).$

相反地, $\Lambda_{\varepsilon}(T,S)=\sigma(T,S)+\rm{B}(0,\varepsilon).$$\lambda \notin \sigma(T,S),$$\varepsilon =$ dist$(\lambda,\sigma(T,S)),$ 则有 $\lambda \in \partial(\sigma(T,S)+\rm{B}(0,\varepsilon)) \subset \partial(\Lambda_{\varepsilon}(T,S)),$$\|(\lambda S-T)^{-1}S\| = \frac{1}{\varepsilon}= \frac{1}{\rm{dist} \it{(\lambda,\sigma(T,S))}}.$

4 算子束 $\lambda J-H$ 拟谱的精细刻画

本节进一步剖析 Hilbert 空间中算子束 $\lambda J-H$ 的拟谱和广义拟谱的性质, 这里的 $H$$J$ 如定义 2.6 所示.

${\bf定理4.1}$$H$ 是一个无穷维 Hamilton 算子, 则下面的叙述成立:

(1) $\lambda \in \Sigma_{\varepsilon,p}(H,S) \Leftrightarrow \lambda \in \Sigma_{\varepsilon,p}(H^*,JSJ)$;

(2) $\lambda \in \Sigma_{\varepsilon,r}(H,S) \Leftrightarrow \lambda \in \Sigma_{\varepsilon,r}(H^*,JSJ)$;

(3) $\lambda \in \Sigma_{\varepsilon,c}(H,S) \Leftrightarrow \lambda \in \Sigma_{\varepsilon,c}(H^*,JSJ)$;

(4) $\lambda \in \Sigma_{\varepsilon}(H,S) \Leftrightarrow \lambda \in \Sigma_{\varepsilon}(H^*,JSJ)$.

${\bf证}$ (1) 若 $\lambda \in \Sigma_{\varepsilon,p}(H,S),$ 那么存在 $E \in \mathcal{B(X)}$$\|E\| < \varepsilon,$ 存在非零元 $x \in \mathcal{D}(H),$ 使得 $(\lambda S-H-E)x=0.$ 因为 $H=JH^{*}J,$

$(\lambda S-JH^{*}J-E)x=J(\lambda JSJ-H^{*}-JEJ)Jx=0$

成立, 其中 $Jx \neq 0,$$\lambda \in \Sigma_{p}(H^*+JEJ,JSJ).$$J$ 是酉算子, $\|JEJ\|=\|E\| < \varepsilon,$ 这表明 $\lambda \in \Sigma_{\varepsilon,p}(H^*,JSJ).$

(2) 若 $\lambda \in \Sigma_{\varepsilon,r}(H,S),$ 那么存在 $E \in \mathcal{B(X)}$$\|E\| < \varepsilon, \lambda \in \Sigma_{r}(H+E,S).$ 则有 $\overline{(\lambda S\!-\!H\!-\!E)}$$\overline{\mathcal{D}(H)}$$\neq \mathcal{X} \oplus \mathcal{X}.$$y \in \mathcal{X} \oplus \mathcal{X} \backslash \overline{(\lambda S-H-E)\mathcal{D}(H)},$ 对任意的 $x \in \mathcal{D}(H),$$((\lambda S-H-E)x,y)=0$ 成立. 由 $H=JH^{*}J$

$\begin{align*} (J(\lambda JSJ-H^{*}-JEJ)Jx,y) &=((\lambda JSJ-H^{*}-JEJ)Jx,J^{*}y) \\ &=((\lambda JSJ-H^{*}-JEJ)Jx,-Jy)=0. \end{align*}$

因此 $((\lambda JSJ-H^{*}-JEJ)Jx,Jy)=0$, 其中 $Jy\neq 0,$ 这表明 $\overline{(\lambda JSJ-H^{*}-JEJ)\mathcal{D}(H)} \neq \mathcal{X} \oplus \mathcal{X}.$

下面证 $(\lambda JSJ-H^{*}-JEJ)$ 是单射. $J$ 是可逆的, 若 $(\lambda JSJ-H^{*}-JEJ)z=0$,

$ (\lambda JSJ-H^{*}-JEJ)z=0 \Leftrightarrow J(\lambda S-H-E)Jz=0 \Leftrightarrow (\lambda S-H-E)Jz=0. $

由于 $\lambda S-H-E$ 是单射, 那么 $Jz=0,$ 从而 $z=0,$ 这表明 $\lambda JSJ-H^{*}-JEJ$ 是单射. 综上, $\lambda \in \Sigma_{r}(H^*+JEJ,JSJ).$$J$ 是酉算子, $\|JEJ\|=\|E\| < \varepsilon,$ 这表明 $\lambda \in \Sigma_{\varepsilon,r}(H^*,JSJ).$

充分性的证明类似.

