数学物理学报, 2022, 42(3): 784-806 doi:

论文

脉冲离散Ginzburg-Landau方程组的统计解及其极限行为

赵才地,1, 姜慧特1, 李春秋1, TomásCaraballo2

1 温州大学数理学院 浙江温州 325035

2 塞维利亚大学数学系 西班牙塞维利亚 41012

Statistical Solutions and Its Limiting Behavior for the Impulsive Discrete Ginzburg-Landau Equations

Zhao Caidi,1, Jiang Huite1, Li Chunqiu1, Tomás Caraballo2

1 Department of Mathematics, Wenzhou University, Zhejiang Wenzhou 325035

2 Departmento de Ecuaciones Diferenciales y Análisis Numérico, Facultad de Mathmáticas, Universidad de Sevilla, c/Tarfia s/n, 41012-Sevilla, Spain

通讯作者: 赵才地, E-mail: zhaocaidi2013@163.com

收稿日期: 2021-04-23  

基金资助: 国家自然科学基金.  11971356
浙江省自然科学基金.  LY17A010011

Received: 2021-04-23  

Fund supported: the NSFC.  11971356
the NSF of Zhejiang Province.  LY17A010011

Abstract

In this article we first prove the global well-posedness of the impulsive discrete Ginzburg-Landau equations. Then we establish that the generated process by the solution operators possesses a pullback attractor and a family of invariant Borel probability measures. Further, we formulate the definition of statistical solution for the addressed impulsive system and prove the existence. Our results reveal that the statistical solution of the impulsive system satisfies merely the Liouville type theorem piecewise, which implies that the Liouville type equation for impulsive system will not always hold true on the interval containing any impulsive point. Finally, we prove that the statistical solution of the impulsive discrete Ginzburg-Landau equations converges to that of the impulsive discrete Schrödinger equations.

Keywords: Statistical solution ; Impulsive differential equation ; Liouville type theorem ; Discrete complex Ginzburg-Landau equation ; Discrete Schrödinger equation

PDF (497KB) 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

赵才地, 姜慧特, 李春秋, TomásCaraballo. 脉冲离散Ginzburg-Landau方程组的统计解及其极限行为. 数学物理学报[J], 2022, 42(3): 784-806 doi:

Zhao Caidi, Jiang Huite, Li Chunqiu, Tomás Caraballo. Statistical Solutions and Its Limiting Behavior for the Impulsive Discrete Ginzburg-Landau Equations. Acta Mathematica Scientia[J], 2022, 42(3): 784-806 doi:

1 引言

本文研究下面脉冲离散Ginzburg-Landau方程组解的存在性, 分段Liouville型定理及其极限行为

$ \begin{equation} {\rm i}\frac{{\mathrm d}u_m}{{\mathrm d}t} -(\alpha-{\rm i}\epsilon)(2u_{m}-u_{m+1}-u_{m-1}) +{\rm i}\kappa u_{m} +\beta|u_m|^{2\gamma}u_m=g_m(t), \; t>\tau, \, t\neq \tau_j, \, \, m, j\in {\Bbb Z}, \end{equation} $

其中脉冲条件和初值条件分别为

$ \begin{equation} u_m(\tau_j^+)-u_m(\tau_j)=\phi_{mj}(u_m(\tau_j)), \; m, j\in {\mathbb Z}, \; \tau_j\in {\mathbb R}, \end{equation} $

$ \begin{equation} u_m(\tau^+)=\lim\limits_{s\rightarrow \tau^+}u_m(s)=u_{m, \tau^+}, \; m\in {\mathbb Z}, \end{equation} $

这里$ \tau\in {\mathbb R} $是初始时间, $ u_m(\cdot)\in {\mathbb C} $是未知的函数, $ g_m(\cdot) $$ \phi_{mj}(\cdot) $是给定的满足一定条件的复值函数, $ \{\tau_j\}_{j\in {\Bbb Z}} $是一列给定的实数, $ \eta $是给定的正数, 满足条件

$ \begin{eqnarray} \tau_{j+1}-\tau_j\geqslant \eta, \; j\in {\mathbb Z}, \; \mbox{和}\; \lim\limits_{j\rightarrow +\infty}\tau_j=+\infty, \; \lim\limits_{j\rightarrow -\infty}\tau_j=-\infty. \end{eqnarray} $

另外$ \alpha $, $ \epsilon $, $ \kappa $$ \beta $都是正的常数, 且$ 1<\gamma<+\infty $; $ {\mathbb Z} $, $ {\mathbb R} $$ {\mathbb C} $分别表示整数集, 实数集和复数集; $ {\rm i} $是虚数单位.当$ \epsilon=0 $时, 方程组(1.1)–(1.3)转化成下面脉冲离散非线性Schrödinger方程组

$ \begin{eqnarray} \left\{ \begin{array}{ll} {\rm i}\frac{{\mathrm d}u_m}{{\mathrm d}t} -\alpha(2u_{m}-u_{m+1}-u_{m-1}) +{\rm i}\kappa u_{m} +\beta|u_m|^{2\gamma}u_m=g_m(t), \; t>\tau, \, t\neq \tau_j, \, \, m, j\in {\Bbb Z}, \\ u_m(\tau_j^+)-u_m(\tau_j)=\phi_{mj}(u_m(\tau_j)), \; m, j\in {\mathbb Z}, \; \tau_j\in {\mathbb R}, \\ u_m(\tau^+)=\lim\limits_{s\rightarrow \tau^+}u_m(s)=u_{m, \tau^+}, \; m\in {\mathbb Z}. \end{array} \right. \end{eqnarray} $

问题(1.1)–(1.3)可以看作是下面Ginzburg-Landau方程关于空间变量$ x\in {{\Bbb R}} $的离散近似

$ \begin{eqnarray} \left\{ \begin{array}{ll} {\rm i}\frac{\partial u}{\partial t} +(\alpha-{\rm i}\epsilon)\Delta u+{\rm i}\kappa u+\beta|u|^{2\gamma}u=g(t), \; t>\tau, \; t\neq\tau_j, \, j\in {\mathbb Z}, \\ u(\tau_j^+)-u(\tau_j)=\phi_j(u(\tau_j)), \; j\in {\Bbb Z}, \;\tau_j\in {{\Bbb R}} , \\ u(\tau)=u_\tau. \end{array} \right. \end{eqnarray} $

类似地, 问题(1.5)可以看作下面非线性Schrödinger方程关于空间变量$ x\in {{\Bbb R}} $的离散近似

$ \begin{eqnarray} \left\{ \begin{array}{ll} {\rm i}\frac{\partial u}{\partial t} +\alpha\Delta u+{\rm i}\kappa u+\beta|u|^{2\gamma}u=g(t), \; t>\tau, \; t\neq\tau_j, \, j\in {\mathbb Z}, \\ u(\tau_j^+)-u(\tau_j)=\phi_j(u(\tau_j)), \; j\in {\Bbb Z}, \;\tau_j\in {{\Bbb R}} , \\ u(\tau)=u_\tau. \end{array} \right. \end{eqnarray} $

Ginzburg-Landau方程和Schrödinger方程都是量子力学和等离子体物理学中的重要方程, 已经有很多文献研究它们解的各种性质[26, 29].然而, 据我们所知, 目前没有文献研究脉冲离散Ginzburg-Landau方程和脉冲离散Schrödinger方程解的性质.

脉冲问题(1.1)–(1.3)的基本特点是解不具有连续性, 在$ t\in {{\Bbb R}} $$ t\neq \tau_j $时解是左连续的, 脉冲点$ \{\tau_j\}_{j\in {\mathbb Z}} $属于解的第一类间断点.事实上, 条件(1.2)描述了脉冲对系统的影响, 其导致了解的分段连续性.脉冲微分方程在模拟科学和技术领域中有非常重要的应用[1, 2, 6, 18, 36, 37].目前已有一些文献研究了脉冲系统的动力学行为, 例如Ciesielski在文献[15-17]中对脉冲动力系统的半连续性和稳定性进行了详细研究.此外, 已有一些文献研究了脉冲反应-扩散方程组的长期行为[23, 24, 34, 41].然而, 据我们所知, 目前很少有文献研究脉冲微分方程或脉冲格点系统的统计解.

格点系统是指某些变量具有离散性的时空系统, 包括耦合微分方程、耦合映射格点和元胞自动机[14].在某些情况下, 偏微分方程的空间离散化会出现格点系统.格点系统已被广泛地应用于电气工程[13], 化学反应理论[27]等多个领域.格点系统的渐近理论已被广泛地研究, 参见文献[3, 10, 22, 38, 52].最近, 文献[43]研究了格点Klein-Gordon-Schrödinger方程组的不变测度.

不变测度和统计解对于人们理解湍流很有帮助, 这是由于湍流中的一些重要物理量(比如质量, 速度)本质上都依赖于这些量关于时间的均值[9, 12, 20, 33].目前已有一些文献研究了演化方程的统计解.例如, 文献[20, 21]系统地研究了三维Navier-Stokes方程组的统计解; 文献[31]构造了二维Navier-Stokes方程组的统计解; 文献[7, 9]给出了一般演化方程组统计解存在的抽象框架; 文献[11, 28, 40, 42]研究了三维全局修正的Navier-Stokes方程组的不变测度和统计解; 文献[46, 50]研究了非自治磁微极流和Klein-Gordon-Schrödinger方程组的统计解.文献[45]中应用无穷动力系统的理论给出了一般自治演化方程轨道统计解存在的充分条件.最近, 文献[30, 39, 53]研究了一阶格点系统和离散长波-短波共振方程组的不变测度.特别地, 文献[51]研究了脉冲格点反应-扩散方程组的统计解和分段Liouville型定理.

本文的主要结果是证明脉冲格点Ginzburg-Landau方程组统计解的存在性, 分段Liouville型定理及其极限行为.首先, 我们证明脉冲问题(1.1)–(1.3)解的全局适定性, 接着证明解算子生成的过程存在拉回吸引子和一族Borel不变概率测度, 进而给出脉冲系统统计解的定义并证明其存在性.最后, 我们证明当$ \epsilon\rightarrow 0^+ $时, 脉冲离散Ginzburg-Landau方程组的统计解收敛于脉冲离散Schrödinger方程组(1.5)的统计解.我们的结果揭示了脉冲系统的统计解只分段地满足Liouville型定理.

在本文研究过程中, 脉冲会带来两个困难.首先, 脉冲会导致解不连续.这种不连续性在我们讨论解的估计时带来一些困难, 而且在我们研究解算子生成的过程$ \{U(t, \tau)\}_{t\geqslant \tau} $的拉回有界吸收性和拉回渐近零性时也产生了困难.造成这些困难的原因是在包含脉冲时间点的区间上Gronwall不等式不再成立.为了研究解的分段连续性, 我们将文献[2, 引理2.2]中的脉冲不等式推广到函数集$ PC^1({\mathbb R}, {\mathbb R}) $上(详参§2中的符号).其次, 由于脉冲的原因, 使得讨论的问题是非自治的(即使外力项$ g $是不依赖于时间$ t $的).自治系统与非自治系统本质的区别是系统起止时间参数的连续依赖性.在本文中, 在$ [\tau, +\infty) $上, 对于每个$ \tau\in {{\Bbb R}} $$ u_*\in \ell^2 $, $ \ell^2 $ -值函数$ t\mapsto U(t, \tau)u_* $是左连续的, 但是当$ \tau\rightarrow t_*^- $时, $ \|U(t_*, \tau)u_*-u_*\|\rightarrow 0 $的收敛性仍然依赖于初始时刻$ \tau $.事实上, 当$ \tau\rightarrow t_*^- $时, 随着初始时刻$ \tau $的变化, 问题的解$ U(t_*, \tau)u_* $也会随之变化.这是由脉冲效应引起的非自治性质造成的.我们将利用所讨论的脉冲系统的结构来证明$ \ell^2 $ -值映射$ \tau\mapsto U(t, \tau)u_* $有以下分段连续性:当$ \tau<t $, $ j\in {\Bbb Z} $, $ \tau\neq\tau_j $时, 映射是连续的; 当$ \tau\in (-\infty, t] $时, 映射是左连续的; 脉冲时间点$ \{\tau_j\}_{j\in {\Bbb Z}} $属于该映射的第一类间断点.

值得一提的是, 脉冲效应也会导致不变测度$ \{{\mathit{ m}}_t\}_{t\in {{\Bbb R}} } $的不连续性, 也即是说, 对于给定的$ \psi\in C(\ell^2) $, 函数$ t\rightarrow \int_{\ell^2} \psi (u){\mathrm d}{\mathit{ m}}_t(u) $并不总是连续的.因此, 我们尝试修订文献[9, 定义3.2]中统计解的定义, 使得修订后的定义适用于脉冲离散Ginzburg-Landau方程组.然而, 我们修订后的统计解的定义也适用于脉冲离散Schrödinger方程组(1.5)和其他脉冲微分方程.在我们修订后的统计解定义中, 函数$ t\rightarrow \int_{\ell^2} \psi (u){\mathrm d}{\mathit{ m}}_t(u) $是分段连续的, 统计解$ \{{\mathit{ m}}_t\}_{t\in {{\Bbb R}} } $只分段地满足Liouville型定理.这些现象本质上都是由脉冲效应导致的.