(3) 由(1) (2) 可推得, (4) 由 (1) (2) (3) 可推得.

${\bf定理4.2}$$JS^{*}J=\beta S, \beta \in \{-1,1\}$, 则

(1) $\lambda \in \Sigma_{\varepsilon,pr}(H,S)=\Sigma_{\varepsilon,p}(H,S) \cup \Sigma_{\varepsilon,r}(H,S) \Leftrightarrow \beta \overline{\lambda} \in \Sigma_{\varepsilon,pr}(H,S)$;

(2) $\lambda \in \Sigma_{\varepsilon,c}(H,S) \Leftrightarrow \beta \overline{\lambda} \in \Sigma_{\varepsilon,c}(H,S)$;

(3) $\lambda \in \Sigma_{\varepsilon}(H,S) \Leftrightarrow \beta \overline{\lambda} \in \Sigma_{\varepsilon}(H,S)$;

(4) $\lambda \in \rho_{\varepsilon}(H,S) \Leftrightarrow \beta \overline{\lambda} \in \rho_{\varepsilon}(H,S)$.

${\bf证}$ (1) 设 $\lambda \in \Sigma_{\varepsilon,pr}(H,S)$, 由定理 3.1, 有 $\overline{\lambda} \in \Sigma_{\varepsilon,pr}(H^*,S^*).$ 再由定理 4.1, 有 $\overline{\lambda} \in$$ \Sigma_{\varepsilon,pr}(H,JS^*J).$$JS^{*}J=\beta S$, 那么 $\beta \overline{\lambda} \in \Sigma_{\varepsilon,pr}(H,S).$

其他情况的证明类似, 此处从略.

${\bf推论 4.1}$ (1) 若 $JS^{*}J=S,$$S$ 是一个 Hamilton 算子 (包括 $S=J$ 的情况), 则 $\Sigma_{\varepsilon,pr}(H,S),$$\Sigma_{\varepsilon,c}(H,S),$$\Sigma_{\varepsilon}(H,S)$$\rho_{\varepsilon}(H,S)$ 分别关于实轴对称;

(2) 若 $JS^{*}J=-S,$$S$ 是一个反 Hamilton 算子 (包括 $S=I$ 的情况), 则 $\Sigma_{\varepsilon,pr}(H,S),$$\Sigma_{\varepsilon,c}(H,S),$$ \Sigma_{\varepsilon}(H,S)$$\rho_{\varepsilon}(H,S)$ 分别关于虚轴对称.

${\bf定理4.3}$$H=\begin{pmatrix} A & 0 \\ 0 & -A^{*} \end{pmatrix} : \mathcal{D}(A)\times \mathcal{D}(A^*) \subset \mathcal{X} \oplus \mathcal{X} \rightarrow \mathcal{X} \oplus \mathcal{X}$ 是一个无穷维 Hamilton 算子, $J$ 是辛单位算子, $A$$\mathcal{X}$ 中的稠定闭线性算子, $0 \in \rho(A),$

$\Sigma_{\varepsilon}(H,J) \subseteq \{\lambda \in \mathbb{C}: \lambda ^2 \in \Sigma_{c_{\lambda}^2 \varepsilon} (-A^*,-A^{-1}) \} \cup \bigg\{\lambda \in \mathbb{C}: \|A^{-1}\|>\frac{1}{c_{\lambda}^2 \varepsilon}\bigg\},$

其中 $c_{\lambda}=1+\|\lambda A^{-1}\|$.