本文安排如下.在第2节, 我们先介绍一些符号, 然后证明问题(1.1)–(1.3)的全局适定性.在第3节, 我们先证明问题(1.1)–(1.3)的解算子在$ \ell^2 $上生成一个连续的过程$ \{U(t, \tau)\}_{t\geqslant \tau} $且过程存在一个有界的拉回吸收集.然后我们证明过程$ \{U(t, \tau)\}_{t\geqslant \tau} $具有拉回渐近零性质且存在拉回吸引子.在第4节, 我们先证明过程$ \{U(t, \tau)\}_{t\geqslant \tau} $关于$ \tau $的某些分段连续性, 接着, 改进文献[32, 定理3.1]的结果, 构造过程$ \{U(t, \tau)\}_{t\geqslant \tau} $的一族Borel不变概率测度, 然后证明这族不变测度分段地满足Liouville型定理并且是脉冲方程(1.1)的统计解.最后, 在第5节中, 我们证明当$ \epsilon\rightarrow 0^+ $时, 脉冲离散Ginzburg-Landau方程组的统计解收敛于脉冲离散Schrödinger方程组的统计解.

2 解的全局适定性

在这一节中, 我们首先介绍一些符号和算子, 然后我们证明问题(1.1)–(1.3)和(1.5)的全局适定性.文后我们用$ {\mathbb R}_+ $$ {\mathbb Z}_+ $分别表示正实数集和正整数集.记

并在其上定义内积和范数:

其中$ \overline{v}_m $$ v_m $的共轭.则$ (\ell^2, (\cdot, \cdot)) $是一个Hilbert空间.在$ \ell^2 $上定义三个线性算子$ B, B^* $$ A $:

易证$ B^* $是算子$ B $的自伴算子, 且有

$ \begin{eqnarray} (Au, v)=(B^*Bu, v)=(Bu, Bv), \; \forall\, u, v\in \ell^2, \end{eqnarray} $

$ \begin{eqnarray} \left\{ \begin{array}{ll} \|Bu\|^2=\|B^*u\|^2\leqslant 4\|u\|^2, \;& \forall\, u\in \ell^2, \\ \|Au\|^2\leqslant 16\|u\|^2, \; &\forall\, u\in \ell^2. \end{array} \right. \end{eqnarray} $

为了描述所讨论的脉冲微分问题解的某种连续性, 对于满足条件(1.4)的给定的脉冲时间点$ \{\tau_j\}_{j\in {\Bbb Z}} $, 我们引入下面两个函数空间$ PC(I; {{\Bbb R}} ) $$ PC(I; \ell^2) $:

另外, $ PC^1(I; {{\Bbb R}} ) $$ PC^1(I; \ell^2) $分别表示一阶导数属于$ PC(I; {{\Bbb R}} ) $$ PC(I; \ell^2) $的全体函数.

为了将问题(1.1)–(1.3)和(1.5)写成向量的形式, 我们记

应用上面介绍的符号和算子, 问题(1.1)–(1.3)可以写成如下形式

$ \begin{equation} {\rm i}\frac{{\mathrm d} u}{{\mathrm d}t}-(\alpha-{\rm i}\epsilon) Au+{\rm i}\kappa u+\beta|u|^{2\gamma}u=g(t), \; t>\tau, \, t\neq \tau_j, \, j\in {\Bbb Z}, \end{equation} $

$ \begin{equation} u(\tau_j^+)-u(\tau_j)=\phi_j(u(\tau_j)), \; \, j\in {\Bbb Z}, \end{equation} $

$ \begin{equation} u(\tau^+)=u_{\tau^+}. \end{equation} $

类似地, 问题(1.5)可以写成

$ \begin{eqnarray} \left\{ \begin{array}{ll} {\rm i}\frac{{\mathrm d} u}{{\mathrm d}t}-\alpha Au+{\rm i}\kappa u+\beta|u|^{2\gamma}u=g(t), \; t>\tau, \, t\neq \tau_j, \, j\in {\Bbb Z}, \\ u(\tau_j^+)-u(\tau_j)=\phi_j(u(\tau_j)), \; \, j\in {\Bbb Z}, \\ u(\tau^+)=u_{\tau^+}. \end{array} \right. \end{eqnarray} $

为了证明问题(2.3)–(2.5)和(2.6)解的全局适定性(也为了保证拉回吸引子的存在性), 我们假设函数$ g(\cdot) $$ \phi_j(\cdot)=(\phi_{mj}(\cdot))_{m\in {\Bbb Z}} $满足下面条件.

(H1) 对任意的$ m, j\in {\Bbb Z} $, $ \phi_{mj}(0)=0 $, 存在一个常数$ L>0 $使得

$ \begin{equation} |\phi_{mj}(z')-\phi_{mj}(z'')| \leqslant L|z'-z''|, \; \forall\, z', z''\in {\Bbb C}, \end{equation} $

$ \begin{equation} \sigma:=\kappa-\frac{1}{\eta}\ln(2+2L^2)>0, \end{equation} $

其中常数$ \kappa $来自于方程(1.1), $ \eta $来自于(1.4)式.

(H2) $ g(\cdot)\in C(\ell^2) $, 且对任意的$ t\in {{\Bbb R}} $$ \int_{-\infty}^t {\rm e}^{\sigma s}\|g(s)\|^2{\mathrm d}s<+\infty $.

引理2.1  假设$ \alpha $, $ \epsilon $, $ \kappa $, $ \beta $都是正的常数, $ 1<\gamma<+\infty $$ \rm ({H1})–({H2}) $成立.则对于每个初始时刻$ \tau $和初始值$ u_{\tau^+}\in\ell^2 $, 问题(2.3)–(2.5)存在唯一的解$ u $满足

其中$ T_*>\tau $, 且当$ T_*<+\infty $时有$ \lim\limits_{t\rightarrow T_*^-}\|u(t)\|=+\infty $.

  将方程(2.3)改写成下面形式

$ \begin{eqnarray} \frac{{\mathrm d} u}{{\mathrm d}t} = F^\epsilon(u, t) := (-\epsilon-{\rm i}\alpha) Au-\kappa u+{\rm i}\beta f(|u|^{2})u-{\rm i}g(t), \; t>\tau, \, t\neq \tau_j, \, j\in {\Bbb Z}, \end{eqnarray} $

其中

$ \begin{equation} f(x)=x^\gamma, \; x\in {{\Bbb R}} _+. \end{equation} $

因为$ u\longmapsto A u $$ u\longmapsto \kappa u $是从$ \ell^2 $$ \ell^2 $的有界线性算子, 且$ g(\cdot)\in C(\ell^2) $, 因此我们只需要证明非线性项$ f(|u|^2)u $满足局部Lipschitz条件.事实上, 由于$ f'(x)=\gamma x^{\gamma-1} $$ {{\Bbb R}} _+ $上连续, 由文献[26, 引理2.1]可知由$ \tilde{f}(u)=f(|u|^2)u=(|u_m|^{2\gamma}u_m)_{m\in {\Bbb Z}} $定义的算子$ \tilde{f}(\cdot) $$ \ell^2 $的有界集满足Lipschitz条件.也即, 对于$ \ell^2 $中的一个有界集$ {\cal B} $, 有

$ \begin{eqnarray} \|\tilde{f}(u)-\tilde{f}(v)\|^2 \leqslant M_f({\mathcal B})\|u-v\|^2, \; \forall\, u, v\in {\mathcal B}, \end{eqnarray} $

其中$ M_f({\mathcal B}) $是只依赖于$ {\mathcal B} $的常数.根据脉冲微分方程的经典理论(参见文献[2, 定理2.3和2.6]), 引理2.1得证.

引理2.2  假设函数$ y(\cdot)\in PC^1({{\Bbb R}} ; {{\Bbb R}} ) $满足

$ \begin{eqnarray} \left\{ \begin{array}{ll} \frac{{\mathrm d}y(t)}{{\mathrm d}t}+ay(t) \leqslant q(t), &\; t\neq \tau_j, \, j\in {\Bbb Z}, \\ y(\tau_j^+)-y(\tau_j) \leqslant by(\tau_j), &\; j\in {\Bbb Z}, \\ y(s^+) \leqslant y_{0}, &\; s\in {{\Bbb R}} , \end{array} \right. \end{eqnarray} $

其中$ q(\cdot)\in PC({{\Bbb R}} ; {{\Bbb R}} ) $, $ a>0 $, $ b>0 $$ y_{0} $均是常数.则

$ \begin{eqnarray} y(t) \leqslant y_{0}(1+b)^{n(s, t)}{\rm e}^{-a(t-s)} +\int_s^t(1+b)^{n[\vartheta, t)}{\rm e}^{-a(t-\vartheta)}q(\vartheta){\mathrm d}\vartheta, \; \forall \, t>s, \end{eqnarray} $

这里以及文后$ n(s, t) $$ n[\theta, t) $表示脉冲点$ \{\tau_j\}_{j\in {\Bbb Z}} $分别落在区间$ (s, t) $$ [\theta, t) $中的个数.

  该引理是文献[2]中引理2.2的改进.事实上, 对于给定的$ s\in {{\Bbb R}} $, 存在某个$ j_0\in {\Bbb Z} $使得$ s\in (\tau_{j_0}, \tau_{j_0+1}] $, 由假设$ y(t) $$ (s, \tau_{j_0+1}) $上满足(2.12)式中的微分不等式.在$ (s, \tau_{j_0+1}) $上应用Gronwall不等式和$ y(t) $的左连续性, 可知$ y(t) $$ t\in (s, \tau_{j_0+1}] $上满足(2.13)式.然后在区间$ (\tau_{j_0+1}, \tau_{j_0+2}] $考虑(2.12)式, 其初始值满足$ y(\tau^+_{j_0+1}) \leqslant (1+b)y(\tau_{j_0+1}) $.同样在$ (\tau_{j_0+1}, \tau_{j_0+2}) $应用Gronwall不等式和$ y(t) $的左连续性, 可知当$ t\in (\tau_{j_0+1}, \tau_{j_0+2}] $时, $ y(t) $也满足(2.13)式.相似地, 我们可以证明对于$ t\in (\tau_{j_0+m}, \tau_{j_0+m+1}) $, $ y(t) $满足(2.13)式, 其中$ m, j \in {\Bbb Z} $.证明的细节在这里省略.

引理2.3  假设$ \alpha $, $ \epsilon $, $ \kappa $, $ \beta $都是正的常数, $ 1<\gamma<+\infty $$ \rm ({H1})–({H2}) $成立.则对于每个初始时刻$ \tau $和初始值$ u_{\tau^+}\in\ell^2 $, 问题(2.3)–(2.5)的解满足

$ \begin{eqnarray} \|u(t)\|^2 \leqslant \|u_{\tau^+}\|^2{\rm e}^{-\sigma(t-\tau)} +\frac{{\rm e}^{-\sigma t}}{\kappa} \int_\tau^t{\rm e}^{\sigma \theta}\|g(\theta)\|^2{\mathrm d}\theta, \; \tau<t\leqslant T_*, \end{eqnarray} $

其中常数$ \sigma $由(2.8)式给定.

  记$ u(\cdot)=u(\cdot; \tau, u_{\tau^+}) $为问题(2.3)–(2.5)在初始时刻$ \tau $$ u_{\tau^+} $为初值的解.用$ u(\cdot) $与(2.9)式在$ \ell^2 $上作内积并取实部, 可得

$ \begin{eqnarray} \frac{{\mathrm d}}{{\mathrm d}t}\|u(t)\|^2 +\kappa \|u(t)\|^2 \leqslant \frac{\|g(t)\|^2}{\kappa}, \; t\neq\tau_j, j\in {\Bbb Z}. \end{eqnarray} $

由假设(H1)得

$ \begin{eqnarray} \|u(\tau_j^+)\|^2 &=&\sum\limits_{m\in {\Bbb Z}}|u_m(\tau^+_j)|^2 =\sum\limits_{m\in {\Bbb Z}}\big|u_m(\tau_j)+\phi_{mj}(u_m(\tau_j))\big|^2 \\ &\leqslant& 2\sum\limits_{m\in {\Bbb Z}}|u_m(\tau_j)|^2 +2\sum\limits_{m\in {\Bbb Z}}|\phi_{mj}(u_m(\tau_j))|^2 \\ &\leqslant& (2+2L^2)\sum\limits_{m\in {\Bbb Z}}|u_m(\tau_j)|^2 = (2+2L^2)\|u(\tau_j)\|^2. \end{eqnarray} $

$ y(t)=\|u(t)\|^2 $, 对(2.15)–(2.16)式应用引理2.2可得

$ \begin{equation} \|u(t)\|^2 \leqslant \|u_{\tau^+}\|^2(2+2L^2)^{n(\tau, t)}{\rm e}^{-\kappa(t-\tau)} +\frac{1}{\kappa}\int_\tau^t(2+2L^2)^{n[s, t)} {\rm e}^{-\kappa(t-s)}\|g(s)\|^2{\mathrm d}s, \; \, \forall\, t>\tau. \end{equation} $

而(1.4)式意味着

因此, 由(2.8)式可得

$ \begin{eqnarray} (2+2L^2)^{n(\tau, t)}{\rm e}^{-\kappa(t-\tau)} \leqslant {\rm e}^{-\sigma(t-\tau)} \; \mbox{和}\; (2+2L^2)^{n[s, t)} {\rm e}^{-\kappa(t-s)} \leqslant {\rm e}^{-\sigma(t-s)}. \end{eqnarray} $

将(2.18)式代入(2.17)式得到(2.14)式.证明完毕.

结合引理2.1和引理2.3知问题(2.3)–(2.5)存在唯一的全局解.

定理2.1  假设$ \alpha $, $ \epsilon $, $ \kappa $, $ \beta $都是正的常数, $ 1<\gamma<+\infty $$ \rm ({H1})–({H2}) $成立, 则对于每一个给定的初始时刻$ \tau\in {{\Bbb R}} $和初值$ u(s^+)\in \ell^2 $, 问题(2.3)–(2.5)存在唯一的解$ u \in PC([\tau, +\infty);\ell^2)\cap PC^1((\tau, +\infty);\ell^2) $满足

$ \begin{eqnarray} \|u(t)\|^2 \leqslant \|u_{\tau^+}\|^2{\rm e}^{-\sigma(t-\tau)} +\frac{{\rm e}^{-\sigma t}}{\kappa} \int_\tau^t{\rm e}^{\sigma \theta}\|g(\theta)\|^2{\mathrm d}\theta, \; \forall\, t>\tau. \end{eqnarray} $

下面证明问题(2.3)–(2.5)的解连续地依赖于初值.

定理2.2  假设$ \alpha $, $ \epsilon $, $ \kappa $, $ \beta $都是正的常数, $ 1<\gamma<+\infty $$ \rm ({H1})–({H2}) $成立, 用$ u^{(k)}(\cdot)=u^{(k)}(\cdot; \tau, u^{(k)}(\tau^+)) $$ (k=1, 2) $分别表示问题(2.3)–(2.5)在初始时刻$ \tau $$ u^{(k)}_{\tau^+} $$ (k=1, 2) $为初值的解.则有

$ \begin{equation} \|u^{(1)}(t)-u^{(2)}(t)\|^2 \leqslant \|u^{(1)}_{\tau^+}-u^{(2)}_{\tau^+}\|^2 {\rm e}^{-(\sigma+\kappa-2\ell(R)) (t-\tau)}, \; \forall\, t>\tau, \end{equation} $

其中$ \ell(R) $是只依赖于$ u^{(1)}_{\tau^+} $$ u^{(2)}_{\tau^+} $的常数.

  用$ u^{(k)}(\cdot)=u^{(k)}(\cdot; \tau, u^{(k)}(\tau^+)) $$ (k=1, 2) $分别表示问题(2.3)–(2.5)在初始时刻$ \tau $$ u^{(k)}_{\tau^+} $$ (k=1, 2) $为初值的解, 记$ v(\cdot)=u^{(1)}(\cdot)-u^{(2)}(\cdot) $, 则$ v(\cdot) $满足

$ \begin{equation} {\rm i}\frac{{\mathrm d} v}{{\mathrm d}t} -(\alpha-{\rm i}\epsilon) Av+{\rm i}\kappa v+\beta(\tilde{f}(u^{(1)})-\tilde{f}(u^{(2)}))=0, \; t>\tau, \, t\neq \tau_j, \, j\in {\Bbb Z}, \end{equation} $

$ \begin{equation} v(\tau_j^+)-v(\tau_j)=\phi_j(u^{(1)}(\tau_j))-\phi_j(u^{(2)}(\tau_j)), \; \, j\in {\Bbb Z}, \end{equation} $

$ \begin{equation} v(\tau^+)=u^{(1)}_{\tau^+}-u^{(2)}_{\tau^+}, \; \tau\in {{\Bbb R}} . \end{equation} $

$ -{\rm i}v $与(2.21)式两边在$ \ell^2 $上作内积并取实部, 得

$ \begin{eqnarray} \frac12\frac{{\mathrm d}}{{\mathrm d}t}\|v\|^2 +\epsilon \|B v\|^2+\kappa \|v\|^2+{\bf{Im}}\beta(\tilde{f}(u^{(1)})-\tilde{f}(u^{(2)}), v) =0, \; t\neq\tau_j, j\in {\Bbb Z}. \end{eqnarray} $

由文献[26, (21)式]可知

$ \begin{equation} \big|{\bf{Im}} (\tilde{f}(u^{(1)})-\tilde{f}(u^{(2)}), v)\big| \leqslant \ell(R)\|v\|^2, \end{equation} $

其中$ \ell(R) $是只依赖于$ u^{(1)}_{\tau^+} $$ u^{(2)}_{\tau^+} $的常数.由(2.24)和(2.25)式得

$ \begin{eqnarray} \frac{{\mathrm d}}{{\mathrm d}t}\|v\|^2 +2(\kappa-\ell(R)) \|v\|^2 \leqslant 0, \; t\neq\tau_j, \, j\in {\Bbb Z}. \end{eqnarray} $

由假设$ \rm {(H1)} $和直接计算, 有

$ \begin{eqnarray} \|v(\tau_j^+)\|^2 &=& \sum\limits_{m\in {\Bbb Z}}|v_m(\tau_j^+)|^2{}\\ &=& \sum\limits_{m\in {\Bbb Z}}\big |v_m(\tau_j) +\phi_{mj}(u^{(1)}_m(\tau_j))-\phi_{mj}(u^{(2)}_m(\tau_j))\big|^2 \\ &\leqslant& 2\sum\limits_{m\in {\Bbb Z}}|v_m(\tau_j)|^2 + 2\sum\limits_{m\in {\Bbb Z}}\big|\phi_{mj}(u^{(1)}_m(\tau_j)) -\phi_{mj}(u^{(2)}_m(\tau_j))\big|^2 \\ &\leqslant& 2\sum\limits_{m\in {\Bbb Z}}|v_m(\tau_j)|^2 + 2\sum\limits_{m\in {\Bbb Z}}L^2 \big|u^{(1)}_m(\tau_j)-u^{(2)}_m(\tau_j)\big|^2\\ &\leqslant& (2+2L^2)\|v(\tau_j)\|^2. \end{eqnarray} $

$ y(t)=\|v(t)\|^2 $, 由(2.26)–(2.27)式和引理2.2得

$ \begin{eqnarray} \|v(t)\|^2 \leqslant \|v_{\tau^+}\|^2(2+2L^2)^{n(\tau, t)}{\rm e}^{-2(\kappa-\ell(R))(t-\tau)}, \; \, \forall\, t>\tau. \end{eqnarray} $

由(2.18)式知

$ \begin{eqnarray} (2+2L^2)^{n(\tau, t)}{\rm e}^{-2(\kappa-\ell(R))(t-\tau)} \leqslant {\rm e}^{-(\sigma+\kappa-2\ell(R))(t-\tau)}. \end{eqnarray} $

将(2.29)式代入(2.28)式即得(2.20)式.证明完毕.

3 拉回吸引子的存在性

定理2.1表明问题(2.3)–(2.5)的解算子

$ \begin{eqnarray} U_{\epsilon}(t, \tau): u(\tau^+)\in \ell^2 \longmapsto U_{\epsilon}(t, \tau)u(\tau^+)=u(t;\tau, u(\tau^+)) \end{eqnarray} $

$ \ell^2 $上生成一个过程.文后我们用$ u^\epsilon(\cdot;\tau, u_{\tau^+}) $表示问题(2.3)–(2.5)在初始时刻$ \tau $$ u(\tau^+) $为初值的解.从定理2.2可知过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $$ \ell^2 $上是连续的, 也即是说, 对于每个给定的$ t $$ \tau $, 其中$ \tau\leqslant t $, 映射$ U_\epsilon(t, \tau): \ell^2\longmapsto \ell^2 $是连续的.

文后我们用$ {\cal P}(\ell^2) $表示$ \ell^2 $的所有子集族, 考虑时间$ t $参数化的非空集合族$ \widehat{D}_0=\{D_0(t) :t\in{{\Bbb R}} \}\subseteq{\cal P}(\ell^2) $.

$ \begin{eqnarray} {\mathcal D}_\sigma =\Big\{\widehat{D} =\{D(t):t\in{{\Bbb R}} \}\subseteq{\cal P}(\ell^2) |\lim\limits_{\tau\rightarrow -\infty} ({\rm e}^{\sigma \tau} \sup\limits_ {u\in D(\tau)}\|u(\cdot)\|^2)=0\Big\}. \end{eqnarray} $

下一步我们回顾与过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $的拉回吸引子相关的一些定义.

定义3.1[31]  对任意的$ \epsilon>0 $.

(1) 设$ \widehat{D}_0=\{D_0(s):s\in {\mathbb R}\}\subseteq {\mathcal P}(\ell^2) $, 对每个$ s\in {\mathbb R} $, $ D_0(s)\subset\ell^2 $是有界集.若对任意的$ t\in {\mathbb R} $, $ \widehat{D}=\{D(s):s\in {\mathbb R}\}\in {\mathcal D}_\sigma $, 存在一个$ \tau_{0, \epsilon}(t, \widehat{D})\leqslant t $使得$ U_\epsilon(t, \tau)D(\tau)\subseteq D_0(t) $对所有的$ \tau\leqslant \tau_{0, \epsilon}(t, \widehat{D}) $成立, 则称$ \widehat{D}_0 $是过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $$ \ell^2 $上的有界拉回$ {\mathcal D}_\sigma $ -吸收集.

(2) 对任意给定的$ t\in {{\Bbb R}} $, $ \nu>0 $$ \widehat{D}=\{D(s):s\in {\mathbb R}\}\in {\mathcal D}_\sigma $, 存在$ M_{0, \epsilon}=M_{0, \epsilon}(t, \nu, \widehat{D})\in {\Bbb Z}_+ $$ \tau_{0, \epsilon}=\tau_{0, \epsilon}(t, \nu, \widehat{D})\leqslant t $使得

$ \begin{eqnarray} \sup\limits_{u_{\tau^+}\in D(\tau)}\sum\limits_{|m|\geqslant M_{0, \epsilon}} |(U_\epsilon(t, \tau)u_{\tau^+})_m|^2 \leqslant \nu^2, \; \forall \, \tau\leqslant \tau_{0, \epsilon} \end{eqnarray} $

成立, 则称过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $具有拉回$ {\mathcal D}_\sigma $ -渐近零性质.

(3) 称集合族$ \hat{{\cal A}}^\epsilon_{{\cal D}_\sigma} =\{{\cal A}^\epsilon_{{\cal D}_\sigma}(t):t\in{{\Bbb R}} \} \subseteq{\cal P}(\ell^2) $是过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant\tau} $的一个拉回$ {\cal D}_\sigma $ -吸引子, 若满足

(a) 紧性:对任意的$ t\in {{\Bbb R}} $, $ {\cal A}^\epsilon_{{\cal D}_\sigma}(t) $$ \ell^2 $上的一个非空紧集;

(b) 不变性: $ U_\epsilon(t, \tau){\cal A}^\epsilon_{{\cal D}_\sigma}(\tau) ={\cal A}^\epsilon_{{\cal D}_\sigma}(t) $, $ \forall \, \tau\leqslant t $;

(c) 拉回吸引性: $ \hat{{\cal A}}^\epsilon_{{\cal D}_\sigma} $满足

其中$ {\rm dist}_{\ell^2}(\cdot, \cdot) $表示$ \ell^2 $上的$ \rm Hausdorff $半距离.

引理3.1  假设$ \alpha $, $ \epsilon $, $ \kappa $, $ \beta $都是正的常数, $ 1<\gamma<+\infty $$ \rm ({H1})–({H2}) $成立, 则由(3.1)式定义的过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $$ \ell^2 $上存在有界拉回$ {\mathcal D}_\sigma $ -吸收集.

  记$ R_\sigma(t)>0 $使得

$ \begin{eqnarray} R^2_\sigma(t) = 1+ \frac{{\rm e}^{-\sigma t}}{\kappa} \int_{-\infty}^t{\rm e}^{\sigma \theta}\|g(\theta)\|^2{\mathrm d}\theta, \; t\in {{\Bbb R}} , \end{eqnarray} $

则依赖于时间$ t $的集合族$ \hat{{\mathcal B}}(t)=\{{\mathcal B}(t)={\mathcal B}(0, R_\sigma(t)): t\in {{\Bbb R}} \} $是过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $的有界拉回$ {\mathcal D}_\sigma $ -吸收集, 这里$ {\mathcal B}(0, R_\sigma(t)) $表示$ \ell^2 $中以原点为球心, 以$ R_\sigma(t) $为半径的闭球.事实上, 对任意给定的$ \widehat{D}=\{D(t):t\in {{\Bbb R}} \}\in{\cal D}_\sigma $$ u_{\tau^+}\in D(\tau) $, $ \tau\in {{\Bbb R}} $, 由(2.19)和(3.2)式知存在时间$ \tau_0=\tau_0(t, \widehat{D})<t $使得

$ \begin{eqnarray} \|U_\epsilon(t, \tau)u_{\tau^+}\|^2 &\leqslant& \|u_{\tau^+}\|^2{\rm e}^{-\sigma(t-\tau)} +\frac{{\rm e}^{-\sigma t}}{\kappa} \int_\tau^t{\rm e}^{\sigma \theta}\|g(\theta)\|^2{\mathrm d}\theta \\ &\leqslant& 1+\frac{{\rm e}^{-\sigma t}}{\kappa} \int_{-\infty}^t{\rm e}^{\sigma \theta}\|g(\theta)\|^2{\mathrm d}\theta, \; \forall\, \tau<\tau_0. \end{eqnarray} $

证明完毕.

引理3.2  假设$ \alpha $, $ \epsilon $, $ \kappa $, $ \beta $都是正的常数, $ 1<\gamma<+\infty $$ \rm ({H1})–({H2}) $成立, 则由(3.1)式定义的过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $$ \ell^2 $上满足拉回$ {\mathcal D}_\sigma $ -渐近零性质.