${\bf证}$ 由于 $0 \in \rho(A),$

$\lambda J-H=\left( \begin{array}{cc} I & 0 \\ \lambda A^{-1} & I \\ \end{array} \right) \left( \begin{array}{cc} -A & 0 \\ 0 & A^*-\lambda^2 A^{-1} \\ \end{array} \right) \left( \begin{array}{cc} I & -\lambda A^{-1} \\ 0 & I \\ \end{array} \right) $

得到 $\sigma (H,J)=\{\lambda \in \mathbb{C}: \lambda ^2 \in \sigma(-A^*,-A^{-1}) \}.$ 由于

$(\lambda J-H)^{-1}=\left( \begin{array}{cc} I & \lambda A^{-1} \\ 0 & I \\ \end{array} \right) \left( \begin{array}{cc} -A^{-1} & 0 \\ 0 & (A^*-\lambda^2 A^{-1})^{-1} \\ \end{array} \right) \left( \begin{array}{cc} I & 0 \\ -\lambda A^{-1} & I \\ \end{array} \right), $

所以

$\left \| \left( \begin{array}{cc} I & \lambda A^{-1} \\ 0 & I \\ \end{array} \right) \right \| \leq \left \|\left( \begin{array}{cc} I & 0 \\ 0 & I \\ \end{array} \right) \right \|+ \left \| \left( \begin{array}{cc} 0 & \lambda A^{-1} \\ 0 & 0 \\ \end{array} \right) \right \| \leq c_{\lambda}. $

同理

$ \left \| \left( \begin{array}{cc} I & 0 \\ -\lambda A^{-1} & I \\ \end{array} \right) \right \| \leq \left \|\left( \begin{array}{cc} I & 0 \\ 0 & I \\ \end{array} \right) \right \|+\left \|\left( \begin{array}{cc} 0 & 0 \\ -\lambda A^{-1} & 0 \\ \end{array} \right) \right \| \leq c_{\lambda}. $

那么

$ \|(\lambda J-H)^{-1}\| \leq c_{\lambda}^2 \max\{ \|A^{-1}\|, \|(A^*-\lambda^2 A^{-1})^{-1}\| \}.$

$\varepsilon >0,$

$\Sigma_\varepsilon (H,J) \subseteq \{\lambda \in \mathbb{C}: \lambda ^2 \in \Sigma_{c_{\lambda}^2 \varepsilon} (-A^*,-A^{-1}) \} \cup \bigg\{\lambda \in \mathbb{C}: \|A^{-1}\|>\frac{1}{c_{\lambda}^2 \varepsilon}\bigg\}.$

下述定理刻画了 $\lambda J-H$ 的拟谱、广义拟谱和最小拟谱三者之间的关系.

${\bf定理4.4}$$H$ 是一个无穷维 Hamilton 算子, $J$ 是辛单位算子, 则

$\Sigma_{\varepsilon}(H,J)=\Lambda_{\varepsilon}(H,J)=\sigma(H,J)+\rm{B}(0,\varepsilon).$

${\bf证}$ 先证明 $\Sigma_{\varepsilon}(H,J)=\Lambda_{\varepsilon}(H,J).$$\|(\lambda J-H)^{-1}\|=\|(\lambda-J^{-1}H)^{-1}J^{-1}\| \leq \|(\lambda -J^{-1}H)^{-1}\|=\|(\lambda J-H)^{-1}J\|$$\|(\lambda J-H)^{-1}J\|\leq \|(\lambda J-H)^{-1}\|,$$\|(\lambda J-H)^{-1}J\|=\|(\lambda J-H)^{-1}\|.$ 因此 $\Sigma_{\varepsilon}(H,J)=\Lambda_{\varepsilon}(H,J).$

再证明 $\Lambda_{\varepsilon}(H,J)=\sigma(H,J)+\rm{B}(0,\varepsilon).$ 由定理 3.6知, $ \sigma(H,J)+\rm{B}(0,\varepsilon) \subset \Lambda_{\varepsilon}\it{(H,J)}.$ 相反地, 由 $\lambda J-H=J(\lambda-J^{-1}H),$ 表明 $\sigma(H,J)=\sigma(J^{-1}H).$$\lambda \in \Lambda_{\varepsilon}(H,J) \backslash \sigma(H,J),$