  选择一个光滑函数$ \chi(\cdot)\in C^1({\mathbb R}_+;{\mathbb R}_+) $使满足

$ \begin{eqnarray} \left\{ \begin{array}{ll} \chi(x)=0, & 0\leqslant x\leqslant 1, \\ 0\leqslant \chi(x)\leqslant 1, & 1\leqslant x\leqslant 2, \\ \chi(x)=1, & x\geqslant 2, \\ |\chi'(x)|\leqslant \chi_0 , & x\geqslant 0. \end{array} \right. \end{eqnarray} $

其中$ \chi_0 $是某个正数.考虑任意给定的$ \widehat{D}=\{D(t): t\in {{\Bbb R}} \}\in{\cal D}_\sigma $, $ t, \tau \in {{\Bbb R}} $$ t\geqslant \tau $, 记$ u^\epsilon(\cdot)=u^\epsilon(\cdot;\tau, u_{\tau^+})=U_\epsilon(t, \tau)u_{\tau^+} $为问题(2.3)–(2.5)在初始时刻$ \tau\in {{\Bbb R}} $$ u_{\tau^+} $为初值的解.设$ M\in {\Bbb Z}_+ $, 记$ v^\epsilon_m(\cdot)=\chi(\frac{|m|}{M})u^\epsilon_m(\cdot) $, $ m\in {\Bbb Z} $.$ v^\epsilon=(v^\epsilon_m)_{m\in {\Bbb Z}} $与方程(2.9)在$ \ell^2 $上作内积并取实部, 得

$ \begin{eqnarray} && {\bf{Re}} (\frac{{\mathrm d} u^\epsilon}{{\mathrm d} t}, v^\epsilon) +{\bf{Re}}({\rm i}\alpha+\epsilon)\sum\limits_{m\in {\Bbb Z}}(Bu^\epsilon)_m(B\overline{v^\epsilon})_m +\kappa\sum\limits_{m\in {\Bbb Z}}u^\epsilon_m \overline{v^\epsilon_m} {}\\ &=& {\bf{Im}}\sum\limits_{m\in {\Bbb Z}}g_m(t)\overline{v^\epsilon_m}, \; t>\tau, t\neq\tau_j, j\in {\Bbb Z}. \end{eqnarray} $

通过简单计算, 有

$ \begin{eqnarray} \left\{ \begin{array}{ll} {\bf{Re}}(\frac{{\mathrm d} u^\epsilon}{{\mathrm d} t}, v^\epsilon) = \frac12\frac{{\mathrm d}}{{\mathrm d} t} \sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|u^\epsilon_m|^2, \\ { } \kappa\sum\limits_{m\in {\Bbb Z}}u^\epsilon_m \overline{v^\epsilon_m} = \kappa \sum\limits_{m\in {\Bbb Z}} \chi(\frac{|m|}{M})|u^\epsilon_m|^2, \\ \Big|{\bf{Im}} \sum\limits_{m\in {\Bbb Z}} \chi(\frac{|m|}{M})g_m(t) \overline{u^\epsilon_m}\Big| \leqslant \frac{\kappa}{2}\sum\limits_{m\in {\Bbb Z}} \chi(\frac{|m|}{M})|u^\epsilon_m|^2 +\frac{1}{2\kappa}\sum\limits_{m\in {\Bbb Z}} \chi(\frac{|m|}{M})|g_m(t)|^2. \end{array} \right. \end{eqnarray} $

同时, 由于

我们有

$ \begin{eqnarray} {\bf{Re}}\; ({\rm i}\alpha+\epsilon) \sum\limits_{m\in {\Bbb Z}}(Bu^\epsilon)_m(B\bar{v}^\epsilon)_m &\geqslant& -\alpha{\bf{Im}}\sum\limits_{m\in {\Bbb Z}} \Big( \chi(\frac{|m+1|}{M})\bar{u}^\epsilon_{m+1}u^\epsilon_m +\frac{|m|}{M}u^\epsilon_{m+1}\bar{u}^\epsilon_{m} \Big)\\ &=& -\alpha{\bf{Im}}\sum\limits_{m\in {\Bbb Z}} \Big( \chi(\frac{|m+1|}{M})-\frac{|m|}{M}\Big) \bar{u}^\epsilon_{m+1}u^\epsilon_m \\ &=& -\alpha{\bf{Im}}\sum\limits_{m\in {\Bbb Z}} \chi'(\frac{\tilde{m}}{M})\frac{1}{M} \bar{u}^\epsilon_{m+1}u^\epsilon_m\\ &\geqslant& -\frac{\alpha\chi_0}{M}R^2_\sigma(t), \; \forall\, \tau\leqslant\tau_0, \end{eqnarray} $

其中$ \tilde{m} $是取值于$ |m| $$ |m+1| $之间的常数.结合(3.7), (3.8)和(3.9)式, 得到

$ \begin{equation} \frac{{\mathrm d}}{{\mathrm d} t} \sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|u^\epsilon_m(t)|^2 +\kappa\sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|u^\epsilon_m(t)|^2 \leqslant \frac{1}{\kappa}\sum\limits_{m\in {\Bbb Z}} \chi(\frac{|m|}{M})|g_m(t)|^2 +\frac{2\chi_0}{M}R^2_\sigma(t), \; t\neq\tau_j, \end{equation} $

其中$ t>\tau_0>\tau $.对于任意的$ \nu>0 $, 由(3.4)式和$ \rm (H2) $知存在常数$ M_1=M_1(t, \nu)\in {\Bbb Z}_+ $, 使得

$ \begin{equation} \frac{4\chi_0}{M}R^2_\sigma(t) \leqslant \sigma\nu^2/3, \; \forall\, M\geqslant M_1. \end{equation} $

$ y(t)=\sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|u^\epsilon_m(t)|^2 $, $ M\geqslant M_1 $, 直接计算可得

$ \begin{eqnarray} y(\tau_j^+) &=& \sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|u^\epsilon_m(\tau_j^+)|^2 =\sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})\big|u^\epsilon_m(\tau_j)+\phi_{mj}(u^\epsilon_m(\tau_j))\big|^2 \\ &\leqslant& 2\sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|u^\epsilon_m(\tau_j)|^2 +\sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|\phi_{mj}(u^\epsilon_m(\tau_j))|^2\\ &\leqslant& (2+2L^2)\sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|u^\epsilon_m(\tau_j)|^2 =(2+2L^2)y(\tau_j), \; j\in {\mathbb Z}. \end{eqnarray} $

注意到

$ \begin{eqnarray} y(\tau^+) = \sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|u_m(\tau^+)|^2 \leqslant \|u(\tau^+)\|^2, \end{eqnarray} $

由引理2.3, (3.10)–(3.13)式和(2.18)式就得到

$ \begin{eqnarray} \sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|u^\epsilon_m(t)|^2 &\leqslant& \|u_{\tau^+}\|^2(2+2L^2)^{n(\tau, t)}{\rm e}^{-\kappa(t-\tau)} {}\\ && + \int_\tau^t(2+2L^2)^{n[s, t)} {\rm e}^{-\kappa(t-s)} \bigg(\frac{1}{\kappa}\sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|g_m(s)|^2 +\frac{\sigma\nu^2}{3}\bigg){\mathrm d}s\\ &\leqslant& \|u_{\tau^+}\|^2{\rm e}^{-\sigma(t-\tau)} +\frac{{\rm e}^{-\sigma t}}{\kappa}\int_\tau^t{\rm e}^{\sigma s} \sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|g_m(s)|^2{\mathrm d}s {}\\ && +\frac{\nu^2}{3}, \; \, \forall\, t>\tau_1> \tau, M\geqslant M_1. \end{eqnarray} $

又由{(H2)}知存在常数$ M_2=M_2(\nu, t)\in {\Bbb Z}_+ $, 使得

$ \begin{equation} \frac{{\rm e}^{-\sigma t}}{\kappa}\int_\tau^t{\rm e}^{\sigma s} \sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|g_m(s)|^2{\mathrm d}s \leqslant \frac{{\rm e}^{-\sigma t}}{\kappa}\int_{-\infty}^t{\rm e}^{\sigma s} \sum\limits_{m\in {\Bbb Z}}\chi(\frac{|m|}{M})|g_m(s)|^2{\mathrm d}s \leqslant \frac{\nu^2}{3}, \; \forall\, M\geqslant M_2. \end{equation} $

同时, 因为$ u_{\tau^+}\in D(\tau) $$ \hat{D}=\{D(t): t\in {{\Bbb R}} \}\in {\mathcal D}_\sigma $, 所以由(3.2)式知, 对上述$ \nu>0 $, 存在$ \tau_2=\tau_2(t, \nu, \hat{D})<t $使得

$ \begin{equation} \|u_{\tau^+}\|^2{\rm e}^{-\sigma(t-\tau)} \leqslant \frac{\nu^2}{3}, \; \forall\, \tau\leqslant \tau_2. \end{equation} $

由(3.14)–(3.16)式推得

证明完毕.

定理3.1  假设$ \alpha $, $ \epsilon $, $ \kappa $, $ \beta $都是正的常数, $ 1<\gamma<+\infty $$ \rm ({H1})–({H2}) $成立.则对任意的$ \epsilon>0 $, 由(3.1)式定义的过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $存在满足定义3.1(3)的拉回$ {\mathcal D}_\sigma $ -吸引子.

  因为对任意的$ \epsilon>0 $, 过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $$ \ell^2 $上连续.定理3.1的结论可以由引理3.1, 引理3.2和文献[43, 定理2.1]直接得到.

4 不变测度与统计解的构造

在这一节中, 我们证明对任意给定的$ t \in {{\Bbb R}} $$ u_*\in\ell^2 $, 映射$ \tau\mapsto U_\epsilon(t, \tau)u_* $属于空间$ PC((-\infty, t]; \ell^2) $; 且对任意的$ \epsilon>0 $, 这个$ \ell^2 $ -值映射是有界的.接着我们改进[32, 定理3.1]中的定义, 通过广义Banach极限和定理3.1中得到的拉回吸引子$ \hat{{\mathcal A}^\epsilon}(t) $, 构造过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $的一族Borel不变概率测度$ \{{{\mathit{ m}}}^\epsilon_t\}_{t\in {{\Bbb R}} } $, 并且证明这族Borel概率不变测度$ \{{{\mathit{ m}}}^\epsilon_t\}_{t\in {{\Bbb R}} } $只分段地满足Liouville型定理, 并且是脉冲离散Ginzburg-Landau方程组的统计解.

文后我们用$ c_1\lesssim c_2 $来表示存在一个常数$ b>0 $, 使得$ c_1\leqslant bc_2 $, 其中$ b $只依赖于问题中的系数.此外, $ \int_{\ell^2}\psi(u){\mathrm d}\rho_t(u) $表示Bochner积分, 其中$ \rho_t $$ \ell^2 $上的Borel概率测度, $ \psi\in C(\ell^2) $ ($ \ell^2 $上的连续实值函数全体).

引理4.1  假设$ \alpha $, $ \epsilon $, $ \kappa $, $ \beta $都是正的常数, $ 1<\gamma<+\infty $$ \rm ({H1})–({H2}) $成立.则对任意的$ \epsilon>0 $, $ t_*\in {{\Bbb R}} $$ u_*\in\ell^2 $, 映射$ \tau\mapsto U_\epsilon(t_*, \tau)u_* $$ (-\infty, t_*] $上是有界的.

  引理4.1的结论可由定理2.1和假设$ \rm ({H2}) $直接得到.事实上, 对任意给定的$ t_*\in {{\Bbb R}} $$ u_*\in\ell^2 $, 有

$ \begin{eqnarray} \|u^\epsilon(t_*;\tau, u_*)\|^2 \leqslant \|u_*\|^2 +\frac{{\rm e}^{-\sigma t_*}}{\kappa} \int_{-\infty}^{t_*}{\rm e}^{\sigma \theta}\|g(\theta)\|^2{\mathrm d}\theta, \; \forall\, t_*>\tau. \end{eqnarray} $

由假设$ \rm ({H2}) $知上面不等式的右边是与$ \tau $无关的有界量.证明完毕.

引理4.2  假设$ \alpha $, $ \epsilon $, $ \kappa $, $ \beta $都是正的常数, $ 1<\gamma<+\infty $$ \rm ({H1})–({H2}) $成立.考虑给定的$ \tau_*\in {{\Bbb R}} $$ u_*\in\ell^2 $, 则对于任意的$ \epsilon>0 $, 存在一个充分小的正数$ \delta=\delta(\nu, \tau_*, u_*) $, 使得

$ \begin{eqnarray} \|U_\epsilon(s, \tau)u_*-u_*\|<\nu, \; \forall\, \tau\in (\tau_*, \tau_*+\delta), \; \forall\, s\in (\tau, \tau_*+\delta). \end{eqnarray} $

  给定$ \tau_*\in {{\Bbb R}} $$ \tau_*\in \ell^2 $.不失一般性, 假设存在某个$ j_0\in {\Bbb Z} $使得$ \tau_*\in (\tau_{j_0}, \tau_{j_0+1}] $.我们分两种情形证明.

情形1  $ \tau_*\in (\tau_{j_0}, \tau_{j_0+1}) $.此时记$ 2d=\min\{\tau_*-\tau_{j_0}, \tau_{j_0+1}-\tau_*\}>0 $, 考虑$ \tau_*< \tau\leqslant s\leqslant \tau_*+d $.我们首先证明

$ \begin{eqnarray} \int_\tau^s\bigg\|\frac{{\mathrm d} U_\epsilon(\theta, \tau)u_*}{{\mathrm d}\theta}\bigg\|^2{\mathrm d}\theta \lesssim c_* = \|u_*\|^2 +\int_{\tau_*-d}^{\tau_*+d}\|g(\theta)\|^2{\mathrm d}\theta. \end{eqnarray} $

事实上, 从(2.2)和(2.3)式可知对任意的$ \epsilon\in (0, 1) $, $ \tau\leqslant \theta\leqslant s $, 有

$ \begin{eqnarray} \bigg \|\frac{{\mathrm d}U_\epsilon(\theta, \tau)u_*}{{\mathrm d}\theta}\bigg\|^2 \lesssim \|AU_\epsilon(\theta, \tau)u_*\|^2+\|U_\epsilon(\theta, \tau)u_*\|^2 +\|\tilde{f}(U_\epsilon(\theta, \tau)u_*)\|^2+\|g(\theta)\|^2. \end{eqnarray} $

由(2.11)和(3.5)式可得

$ \begin{eqnarray} \|\tilde{f}(U_\epsilon(\theta, \tau)u_*)\|^2 \leqslant M^*_f\|U_\epsilon(\theta, \tau)u_*\|^2, \end{eqnarray} $

其中$ M^*_f $是只依赖于$ L_* $($ \tau $$ s $无关)的一个常数, 这里

对(4.4)式关于$ \theta $$ [\tau, s] $上积分, 并应用(4.5)式, 可得

$ \begin{eqnarray} \int_\tau^s\bigg\|\frac{{\mathrm d} U_\epsilon(\theta, \tau)u_*}{{\mathrm d}\theta}\bigg\|^2{\mathrm d}\theta &\lesssim& (1+M^*_f)\int_\tau^s\|U_\epsilon(\theta, \tau)u_*\|^2{\mathrm d}\theta +\int_\tau^s\|g(\theta)\|^2{\mathrm d}\theta\\ &\lesssim& \|u_*\|^2 +\int_\tau^s\|g(\theta)\|^2{\mathrm d}\theta \lesssim \|u_*\|^2 +\int_{\tau_*-d}^{\tau_*+d}\|g(\theta)\|^2{\mathrm d}\theta. \end{eqnarray} $

注意到

$ \begin{eqnarray} \|U_\epsilon(s, \tau)u_*-u_*\|^2 &=& \big(U_\epsilon(s, \tau)u_*-u_*, U_\epsilon(s, \tau)u_*-u_*\big) \\ &=& \|U_\epsilon(s, \tau)u_*\|^2-\|u_*\|^2-2{\bf{Re}}\, \big(U_\epsilon(s, \tau)u_*-u_*, u_*\big) \\ &=& \int_\tau^s\frac{{\mathrm d} \|U_\epsilon(\theta, \tau)u_*\|^2}{{\mathrm d}\theta}{\mathrm d}\theta -2{\bf{Re}}\, \big(U_\epsilon(s, \tau)u_*-u_*, u_*\big). \end{eqnarray} $

一方面, 由(2.15)式知

$ \begin{eqnarray} \int_{\tau}^{s}\frac{{\mathrm d} \|U_\epsilon(\theta, \tau)u_*\|^2}{{\mathrm d}\theta}{\mathrm d}\theta \lesssim \int_{\tau}^{s}\|g(\theta)\|^2{\mathrm d}\theta. \end{eqnarray} $

由假设(H2)和$ g(\cdot)\in L^2_{\rm loc}({{\Bbb R}} ; \ell^2) $知, 对任意的$ \epsilon>0 $, 存在$ \delta'=\delta'(\nu, \tau_*, g)\in (0, d) $使得

$ \begin{eqnarray} \int_\tau^s\frac{{\mathrm d} \|U_\epsilon(\theta, \tau)u_*\|^2}{{\mathrm d}\theta}{\mathrm d}\theta \lesssim \int_{\tau}^{s}\|g(\theta)\|^2{\mathrm d}\theta < \frac{\nu^2}{2}, \; \tau_*<\tau\leqslant s\leqslant \tau_*+\delta'. \end{eqnarray} $

另一方面, 从(4.3)式可知常数$ c_* $$ \tau $$ s $无关.由Cauchy不等式和(4.3)式得

$ \begin{eqnarray} \Big|\big(U_\epsilon(\theta, \tau)u_*-u_*, u_*\big)\Big| &=& \bigg|\bigg(\int_\tau^s \frac{{\mathrm d}U_\epsilon(\theta, \tau)u_*}{{\mathrm d}\theta}{\mathrm d}\theta, u_*\bigg)\bigg| \leqslant \|u_*\|\int_\tau^s\bigg\| \frac{{\mathrm d}U_\epsilon(\theta, \tau)u_*} {{\mathrm d}\theta}\bigg\|{\mathrm d}\theta \\ &\leqslant& \|u_*\|\bigg(\int_\tau^s\| \frac{{\mathrm d}U_\epsilon(\theta, \tau)u_*} {{\mathrm d}\theta}\|^2{\mathrm d}\theta \bigg)^{1/2}(s-\tau)^{1/2}{}\\ &\leqslant& c_*^{1/2}\|u_*\|(s-\tau)^{1/2}. \end{eqnarray} $

(4.10)式表明存在$ \delta''=\delta''(\nu, \tau_*, u_*)\in (0, d) $使得

$ \begin{eqnarray} \Big|\big(U_\epsilon(\theta, \tau)u_*-u_*, u_*\big)\Big| < \frac{\nu^2}{4}, \; \tau_*<\tau\leqslant s\leqslant \tau_*+\delta''. \end{eqnarray} $

$ \delta=\min\{\delta', \delta''\} $, 则由(4.8), (4.10)和(4.11)式就得到(4.2)式. $ \tau_*\in (\tau_{j_0}, \tau_{j_0+1}) $的情形证明完毕.

情形2  $ \tau_*=\tau_{j_0+1} $.此时记$ 2d=\tau_{j_0+2}-\tau_{j_0+1} $, 考虑$ \tau_*<\tau\leqslant s\leqslant \tau_*+d $.与情形$ 1 $证明的主要区别在于常数$ c_* $$ L_* $分别替换为

此时$ c'_* $$ L'_* $也都是不依赖于$ \tau $$ s $的常数.其余的证明与情形$ 1 $的证明类似, 略去.

与引理4.2相似, 我们有

引理4.3  假设$ \alpha $, $ \epsilon $, $ \kappa $, $ \beta $都是正的常数, $ 1<\gamma<+\infty $$ \rm ({H1})–({H2}) $成立.考虑给定的$ \tau_*\in {{\Bbb R}} $$ u_*\in\ell^2 $, 则对任意的$ \nu>0 $, 存在一个充分小的正数$ \delta=\delta(\nu, \tau_*, u_*) $, 使得对任意的$ \epsilon\in (0, 1) $

$ \begin{eqnarray} \|U_\epsilon(s, \tau)u_*-u_*\|<\nu, \;\forall\, \tau\in (\tau_*-\delta, \tau_*], \; \forall\, s\in [\tau, \tau_*]. \end{eqnarray} $

引理4.4  假设$ \alpha $, $ \kappa $, $ \beta $都是正的常数, $ \epsilon\in(0, 1) $, $ \gamma\in(1, +\infty) $$ \rm ({H1})–({H2}) $成立.则对任意给定的$ t_*\in {{\Bbb R}} $$ u_*\in\ell^2 $, $ \ell^2 $ -值映射$ \tau\mapsto U_\epsilon(t_*, \tau)u_* $属于$ PC((-\infty, t_*]; \ell^2) $, 也即

(1) $ U_\epsilon(t_*, \tau)u_* $$ \tau\in (-\infty, t_*] $上左连续$ \rm ; $

(2) $ U_\epsilon(t_*, \tau)u_* $$ \tau\in (-\infty, t_*] $, $ \tau\neq\tau_j\in (-\infty, t_*] $, $ j\in {\Bbb Z} $处连续$ \rm ; $

(3) $ U_\epsilon(t_*, \tau)u_* $在脉冲点$ \tau_j\in (-\infty, t_*] $, $ j\in {\Bbb Z} $处右极限存在.

  首先证明(1).任意给定$ s_*\in (-\infty, t_*] $.我们证明$ U_\epsilon(t_*, \tau)u_* $$ \tau=s_* $处左连续.事实上, 不失一般性, 假设对于某个$ j_0\in {\Bbb Z} $$ s_*\in (\tau_{j_0}, \tau_{j_0+1}] $.对任意的$ \tau\in (\tau_{j_0}, s_*] $, 由过程的不变性质得

$ \begin{eqnarray} \|U_\epsilon(t_*, \tau)u_*-U_\epsilon(t_*, s_*)u_*\| = \|U_\epsilon(t_*, s_*)U(s_*, \tau)u_*-U_\epsilon(t_*, s_*)u_*\|. \end{eqnarray} $

注意到$ t_* $$ s_* $给定, $ U_\epsilon(t_*, s_*): \ell^2\mapsto \ell^2 $是连续映射.故$ U_\epsilon(t_*, \tau)u_* $在点$ \tau=s_* $处的连续性可由(4.12)和(4.13)式得到.

接着证明(2).只需证明在(1)结论下, 对于$ j_0\in {\Bbb Z} $, $ U_\epsilon(t_*, \tau)u_* $在区间$ (\tau_{j_0}, \tau_{j_0+1})\cap(-\infty, t_*] $上是右连续的.给定$ s_*\in (\tau_{j_0}, \tau_{j_0+1})\cap(-\infty, t_*] $$ s_*<\tau<\tau_{j_0+1}\leqslant t_* $, 应用过程的不变性质和(2.20)式, 得

$ \begin{eqnarray} \|U_\epsilon(t_*, s_*)u_*-U_\epsilon(t_*, \tau)u_*\| &=& \|U_\epsilon(t_*, \tau)U(\tau, s_*)u_*-U_\epsilon(t_*, \tau)u_*\| \\ &\leqslant& \|U_\epsilon(\tau, s_*)u_*-u_*\|{\rm e}^{-(\sigma+\kappa-2\ell(u_*))(t_*-\tau)}. \end{eqnarray} $

由(4.14)式和$ U_\epsilon(\cdot, s_*)u_*=u^\epsilon(\cdot)\in C((\tau_{j_0}, \tau_{j_0+1}); \ell^2) $就得到过程$ U_\epsilon(t_*, \tau)u_* $在点$ \tau=s_* $处的右连续性.

最后证明(3).事实上, 应用Cauchy收敛准则, 引理4.2和过程的不变性质得到就可证明$ U_\epsilon(t_*, \tau)u_* $在脉冲点$ \tau_j\in (-\infty, t_*] $处有右极限.证明完毕.

结合引理4.1和引理4.4可知, 对任意的$ \epsilon\in (0, 1) $, 过程$ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $具有下面的PC-$ \tau $ -连续性:对任意给定的$ t_*\in {{\Bbb R}} $$ u_*\in\ell^2 $, $ \ell^2 $ -值映射$ \tau\mapsto U_\epsilon(t_*, \tau)u_*\in PC((-\infty, t_*]; \ell^2) $且在区间$ (-\infty, t_*] $上有界.

定义4.1[20, 32]  记$ B_+ $$ [0, +\infty) $上所有有界实值函数全体构成的集合.广义$ \rm Banach $极限(记作$ \lim\limits_{t\rightarrow +\infty} $)是指定义在$ B_+ $上的线性泛函, 满足

(1) 在$ [0, +\infty) $上, 非负函数$ \zeta(\cdot) $$ \lim\limits_{t\rightarrow +\infty}\zeta(t)\geqslant 0 $;

(2) 若通常的极限$ \lim\limits_{t\rightarrow +\infty}\zeta(t) $存在, 则$ \lim\limits_{t\rightarrow +\infty}\zeta(t) =\lim\limits_{t\rightarrow +\infty}\zeta(t) $.

对于任意的广义Banach极限$ \lim\limits_{t\rightarrow +\infty} $, 下面的性质

$ \begin{eqnarray} |\lim\limits_{t\rightarrow +\infty}\zeta(t)| \leqslant \limsup\limits_{t\rightarrow +\infty}|\zeta(t)|, \; \forall\, \zeta(\cdot)\in B_+, \end{eqnarray} $

在文献[20, (1.38)式]和[12, (2.3)式]中已经得到证明.