$\|(\lambda J-H)^{-1}J\| =\|(\lambda-J^{-1}H)^{-1}\| > \frac{1}{\varepsilon}.$

由于

$ (J^{-1}H)^{*}=(J^{*}H)^{*}=H^{*}J=JHJJ=-JH=J^{-1}H,$

$J^{-1}H$ 是自伴算子, 由命题 2.1,

$\|(\lambda-J^{-1}H)^{-1}\| =r((\lambda-J^{-1}H)^{-1})=\frac{1}{\rm{dist} \it{(\lambda,\sigma(J^{-1}H))}}>\frac{1}{\varepsilon}.$

因此 $\lambda \in \sigma(H,J)+\rm{B}(0,\varepsilon).$

${\bf注 4.1}$ 算子束 $\lambda J-H$$\lambda \in \mathbb{R}$ 时满足 $G_1$ 条件.

${\bf推论4.2}$$H=\begin{pmatrix} 0 & B \\ C & 0 \end{pmatrix} : \mathcal{D}(B) \times \mathcal{D}(C) \subset \mathcal{X} \oplus \mathcal{X} \rightarrow \mathcal{X} \oplus \mathcal{X}$ 是一个无穷维 Hamilton 算子, $B, C$ 均为 $\mathcal{X}$ 中的自伴算子, $J$ 是辛单位算子, 则 $\Sigma_{\varepsilon}(H,J)=\sigma_{\varepsilon}(B) \cup \sigma_{\varepsilon}(-C).$

${\bf证}$ 对于 $\varepsilon > 0,$$\lambda \in \sigma_{\varepsilon}(B),$ 则存在 $E_1$$\|E_1\| < \varepsilon$, 使得 $\lambda \in \sigma(B+E_1).$$E=\left( \begin{array}{cc} 0 & E_1 \\ 0 & 0 \\ \end{array} \right)$$\|E\|=\|E_1\| < \varepsilon,$ 则有 $\lambda \in \sigma(H+E,J)$, 即 $\sigma_{\varepsilon}(B) \subset \Sigma_{\varepsilon}(H,J).$ 同理可证, $\sigma_{\varepsilon}(-C) \subset \Sigma_{\varepsilon}(H,J).$ 因此 $\sigma_{\varepsilon}(B) \cup \sigma_{\varepsilon}(-C) \subset \Sigma_{\varepsilon}(H,J).$

相反地, 已知 $\sigma(H,J)=\sigma(B) \cup \sigma(-C)$, 我们只要证 $\Sigma_{\varepsilon}(H,J) \backslash \sigma(H,J) \subset (\sigma_{\varepsilon}(B) \cup \sigma_{\varepsilon}(-C)) \backslash \\ (\sigma(B) \cup \sigma(-C)).$$\lambda \in \Sigma_{\varepsilon}(H,J) \backslash \sigma(H,J),$ 存在 $D=\left( \begin{array}{cc} D_{11} & D_{12} \\ D_{21} & D_{22} \\ \end{array} \right)$ 且满足 $\|D\|< \varepsilon,$$\lambda \in \sigma(H+D,J).$ 已知

$\lambda J-H-D=\left( \begin{array}{cc} -D_{11} & \lambda-B-D_{12} \\ -\lambda -C-D_{21} & -D_{22} \\ \end{array} \right)= \left( \begin{array}{cc} 0 & \lambda-B \\ -\lambda-C & 0 \\ \end{array} \right)(I-N), $

其中

$N=\left( \begin{array}{cc} (-\lambda-C)^{-1}D_{21} & (-\lambda-C)^{-1}D_{22} \\ (\lambda -B)^{-1}D_{11} & (\lambda -B)^{-1}D_{12} \\ \end{array} \right).$

$I-N$ 不可逆, 根据命题 2.2, 有 $1 \leq \|N\|.$ 显然地, $\max\limits_{j \in \{1,2\}} \{\|(\lambda -B)^{-1}D_{1j}\|, \|(-\lambda -C)^{-1}D_{2j}\|\} \leq \|N\|$$\max\limits_{j \in \{1,2\}} \{\|D_{1j}\|, \|D_{2j}\|\} \leq \|D\|<\varepsilon.$ 分两种情形讨论.