注4.1  由于研究"拉回"渐近行为, 我们需考虑$ \tau\rightarrow -\infty $时的广义Banach极限.对于一个定义在$ (-\infty, 0] $上的实值函数$ \zeta $和一个给定的广义$ \rm Banach $极限$ \lim\limits_{t\rightarrow +\infty} $, 我们定义

$ \begin{eqnarray} \lim\limits_{t\rightarrow -\infty}\zeta(t) =\lim\limits_{t\rightarrow +\infty}\zeta(-t). \end{eqnarray} $

定理4.1  假设$ \alpha $, $ \kappa $, $ \beta $都是正的常数, $ \epsilon\in(0, 1) $, $ \gamma\in(1, +\infty) $$ \rm ({H1})–({H2}) $成立.设$ v(\cdot):{\mathbb R}\mapsto \ell^2 $是一个连续映射, 满足$ v(\cdot)\in {\mathcal D}_\sigma $.则对于一个给定的广义$ \rm Banach $极限$ \lim\limits_{t\rightarrow +\infty} $, 在$ \ell^2 $上存在唯一的一族$ \rm Borel $概率测度$ \{{\textsf m}^\epsilon_t\}_{t\in {{\Bbb R}} } $使得$ {\textsf m}^\epsilon_t $的支集包含在$ {\mathcal A}^\epsilon(t) $中, 且

$ \begin{eqnarray} && \lim\limits_{\tau\rightarrow -\infty}\frac{1}{t-\tau}\int_\tau^t \psi\big(U_\epsilon(t, s)v(s)\big){\mathrm d}s {}\\ &=& \int_{{\mathcal A}^\epsilon(t)}\psi(u){\mathrm d}{\textsf m}^\epsilon_t(u) = \int_{\ell^2}\psi(u){\mathrm d}{\textsf m}^\epsilon_t(u) \end{eqnarray} $

$ \begin{eqnarray} &=& \lim\limits_{\tau\rightarrow -\infty}\frac{1}{t-\tau}\int_\tau^t\int_{\ell^2} \psi\big(U_\epsilon(t, s)u\big){\mathrm d}{\textsf m}^\epsilon_s(u){\mathrm d}s, \end{eqnarray} $

对任意的$ \psi\in C(\ell^2) $成立.同时, $ {\textsf m}^\epsilon_t $在下述意义下具有不变性:

$ \begin{eqnarray} \int_{{\mathcal A}^\epsilon(t)}\psi(u){\mathrm d}{\textsf m}^\epsilon_t(u) = \int_{{\mathcal A}^\epsilon(\tau)}\psi\big(U_\epsilon(t, \tau)u\big){\mathrm d}{\textsf m}^\epsilon_\tau(u), \; t\geqslant \tau. \end{eqnarray} $

  证明的思想与[32, 定理3.1]类似.主要的区别是将过程$ \{U_\varepsilon(t, \tau)\}_{t\geqslant \tau} $$ \tau $ -连续性替换为PC-$ \tau $ -连续性.显然, $ \tau $ -连续性包含PC-$ \tau $ -连续性.固定$ \psi(\cdot)\in C(\ell^2) $和连续映射$ v(\cdot): {{\Bbb R}} \mapsto \ell^2 $使得$ v(\cdot)\in {\mathcal D}_\sigma $.对任意的$ \epsilon\in (0, 1) $和给定的$ t\in {{\Bbb R}} $, 我们断言在任意闭区间$ [t_0, t] $上, 映射$ s\mapsto \psi(U_\epsilon(t, s)v(s)) $$ [t_0, t] $上有界, 其中$ t_0<t $.事实上, 一方面, 从(1.4)式可知落在区间$ [t_0, t] $上的脉冲点$ \{\tau_j\}_{j\in {\Bbb Z}} $的个数是有限的.我们用$ \tau_{j_0+1} $, $ \tau_{j_0+2} $, $ \cdots $, $ \tau_{j_0+N} $来表示这些脉冲点, 其中$ N\in {\Bbb Z}_+ $.由引理4.4知函数$ s\mapsto \psi(U_\epsilon(t, s)v(s)) $$ [t_0, t]\backslash \{\tau_{j_0+1}, \tau_{j_0+2}, \cdots, \tau_{j_0+N}\} $上连续, 在$ (t_0, t] $上左连续, 且在点$ t_0 $, $ \tau_{j_0+1} $, $ \cdots $, $ \tau_{j_0+N} $处存在右极限.因此, $ \psi(U_\epsilon(t, s)v(s)) $在紧集$ [t_0, t] $上有界.另一方面, 由引理3.1可知$ \psi(U_\epsilon(t, s)v(s)) $在区间$ (-\infty, t_0+1] $上有界.由于$ v(\cdot)\in {\mathcal D}_\sigma $, $ \{U_\epsilon(t, \tau)\}_{t\geqslant \tau} $有拉回$ {\mathcal D}_\sigma $ -吸收性.因此, 函数$ \psi(U_\epsilon(t, s)v(s)) $$ (-\infty, t] $上有界, 且映射$ \tau\longmapsto \frac{1}{t-\tau}\int_\tau^t\psi(U_\epsilon(t, s)v(s)){\mathrm d}s $$ (-\infty, t] $上有界.据此, 我们定义

剩余的证明参考文献[32, 定理3.1], 细节在此略去.

下面研究方程(2.3)的统计解.首先介绍与统计解相关的试验函数的类$ {\mathcal T} $的定义.为此, 将方程(2.3)写作如下形式

$ \begin{eqnarray} \frac{{\mathrm d} u^\epsilon}{{\mathrm d}t}= F^\epsilon(u^\epsilon, t)=-{\rm i}g(t)-(\epsilon-{\rm i}\alpha) Au^\epsilon-\kappa u^\epsilon+{\rm i}\beta\tilde{f}(u^\epsilon), \; \, t\neq \tau_j, \, j\in {\Bbb Z}. \end{eqnarray} $

不难看出$ F(u^\epsilon, t):\ell^2\times {{\Bbb R}} \longmapsto \ell^2 $.我们期望试验函数$ \Phi\in {\mathcal T} $满足

$ \begin{eqnarray} \frac{{\mathrm d}}{{\mathrm d}t}\Phi(u^\epsilon(t))=(\Phi'(u^\epsilon), F(u^\epsilon, t)), \; \, t\neq \tau_j, \, j\in {\Bbb Z}, \end{eqnarray} $

其中$ u^\epsilon(t) $, $ \epsilon\in (0, 1) $是方程组(2.3)–(2.5)的解.

定义4.2[20, 45]  试验函数类$ {\mathcal T} $是定义在$ \ell^2 $上的实值泛函$ \Phi=\Phi(\cdot) $的集合, 这些泛函在$ \ell^2 $的有界子集上有界, 且满足

(a) 对任意的$ u\in \ell^2 $, $ \rm Fréchet $导数$ \Phi'(u) $存在, 即:对任意的$ u\in \ell^2 $, 存在$ \Phi'(u) $使得

(b) 对任意的$ u\in \ell^2 $, 有$ \Phi'(u)\in \ell^2 $, 且作为$ \ell^2 $$ \ell^2 $的映射, $ u\longmapsto \Phi'(u) $是连续且有界的$ \rm ; $

(c) 方程(4.20)的每一个全局解$ u^\epsilon(t) $都满足(4.21)式.

我们可以考虑文献[20, p178]中定义在$ \ell^2 $上的试验函数.考虑给定的$ \varphi\in \ell^2 $, $ \gamma $$ {\mathbb C} $上具有紧支集的连续可微的实值函数.对任意的$ u\in \ell^2 $, 定义

显然$ \Phi(\cdot) $是从$ \ell^2 $$ {\mathbb R} $上的连续函数, 并且函数$ \Phi(\cdot) $还在$ \ell^2 $上Fréchet可导:

$ \begin{eqnarray} \Phi'(u) = \gamma'((\varphi, u))\varphi\in \ell^2, \end{eqnarray} $

上面的分析说明满足定义4.2的试验函数是存在的.

下面介绍方程(4.20)统计解的定义, 并证明它的存在性.

定义4.3  对任意的$ \epsilon\in (0, 1) $, $ \ell^2 $上的一族$ \rm Borel $概率测度$ \{\rho^\epsilon_t\}_{t\in {\mathbb R}} $是方程(4.20)的统计解, 若满足下面条件:

(a) 对任意的$ \psi\in C(\ell^2) $, 映射$ t\mapsto \int_{\ell^2}\psi(u^\epsilon) {\mathrm d}\rho^\epsilon_t(u^\epsilon)\in PC({{\Bbb R}} ;{{\Bbb R}} ) $;

(b) 对任意的$ w\in \ell^2 $, 映射$ u^\epsilon\mapsto \big(w, F^\epsilon(u^\epsilon, t)\big) $对于几乎所有的$ t\in {\mathbb R} $$ \rho^\epsilon_t $ -可积的, 且对任意$ w\in \ell^2 $, 映射$ t\mapsto \int_{\ell^2}\big(w, F^\epsilon(u^\epsilon, t)\big) {\mathrm d}\rho^\epsilon_t(u^\epsilon) $属于$ L^1_{\rm loc}({\mathbb R}) $$ \rm ; $

(c) 对任意的试验函数$ \Phi\in {\mathcal T} $, 有

其中$ t, \tau\in (\tau_j, \tau_{j+1}) $, $ t\geqslant \tau $, $ j\in {\Bbb Z} $.

本文的主要结果如下:

定理4.2  假设$ \alpha $, $ \kappa $, $ \beta $都是正的常数, $ \epsilon\in(0, 1) $, $ \gamma\in(1, +\infty) $, 且$ \rm ({H1})–({H2}) $成立.则对任意的$ \epsilon\in (0, 1) $, 定理$ \rm {4.1} $中得到的$ \rm Borel $概率测度$ \{{\textsf m}^\epsilon_t\}_{t\in {{\Bbb R}} } $是方程(4.20)的统计解.

  我们证明对任意的$ \epsilon\in (0, 1) $, 由定理4.1得到的Borel概率测度$ \{m^\epsilon_t\}_{t\in {{\Bbb R}} } $满足定义4.3(a)–(c).

首先证明(a).考虑任意的$ j\in {\Bbb Z} $和给定的$ t_*\in (\tau_j, \tau_{j+1}] $.$ t_*\in (\tau_j, \tau_{j+1}) $, 我们证明对任意的$ \psi\in C(\ell^2) $, 有

$ \begin{eqnarray} \lim\limits_{t\rightarrow t^*} \int_{\ell^2} \psi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_t(u^\epsilon) = \int_{\ell^2} \psi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_{t_*}(u^\epsilon). \end{eqnarray} $

事实上, 由(4.17)和(4.19)式知对于$ t>t_* $

$ \begin{equation} \int_{\ell^2} \psi(u^\epsilon){\mathrm d}{\it {\mathit{ m}}}^\epsilon_t(u^\epsilon) - \int_{\ell^2} \psi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_{t_*}(u^\epsilon) = \int_{{\mathcal A}^\epsilon(t_*)} \big(\psi(U_\epsilon(t, t_*)u^\epsilon)-\psi(u^\epsilon)\big){\mathrm d}{\mathit{ m}}^\epsilon_{t_*}(u^\epsilon). \end{equation} $

由于当$ t\rightarrow t_*^+ $时, $ U(t, t_*)u^\epsilon\longrightarrow u^\epsilon $$ \ell^2 $中强收敛, 且$ {{\mathcal A}^\epsilon(t_*)} $$ \ell^2 $中紧集, (4.24)式表明

类似地, 对于$ t<t_* $, 有

$ \begin{equation} \int_{\ell^2} \psi(u^\epsilon){\mathrm d}{\it {\mathit{ m}}}^\epsilon_{t_*}(u^\epsilon) - \int_{\ell^2} \psi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_{t}(u^\epsilon) = \int_{{\mathcal A}^\epsilon(t)} \big(\psi(U_\epsilon(t_*, t)u^\epsilon)-\psi(u^\epsilon)\big){\mathrm d}{\mathit{ m}}^\epsilon_{t}(u^\epsilon). \end{equation} $

由于当$ t\rightarrow t_*^- $$ U(t_*, t)u^\epsilon\longrightarrow u^\epsilon $$ \ell^2 $中是强收敛(见引理4.3), 且对任意的$ t\in {{\Bbb R}} $$ {\mathit{ m}}^\epsilon_{t}({{\mathcal A}^\epsilon(t)})\leqslant 1 $, 所以(4.25)式表明

$ \begin{eqnarray} \lim\limits_{t\rightarrow t_*^-}\int_{\ell^2} \psi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_t(u^\epsilon) = \int_{\ell^2} \psi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_{t_*}(u). \end{eqnarray} $

$ t_*=\tau_{j+1} $, 我们应用与(4.26)式相同的证明得到$ \int_{\ell^2} \psi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_t(u^\epsilon) $在点$ t_* $处的左连续性.为了证明$ \lim\limits_{t\rightarrow t_*^+} \int_{\ell^2}\psi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_t(u^\epsilon) $的存在性, 我们考虑$ t_*<t'\leqslant t''<\tau_{j+2} $, 此时有

$ \begin{equation} \int_{\ell^2}\psi(u^\epsilon){\mathrm d}{\it {\mathit{ m}}}^\epsilon_{t''}(u^\epsilon) - \int_{\ell^2} \psi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_{t'}(u^\epsilon) = \int_{{\mathcal A}^\epsilon(t')} \big(\psi(U_\epsilon(t'', t')u^\epsilon)-\psi(u^\epsilon)\big){\mathrm d}{\mathit{ m}}^\epsilon_{t'}(u_\epsilon). \end{equation} $

由于$ \psi\in C(\ell^2) $, $ {\mathit{ m}}^\epsilon_{t'}({{\mathcal A}^\epsilon(t')})\leqslant 1 $, (4.27)式和引理4.2表明$ \lim\limits_{t\rightarrow t_*^+}\int_{\ell^2}\psi(u^\epsilon) {\mathrm d}{\mathit{ m}}^\epsilon_t(u^\epsilon) $存在. (a)得证.