${\bf 情形 1}$$1 \leq \max\limits_{j \in \{1,2\}} \{\|(\lambda -B)^{-1}D_{1j}\|, \|(-\lambda -C)^{-1}D_{2j}\|\},$ 则对 $j \in \{1,2\},$$1 \leq \|(\lambda -B)^{-1}D_{1j}\| <\|(\lambda -B)^{-1}\|\varepsilon$ 或者 $1 \leq \|(-\lambda -C)^{-1}D_{2j}\| <\|(-\lambda -C)^{-1}\|\varepsilon,$ 即得 $\|(\lambda -B)^{-1}\|> \frac{1}{\varepsilon}$ 或者 $\|(-\lambda -C)^{-1}\|> \frac{1}{\varepsilon}.$ 因此, $\lambda \in (\sigma_{\varepsilon}(B) \cup \sigma_{\varepsilon}(-C)) \backslash (\sigma(B) \cup \sigma(-C)).$

${\bf 情形 2}$$ \max\limits_{j \in \{1,2\}} \{\|(\lambda -B)^{-1}D_{1j}\|, \|(-\lambda -C)^{-1}D_{2j}\|\} <1,$ 则对 $j \in \{1,2\},$$\|(\lambda -B)^{-1}D_{1j}\|< 1$$\|(-\lambda -C)^{-1}D_{2j}\|<1.$

考虑 $\|(\lambda -B)^{-1}D_{11}\|< 1,$ 由命题 2.2,

$I-(\lambda -B)^{-1}D_{11}=(\lambda -B)^{-1}(\lambda -B-D_{11})$ 是可逆的, 进而 $\lambda -B-D_{11}$

是可逆的. 类似地, 有 $-\lambda -C-D_{22}$ 是可逆的. 这表明 $\lambda \notin \sigma_{\varepsilon}(B) \cap \sigma_{\varepsilon}(-C),$$ \lambda \in \Sigma_{\varepsilon}(H,J)$ 矛盾. 因此, $\lambda \in (\sigma_{\varepsilon}(B) \cup \sigma_{\varepsilon}(-C)) \backslash (\sigma(B) \cup \sigma(-C)).$ 又由于 $B,C$ 是自伴算子, 有 $\sigma_{\varepsilon}(B)= \sigma(B)+ \rm{B} (0, \varepsilon), \sigma_{\varepsilon}(-C)= \sigma(-C) + \rm{B}(0, \varepsilon).$

${\bf例 4.1}$ 考虑调和方程

$ \left\{ \begin{aligned} &\frac{\partial^{2}u}{\partial x^2}+\frac{\partial^{2}u}{\partial y^2}=0, 0<x,y<1, \\ &u(0,y)=u(1,y)=0, \\ &u(x,0)=f(x), u(x,1)=g(x), \end{aligned} \right. $

化为无穷维 Hamilton 系统

$\dfrac{\partial}{\partial y}\begin{pmatrix}u\\v\end{pmatrix}=\begin{pmatrix} 0&I\\-\dfrac{\partial^2}{\partial x^2}&0\end{pmatrix}\begin{pmatrix}u\\v\end{pmatrix},$

$\mathcal{X}=L^{2}[0,1]$, 则由调和方程导出无穷维 Hamilton 算子

$H=\begin{pmatrix}A&B\\C&-A^*\end{pmatrix}=\begin{pmatrix}0&I\\-\dfrac{\partial^2}{\partial x^2}&0\end{pmatrix}:\mathcal{D}(C)\oplus\mathcal{D}(B)\to\mathcal{X}\oplus\mathcal{X},$

其定义域为

$\mathcal{D}(B)=\mathcal{X},$$\mathcal{D}(C)=\{u \in \mathcal{X}:u' \mbox{绝对连续} , u(0)=u(1)=0, u', u" \in \mathcal{X}\}.$

我们考虑方程 $(\lambda J-H)w=0, w=(u \quad v)^t$ 得到

$\left\{ \begin{aligned} &(\lambda I-I)v=0, \\ &\bigg(\frac{\partial^{2}}{\partial x^2}-\lambda I\bigg)u=0. \end{aligned} \right.$

由于 $\lambda$ 是不定的, 分情况讨论. 结合边界条件 $u(0)=u(1)=0,$

(1) 当 $\lambda=0$ 时, 得到 $u=v=0, (\lambda J-H)$ 是单射; (2) 当 $\lambda=1$ 时, $v$ 取任意非零元, $(\lambda J-H)$ 不是单射; (3) 当 $\lambda>0$$\lambda \neq 1$ 时, $u(x)=c_{1}{\rm e}^{\sqrt{\lambda}x}+c_{2}{\rm e}^{-\sqrt{\lambda}x},$ 得到 $c_1=c_2=0,$ 上述方程只有零解, $(\lambda J-H)$ 是单射; (4)当 $\lambda<0$ 时, $u(x)=c_{1}\cos\sqrt{-\lambda}x+c_{2}\sin\sqrt{-\lambda}x,$ 得到 $c_{1}=0,$$u(1)=c_{2}\sin\sqrt{-\lambda}=0,$ 解得 $\lambda_{k}=-(k\pi)^2, k=\pm1, \pm2,\cdots$, 因此 $(\lambda J-H)$ 我们声明, 对 $\lambda \in \sigma_{p}(H,J),$$\overline{(\lambda J-H)(\mathcal{D}(H))} \neq \mathcal{X} \oplus \mathcal{X}$ 并且 $\lambda \notin \sigma_{p}(H,J),$ 因此 $\overline{(\lambda J-H)(\mathcal{D}(H))} = \mathcal{X} \oplus \mathcal{X},$ 进而 $\sigma(H,J)=\sigma_{p}(H,J)=\{1\} \cup \{-(k\pi)^2: k=\pm1, \pm2,\cdots\}.$

另一方面, 用 $D$ 表示 $\dfrac{\rm d}{{\rm d}x},$

$\lambda J-H =\begin{pmatrix}0&\lambda I-I\\-\lambda I+D^2&0 \end{pmatrix},$

那么 $\|(\lambda J-H)^{-1}\|= \max \{ \frac{1}{|\lambda-1|}, \|(D^2-\lambda I)^{-1}\|\}.$ 因为 $D^2$ 是自伴的, 有 $\|(D^2-\lambda I)^{-1}\|=\frac{1}{\rm{dist}(\lambda, \sigma(D^2))},$$ \lambda \notin \sigma(D^2).$ 由上面的讨论知 $\sigma(D^2)=\{-(k\pi)^2: k=\pm1, \pm2,\cdots\}.$ 所以 $\Sigma_\varepsilon (H,J)=\sigma(H,J)+\rm{B}(0,\varepsilon).$ 这表明 $\Sigma_{\varepsilon}(H,J)$ 关于实轴对称且说明了推论 4.1(1) 和定理 4.4 的有效性.

文献 [30] 已论证 $\Sigma_{\varepsilon}(H,I)=\sigma_{\varepsilon}(H)$ 关于虚轴对称, 说明了推论 4.1(2) 的有效性, 本文是文献 [30] 的进一步推广.

参考文献

Nachtigal N M, Reddy S C, Trefethen L N.

How fast are nonsymmetric matrix iterations?

SIAM J Matrix Anal Appl, 1992, 13(3): 778-795

[本文引用: 1]

Kato T. Perturbation Theory for Linear Operators. New York: Springer-Verlag, 1976

[本文引用: 1]

Trefethen L N.

Pseudospectra of linear operators

Society Indus Appl Math, 1997, 39(3): 383-406

Trefethen L N, Embree M. Spectra and Pseudospectra:The Behavior of Nonnormal Matrices and Operator. Princeton: Princeton University Press, 2005

[本文引用: 2]

Böttcher A.

Pseudospectra and singular values of large convolution operators

J Integral Equations Appl, 1994, 6(3): 267-301

[本文引用: 1]

Böttcher A, Wolf H.

Spectral approximation for Segal-Bargmann space Toeplitz operators

Banach Center Publications, 1997, 38(1): 25-48

Davies E B.

Pseudospectra of differential operators

J Operator Theory, 2000, 43(2): 243-262

Cui J L, Li C K, Poon Y T.

Pseudospectra of special operators and pseudospectrum preservers

J Math Anal Appl, 2014, 419(2): 1261-1273

[本文引用: 1]

Dhara K, Kulkarni S H.