接着证明(b).对于任意的$ t\in {\mathbb R} $我们已经证明了$ {m}^\epsilon_t $的支集包含于$ {\mathcal A}^\epsilon(t)\subset \ell^2 $中.现对于任意的$ w\in \ell^2 $, 定义

$ \begin{eqnarray} \Psi(u)=\big(w, F(u, \cdot)\big), \quad u\in \ell^2, \end{eqnarray} $

$ \Psi(\cdot): \ell^2\longmapsto {\mathbb R} $.下面证明$ \Psi(\cdot)\in C(\ell^2) $.$ u_*\in \ell^2 $给定, 考虑$ u\in \ell^2 $, 其中$ \|u_*-u\|\leqslant 1 $.由(2.2), (2.11)和(4.20)式得到

$ \begin{eqnarray} |\Psi(u_*)-\Psi(u)| &=& |\big(w, F^\epsilon(u_*, \cdot)-F^\epsilon(u, \cdot)\big)| \\ &\lesssim& |\big(w, A(u_*-u)\big)| +|(w, u_*-u)| +|\big(w, \tilde{f}(u_*)-\tilde{f}(u)\big)|\\ &\lesssim& M_f(u^*, u)\|w\|\|u_*-u\|, \end{eqnarray} $

(4.29)式表明由(4.28)式定义的实值函数$ \Phi(\cdot) $$ \ell^2 $上连续.从(4.17)和(4.28)式可知映射$ u\mapsto \big(w, F^\epsilon(u, \cdot)\big)=\Psi(u) $$ m^\epsilon_t $ -可积的.同时, 我们已经证明

$ {\mathbb R} $上是分段连续的且$ \{\tau_j\}_{j\in {\Bbb Z}} $是其第一类间断点.因此它属于$ L^1_{\rm loc}({\mathbb R}) $.

最后证明(c).对任意的$ \Phi\in {\mathcal T} $, 从(4.21)式推出对任意的$ \Phi\in {\mathcal T} $, 有

故对于所有的$ t $, $ \tau\in (\tau_j, \tau_{j+1}) $, 其中$ t\geqslant \tau $$ j\in {\Bbb Z} $, 有

$ \begin{eqnarray} \Phi(u^\epsilon(t))-\Phi(u^\epsilon(\tau)) = \int_\tau^t\big(\Phi'(u^\epsilon(\theta)), F(u^\epsilon(\theta), \theta)\big) {\mathrm d}\theta. \end{eqnarray} $

现对于任意的$ s\leqslant \tau $, 记$ u_*\in \ell^2 $, $ u^\epsilon(\theta)=U_\epsilon(\theta, s)u_* $, 其中$ \theta\in[\tau, t] $.由(4.30)式得

$ \begin{eqnarray} \Phi(U_\epsilon(t, s)u_*)-\Phi(U_\epsilon(\tau, s)u_*) = \int_\tau^t\big(\Phi'(U_\epsilon(\theta, s)u_*), F^\epsilon(U_\epsilon(\theta, s)u_*, \theta)\big) {\mathrm d}\theta. \end{eqnarray} $

结合(4.17)–(4.18)和(4.31)式, 应用Fubini定理得到

应用(4.19)式和过程$ U_\epsilon(\theta, s)=U_\epsilon(\theta, \tau)U_\epsilon(\tau, s) $的不变性质, 就有

$ \begin{eqnarray} && \int_{\ell^2}\big(\Phi'(U_\epsilon(\theta, s)u_*), F^\epsilon(U_\epsilon(\theta, s)u_*, \theta)\big) {\mathrm d}{\mathit{ m}}^\epsilon_s(u_*) \\ &=& \int_{\ell^2}\big(\Phi'(U_\epsilon(\theta, \tau)u_*), F^\epsilon(U_\epsilon(\theta, \tau)u_*, \theta)\big) {\mathrm d}{\mathit{ m}}^\epsilon_\tau(u_*), \end{eqnarray} $

其中(4.32)式的右边与$ s $无关.因此当$ \tau_j<\tau\leqslant t<\tau_{j+1} $时, 有

$ \begin{eqnarray} && \int_{{\mathcal A}^\epsilon(t)}\Phi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_t(u^\epsilon) -\int_{{\mathcal A}^\epsilon(\tau)}\Phi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_\tau(u^\epsilon) {}\\ &=& \int_\tau^t\int_{\ell^2}\big(\Phi'(U_\epsilon(\theta, \tau)u_*), F^\epsilon(U_\epsilon(\theta, \tau)u_*, \theta)\big) {\mathrm d}{\mathit{ m}}^\epsilon_\tau(u_*){\mathrm d}\theta\\ &=& \int_\tau^t\int_{\ell^2}\big(\Phi'(u^\epsilon), F^\epsilon(u^\epsilon(s), s)\big) {\mathrm d}{\mathit{ m}}^\epsilon_s(u){\mathrm d}s. \end{eqnarray} $

定理4.2证明完毕.

注4.2  当所讨论的脉冲系统达到统计平衡点时, 系统的统计信息就不再随时间演化而改变, 也即$ \Phi'(u^\epsilon(t))=0 $.此时, (4.33)式表明

$ \begin{eqnarray} \int_{{\mathcal A}^\epsilon(t)}\Phi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_t(u^\epsilon) = \int_{{\mathcal A}^\epsilon(\tau)}\Phi(u^\epsilon){\mathrm d}{\mathit{ m}}^\epsilon_\tau(u^\epsilon), \; \tau_j<\tau\leqslant t<\tau_{j+1}, \forall\, j\in {\Bbb Z}. \end{eqnarray} $

(4.34)式意味着拉回吸引子$ {\mathcal A}^\epsilon(\tau) $的形状可能会随着时间从$ \tau $$ t $的演化而改变, 但是$ {\mathcal A}^\epsilon(\tau) $$ {\mathcal A}^\epsilon(t) $的总测度是相等的.这正是统计力学中$ \rm Liouville $定理的结论.需要注意的是, 等式(4.33)在包含任何脉冲点$ \tau_j $的区间$ (\tau, t) $上并不成立.这表明脉冲会破坏演化系统的某种类型的守恒.这种现象与人们对脉冲系统的直觉是一致的.

5 统计解的极限行为

这一节旨在研究当$ \epsilon\rightarrow 0^+ $时, 脉冲离散Ginzburg-Landau方程组的统计解和脉冲离散Schrödinger方程组的统计解之间的关系.我们先回顾一些关于脉冲离散Schrödinger方程组(即方程组(1.5)的结果).这些结果的证明类似定理2.1, 2.2和定理3.1.

引理5.1  假设$ \alpha $, $ \kappa $, $ \beta $都是正的常数, $ \epsilon\in(0, 1) $, $ \gamma\in(1, +\infty) $, 且$ \rm ({H1})–({H2}) $成立.则对于给定的$ \tau\in {{\Bbb R}} $$ u_{\tau^+}\in \ell^2 $, 问题(1.5)存在唯一的解$ u\in PC([\tau, +\infty);\ell^2)\cap PC^1((\tau, +\infty);\ell^2) $, 并且解算子$ U(t, \tau):u_{\tau^+}\longmapsto u(t; \tau, u_{\tau^+})=U(t, \tau)u_{\tau^+} $$ \ell^2 $上生成连续的过程$ \{U(t, \tau)\}_{t\geqslant \tau} $且存在一个满足定义3.1(3)的拉回$ {\mathcal D}_\sigma $ -吸引子$ \hat{{\mathcal A}}(t)=\{{\mathcal A}(t): t\in {{\Bbb R}} \} $.

我们还可以证明依赖于时间$ t $的一族闭球$ \hat{{\mathcal B}}(t)=\{{\mathcal B}(t)={\mathcal B}(0, R_\sigma(t)): t\in {{\Bbb R}} \} $就是过程$ \{U(t, \tau)\}_{t\geqslant \tau} $的有界拉回$ {\mathcal D}_\sigma $ -吸收集.因此, 对于任意的$ t\in {\mathbb R} $$ \widehat{D}=\{D(s):s\in {\mathbb R}\}\in {\mathcal D}_\sigma $, 存在$ \tau_{0}(t, \widehat{D})\leqslant t $使得$ U_\epsilon(t, \tau)D(\tau)\subseteq D_0(t) $, 其中$ \tau\leqslant \tau_{0}(t, \widehat{D}) $, $ \epsilon\in [0, 1) $, $ U_0(t, \tau)=U(t, \tau) $.同时, 由(2.19)式易看出

$ \begin{eqnarray} \|U_\epsilon(t, \tau) u_{\tau^+}\|^2 &\leqslant& \|u_{\tau^+}\|^2{\rm e}^{-\sigma(t-\tau)} +\frac{{\rm e}^{-\sigma t}}{\kappa} \int_\tau^t{\rm e}^{\sigma \theta}\|g(\theta)\|^2{\mathrm d}\theta\\ &\leqslant& G(t, \tau, \hat{D}){}\\ &:=& \sup\limits_{u_{\tau^+}\in D(\tau)}\|u_{\tau^+}\|^2 +\frac{{\rm e}^{-\sigma t}}{\kappa} \int_\tau^t{\rm e}^{\sigma \theta}\|g(\theta)\|^2{\mathrm d}\theta, \; \forall\, t>\tau, \, \forall\, \epsilon\in [0, 1), \end{eqnarray} $

其中$ \widehat{D}=\{D(s):s\in {\mathbb R}\}\in {\mathcal D}_\sigma $, $ u_{\tau^+}\in D(\tau) $.

引理5.2  假设$ \alpha $, $ \kappa $, $ \beta $都是正的常数, $ \epsilon\in(0, 1) $, $ \gamma\in(1, +\infty) $, 且$ \rm ({H1})–({H2}) $成立.则对于任意的$ \epsilon\in (0, 1) $, 对给定的$ \tau\in {{\Bbb R}} $, $ u_{\tau^+}\in \ell^2 $$ T>\tau $, 存在不依赖于$ \epsilon $的常数$ c=c(T, \tau, \hat{D}) $使得

$ \begin{eqnarray} \|U_\epsilon(t, \tau)u_{\tau^+}-U(t, \tau)u_{\tau^+}\| \leqslant c\epsilon, \; \forall\, t\in [\tau, T], \; \; \forall\, u_{\tau^+}\in \widehat{D}=\{D(s):s\in {\mathbb R}\}\in {\mathcal D}_\sigma. \end{eqnarray} $

  给定$ \widehat{D}=\{D(s):s\in {\mathbb R}\}\in {\mathcal D}_\sigma $, $ \tau\in {{\Bbb R}} $$ u_{\tau^+}\in \widehat{D} $.$ u^\epsilon=u^\epsilon(t)=U_\epsilon(t, \tau)u_{\tau^+} $, $ u=u(t)=U(t, \tau)u_{\tau^+} $, 则$ v=u^\epsilon-u $是下面问题的解

$ \begin{eqnarray} \left\{ \begin{array}{ll} {\rm i}\frac{{\mathrm d} v}{{\mathrm d}t}+\alpha Av+{\rm i}\epsilon Au^\epsilon+{\rm i}\kappa v +\beta|u^\epsilon|^{2\gamma}u^\epsilon -\beta|u|^{2\gamma}u=0, \; t>\tau, \, t\neq \tau_j, \, j\in {\Bbb Z}, \\ v(\tau_j^+)-v(\tau_j)=\phi_j(u^\epsilon(\tau_j))-\phi_j(u(\tau_j)), \; \, j\in {\Bbb Z}, \\ v(\tau^+)=0, \; \tau\in {{\Bbb R}} . \end{array} \right. \end{eqnarray} $

$ -{\rm i}v $与方程(5.3)作内积并取实部, 可得

$ \begin{equation} \frac 12\frac{{\mathrm d}}{{\mathrm d}t}\|v\|^2 +{\bf{Re}}\, \epsilon(B u^\epsilon, B v) +\kappa\|v\|^2 +{\bf{Im}}\, \beta(|u^\epsilon|^{2\gamma}u^\epsilon -|u|^{2\gamma}u, \bar{v}) =0, \; t>\tau, \, t\neq \tau_j, \, j\in {\Bbb Z}. \end{equation} $

由(2.2)式和Cauchy不等式, 得

$ \begin{equation} \big|{\bf{Re}}\, \epsilon(B u^\epsilon, B v)\big| \leqslant 4\epsilon\|u^\epsilon\|\|v\| \leqslant \frac{\kappa}{4}\|v\|^2 +\frac{32\epsilon^2}{\kappa}\|u^\epsilon\|^2. \end{equation} $

下面估计$ {\bf{Im}}\, \beta(|u^\epsilon|^{2\gamma}u^\epsilon -|u|^{2\gamma}u, \bar{v}) $.事实上, 我们有

$ \begin{eqnarray} \big|{\bf{Im}}\, \beta(|u^\epsilon|^{2\gamma}u^\epsilon -|u|^{2\gamma}u, \bar{v})\big| \leqslant \frac{\kappa}{4}\|v\|^2 +\frac{2\beta^2}{\kappa}\sum\limits_{m\in {\Bbb Z}} \big||u^\epsilon_m|^{2\gamma}u^\epsilon_m -|u_m|^{2\gamma}u_m\big|^2 \end{eqnarray} $

以及

$ \begin{eqnarray} && \sum\limits_{m\in {\Bbb Z}}\big||u^\epsilon_m|^{2\gamma}u^\epsilon_m -|u_m|^{2\gamma}u_m\big|^2{}\\ &=& \sum\limits_{m\in {\Bbb Z}}\big|f(|u^\epsilon_m|^2)u^\epsilon_m-f(|u_m|^2)u_m\big|^2 \\ &\leqslant& 2\sum\limits_{m\in {\Bbb Z}}\big|f(|u^\epsilon_m|^2)\big|^2|u^\epsilon_m-u_m\big|^2 + 2\sum\limits_{m\in {\Bbb Z}}\big|f(|u^\epsilon_m|^2)-f(|u_m|^2)|^2|u_m|^2 \\ &\leqslant& 2f^2(\|u^\epsilon\|^2)\|v\|^2 + 2\|u\|^2\underline{\sum\limits_{m\in {\Bbb Z}}\big|f(|u^\epsilon_m|^2)-f(|u_m|^2)|^2}. \end{eqnarray} $