The $(n,\varepsilon)$-pseudospectrum of an element of a Banach algebra

J Math Anal Appl, 2018, 464(1): 939-954

[本文引用: 1]

Möller M, Pivovarchik V. Spectral Theory of Operator Pencils, Hermite-Biehler Functions, and Their Applications. Switzerland: Springer International Publishing, 2015

[本文引用: 1]

Guebbai H.

Generalized spectrum approximation and numerical computation of eigenvalue for Schrödinger operator

Lobachevskii J Math, 2013, 34(1): 45-60

[本文引用: 1]

Frayssé V, Gueury M, Nicoud F, et al. Spectral portraits for matrix pencils. Tech Report. Toulouse: CERFACS, 1996

[本文引用: 1]

Jos L M, Van D.

Pseudospectra for matrix pencils and stability of equilibria

Bit Numer Math, 1997, 37(4): 833-845

[本文引用: 1]

Ammar A, Bouchekoua A, Jeribi A.

A characterization of S-pseudospectra of linear operators in Hilbert space

Filomat, 2023, 37(5): 1331-1339

[本文引用: 2]

Jeribi A. Linear Operators and Their Essential Pseudospectra. Canada: Apple Academic Press, 2018

[本文引用: 2]

Ammar A, Bouchekoua A, Jeribi A.

New result of the pseudospectra of linear operators in Hilbert space

Filomat, 2023, 15(1): 13-26

[本文引用: 1]

Krishna K G, Augustine J.

A pseudospectral mapping theorem for operator pencil

Operators and Matrices, 2022, 16(3): 709-732

[本文引用: 1]

Khellaf A, Guebbai H, Lemita S, et al.

The pseudo-spectrum of operator pencils

Asian Eur J Math, 2020, 13(5): 2050100

[本文引用: 4]

Wu D Y, Chen A.

Invertibility of infinite-dimensional Hamiltonian operator and its applications

Sci China Math, 2010, 40(9): 921-928

[本文引用: 1]

Fan X Y, Chen A.

Spectral distribution and its invertibility of a class of nonnegative Hamilton operator

Acta Math Appl Sin, 2009, 32(1): 14-18

Kurina G A.

Invertibility of nonnegatively Hamilton operators in Hilbert space

Differ Equations, 2001, 37(6): 880-882

[本文引用: 1]

Hou G L, Chen A.

Completeness of generalized eigenfunction systems of infinite dimensional Hamiltonian operators and its application in elasticity

Sci China Math, 2012, 42(1): 57-68

[本文引用: 1]

Chen A, Jin G H, Wu D Y.

On symplectic self-adjointness of Hamiltonian operator matrices

Sci China Math, 2015, 58(4): 821-828

[本文引用: 1]

Wu D Y, Chen A.

Symplectic self-adjoint of infinite dimensional Hamiltonian operator

Acta Math Appl Sin, 2011, 34(5): 918-923

[本文引用: 1]

Chen A, Wu D Y, Hai G J.

Numerical range of infinite dimensional Hamiltonian operator

J Sys Sci & Math Scis, 2013, 33(4): 506-510

[本文引用: 1]

Yu J H, Chen A. Zhao Y J, et al.

Symmetry of a class of $n$-numerical range for Hamilton operator

Math Prac Theory, 2014, 44(2): 193-198

[本文引用: 1]

Huang J J, Chen A, Fan X Y.

The spectral structure of infinite dimensional Hamilton operator

Sci China Math, 2008, 38(1): 71-78

[本文引用: 1]

Wang H, Chen A, Huang J J.

The residual spectrum of upper triangular infinite dimensional Hamiltonian operator

Math Prac Theory, 2010, 40(3): 195-201

Bai E, Chen A.

Spectral distribution and its invertibility of a class of Hamiltonian operator

Acta Math Appl Sin, 2012, 35(6): 1113-1119

Shen R S, Hou G L.

On the pseudospectra and the related properties of infinite dimensional Hamiltonian operator

Linear Multilinear Algebra, 2021, 70(2): 1-12

[本文引用: 4]

Conway J B. A Course in Functional Analysis. New York: Springer-Verlag, 1990

[本文引用: 1]

Kreyszig E. Introductory Functional Analysis with Application. New York: Wiley, 1989

[本文引用: 1]

/