接下来估计(5.7)中的下划线项.由微分中值定理可得

$ \begin{eqnarray} && \sum\limits_{m\in {\Bbb Z}}\big|f(|u^\epsilon_m|^2)-f(|u_m|^2)|^2{}\\ &=& \sum\limits_{m\in {\Bbb Z}}\big|f'(\theta|u^\epsilon_m|^2+(1-\theta)|u_m|^2)|^2 \big||u^\epsilon_m|+|u_m|\big|^2\big||u^\epsilon_m|-|u_m|\big|^2\\ &\leqslant& \gamma^2\sum\limits_{m\in {\Bbb Z}}\big|f'(\|u^\epsilon\|^2+\|u\|^2)|^2 \big(|u^\epsilon_m|^2+|u_m|^2\big)\big||u^\epsilon_m|-|u_m|\big|^2\\ &\leqslant& \gamma^2\big|f'(\|u^\epsilon\|^2+\|u\|^2)|^2\big(\|u^\epsilon\|^2+\|u\|^2\big)\|v\|^2. \end{eqnarray} $

现在对于给定$ \tau\in {{\Bbb R}} $, $ T>\tau $$ \widehat{D}=\{D(s):s\in {\mathbb R}\}\in {\mathcal D}_\sigma $, 由(5.1)式可知

$ \begin{eqnarray} \|u^\epsilon\|^2+\|u\|^2 \leqslant \bar{G}=\bar{G}(T, \tau, \widehat{D}) := 2\sup\limits_{\tau\leqslant t\leqslant T}G(t, \tau, \widehat{D}). \end{eqnarray} $

结合(5.4)–(5.9)式就得到

$ \begin{eqnarray} \frac{{\mathrm d}}{{\mathrm d}t}\|v\|^2 +\bigg(\kappa -\frac{8\beta^2\bar{G}^{4\gamma}}{\kappa} -4\gamma^2\bar{G}^{2\gamma-1}\bigg)\|v\|^2 \lesssim \bar{G}\epsilon^2, \; \tau<t\leqslant T, \, t\neq \tau_j, \, j\in {\Bbb Z}. \end{eqnarray} $

通过直接计算有

$ \begin{eqnarray} \|v(\tau_j^+)\|^2 &=& \sum\limits_{m\in {\Bbb Z}}|v_m(\tau_j^+)|^2{}\\ &=& \sum\limits_{m\in {\Bbb Z}}\big |v_m(\tau_j) +\phi_{mj}(u^{\epsilon}_m(\tau_j))-\phi_{mj}(u_m(\tau_j))\big|^2 \\ &\leqslant& 2\sum\limits_{m\in {\Bbb Z}}|v_m(\tau_j)|^2 + 2\sum\limits_{m\in {\Bbb Z}}\big|\phi_{mj}(u^{\epsilon}_m(\tau_j)) -\phi_{mj}(u_m(\tau_j))\big|^2 \\ &\leqslant& 2\sum\limits_{m\in {\Bbb Z}}|v_m(\tau_j)|^2 + 2\sum\limits_{m\in {\Bbb Z}}L^2 \big|u^{\epsilon}_m(\tau_j)-u_m(\tau_j)\big|^2{}\\ &=& (2+2L^2)\|v(\tau_j)\|^2. \end{eqnarray} $

结合(5.10)–(5.11)式, 引理2.2以及$ v(\tau^+)=0 $, 得到

证明完毕.

现在将方程(1.5)改写成下面形式

$ \begin{eqnarray} \frac{{\mathrm d} u}{{\mathrm d}t}= F(u, t):=-{\rm i}g(t)+{\rm i}\alpha Au-\kappa u+{\rm i}\beta\tilde{f}(u), \; \, t\neq \tau_j, \, j\in {\Bbb Z}. \end{eqnarray} $

脉冲离散Schrödinger方程组(5.12)统计解的定义与定义4.3相似, 只需要把$ F^\epsilon(u^\epsilon, t) $替换成$ F(u, t) $即可.类似定理4.1和4.2, 方程组(5.12)的Borel不变概率测度和统计解也可以通过拉回吸引子和广义Banach极限得到.

定理5.1  假设$ \alpha $, $ \kappa $, $ \beta $都是正的常数, $ \epsilon\in(0, 1) $, $ \gamma\in(1, +\infty) $, 且$ \rm ({H1})–({H2}) $成立.令$ v(\cdot):{\mathbb R}\mapsto \ell^2 $是一个连续映射满足$ v(\cdot)\in {\mathcal D}_\sigma $.则对于给定的广义$ {\rm Banach} $极限$ { } \lim_{t\rightarrow +\infty} $, 在$ \ell^2 $上存在唯一的一族Borel概率测度$ \{{\textsf m}_t\}_{t\in {{\Bbb R}} } $使得测度$ {\textsf m}_t $的支集包含于$ {\mathcal A}(t) $, 且对任意的$ \psi\in C(\ell^2) $

$ \begin{eqnarray} &&\lim\limits_{\tau\rightarrow -\infty}\frac{1}{t-\tau}\int_\tau^t \psi\big(U(t, s)v(s)\big){\mathrm d}s {}\\ &=& \int_{{\mathcal A}(t)}\psi(u){\mathrm d}{\textsf m}_t(u) = \int_{\ell^2}\psi(u){\mathrm d}{\textsf m}_t(u) \end{eqnarray} $

$ \begin{eqnarray} &=& \lim\limits_{\tau\rightarrow -\infty}\frac{1}{t-\tau}\int_\tau^t\int_{\ell^2} \psi\big(U(t, s)u\big){\mathrm d}{\textsf m}_s(u){\mathrm d}s. \end{eqnarray} $

另外, $ {\textsf m}_t $在下述意义具有不变性

且是方程组(5.12)的一个统计解.

这一节的主要结果是证明当$ \epsilon\rightarrow 0^+ $时, 脉冲离散Ginzburg-Landau方程组的统计解$ \{{\mathit{ m}}^\epsilon_t\}_{t\in {{\Bbb R}} } $收敛于脉冲离散Schrödinger方程组的统计解$ \{{\mathit{ m}}_t\}_{t\in {{\Bbb R}} } $.

定理5.2  假设$ \alpha $, $ \kappa $, $ \beta $都是正的常数, $ \epsilon\in(0, 1) $, $ \gamma\in(1, +\infty) $, 且$ \rm ({H1})–({H2}) $成立.则

$ \begin{eqnarray} \lim\limits_{\epsilon\rightarrow 0^+} \int_{\ell^2}\psi(u){\mathrm d}{\textsf m}^\epsilon_t(u) = \int_{\ell^2}\psi(u){\mathrm d}{\textsf m}_t(u), \; \forall\, \psi\in C(\ell^2), \forall\, \, t\in {{\Bbb R}} . \end{eqnarray} $

  令$ v(\cdot):{\mathbb R}\mapsto \ell^2 $是一个连续映射且满足$ v(\cdot)\in {\mathcal D}_\sigma $.对于每一个$ t\in {{\Bbb R}} $和任意的$ \tau<t $, 由引理5.2可得

因此, 对于一个给定的广义$ {\rm Banach} $极限$ \lim\limits_{t\rightarrow +\infty} $, 从(4.17)和(5.13)式可知

证明完毕.

参考文献

Ahmed N U .

Existence of optimal controls for a general class of impulsive systems on Banach spaces

SIAM J Control Optim, 2003, 42, 669- 685

DOI:10.1137/S0363012901391299      [本文引用: 1]

Bainov D D , Simeonov P S . Impulsive Differential Equations: Periodic solutions and applications. New York: John Wiley, 1993

[本文引用: 4]

Beyn W J , Pilyugin S Yu .

Attractors of reaction diffusion systems on infinite lattices

J Dyn Differential Equations, 2003, 15, 485- 515

DOI:10.1023/B:JODY.0000009745.41889.30      [本文引用: 1]

Bonotto E M , Bortolan M C , Caraballo T .

Attractors for impulsive non-autonomous dynamical systems and their relations

J Differential Equations, 2017, 262, 3524- 3550

DOI:10.1016/j.jde.2016.11.036     

Bonotto E M , Demuner D P , Jimenez M Z .

Convergence for non-autonomous semidynamical systems with impulses

J Differential Equations, 2019, 266, 227- 256

DOI:10.1016/j.jde.2018.07.035     

Bouchard B , Dang N M , Lehalle C A .

Optimal control of trading algorithms: A general impulse control approach

SIAM J Financial Math, 2011, 2, 404- 438

DOI:10.1137/090777293      [本文引用: 1]

Bronzi A , Mondaini C , Rosa R .

Trajectory statistical solutions for three-dimensional Navier-Stokes-like systems

SIAM J Math Anal, 2014, 46, 1893- 1921

DOI:10.1137/130931631      [本文引用: 1]

Bronzi A , Rosa R .

On the convergence of statistical solutions of the 3D Navier-Stokes-$\alpha$ model as $\alpha$ vanishes

Discrete Cont Dyn Syst, 2014, 34, 19- 49

DOI:10.3934/dcds.2014.34.19     

Bronzi A , Mondaini C , Rosa R .

Abstract framework for the theory of statistical solutions

J Differential Equations, 2016, 260, 8428- 8484

DOI:10.1016/j.jde.2016.02.027      [本文引用: 3]

Caraballo T , Morillas F , Valero J .

Attractors of stochastic lattice dynamical systems with a multipliative noise and non-Lipschitz nonlinearities

J Differential Equations, 2012, 253, 667- 693

DOI:10.1016/j.jde.2012.03.020      [本文引用: 1]

Caraballo T , Kloeden P , Real J .

Invariant measures and statistical solutions of the globally modified Navier-Stokes equations

Discrete Cont Dyn Syst-B, 2008, 10, 761- 781

[本文引用: 1]

Chekroun M , Glatt-Holtz N .

Invariant measures for dissipative dynamical systems: abstract results and applications

Comm Math Phys, 2012, 316, 723- 761

DOI:10.1007/s00220-012-1515-y      [本文引用: 2]

Chow S N , Paret J M .

Pattern formation and spatial chaos in lattice dynamical systems

IEEE Trans Circuits Syst, 1995, 42, 752- 756

DOI:10.1109/81.473584      [本文引用: 1]

Chow S N. Lattice Dynamical Systems//Macki J W, Zecca P. Dynamical Systems. Berlin: Springer-Verlag, 2003: 1-102

[本文引用: 1]

Ciesielski K .

On semicontinuity in impulsive dynamical systems

Bull Pol Acad Sci Math, 2004, 52, 71- 80

DOI:10.4064/ba52-1-8      [本文引用: 1]

Ciesielski K .

On stability in impulsive dynamical systems

Bull Pol Acad Sci Math, 2004, 52, 81- 91

DOI:10.4064/ba52-1-9     

Ciesielski K .

On time reparametrizations and isomorphisms of impulsive dynamical systems

Ann Polon Math, 2004, 84, 1- 25

DOI:10.4064/ap84-1-1      [本文引用: 1]

Davis M H , Guo X , Wu G .

Impulse control of multidimensional jump diffusions

SIAM J Control Optim, 2010, 48, 5276- 5293

DOI:10.1137/090780419      [本文引用: 1]

Foias C , Prodi G .

Sur les solutions statistiques des équations de Naiver-Stokes

Ann Mat Pura Appl, 1976, 111, 307- 330

DOI:10.1007/BF02411822     

Foias C , Manley O , Rosa R , Temam R . Navier-Stokes Equations and Turbulence. Cambridge: Cambridge University Press, 2001

[本文引用: 6]

Foias C , Rosa R , Temam R .

Properties of stationary statistical solutions of the three-dimensional Navier-Stokes equations

J Dyn Differential Equations, 2019, 31, 1689- 1741

DOI:10.1007/s10884-018-9719-2      [本文引用: 1]

Han X , Kloeden P .

Non-autonomous lattice systems with switching effects and delayed recovery

J Differential Equations, 2016, 261, 2986- 3009

DOI:10.1016/j.jde.2016.05.015      [本文引用: 1]

Iovane G , Kapustyan A V .

Global attractor for impulsive reaction-diffusion equation

Nonlinear Oscil, 2005, 8, 318- 328

DOI:10.1007/s11072-006-0004-7      [本文引用: 1]

Iovane G , Kapustyan A V , Valero J .

Asymptotic behavior of reaction-diffusion equations with non-damped impulsive effects

Nonlinear Anal, 2008, 68, 2516- 2530

DOI:10.1016/j.na.2007.02.002      [本文引用: 1]

Jiang H , Zhao C .

Trajectory statistical solutions and Liouville type theorem for nonlinear wave equations with polynomial growth

Adv Differential Equations, 2021, 26 (3/4): 107- 132

Karachalios N I , Yannacopoulos A N .

gloabl existence and compact attractors for the discrete nonlinear Schrödinger equation

J Differential Equations, 2005, 217, 88- 123

DOI:10.1016/j.jde.2005.06.002      [本文引用: 3]

Keener J P .

Propagation and its failure in coupled systems of discret excitable cells

SIAM J Appl Math, 1987, 47, 556- 572

DOI:10.1137/0147038      [本文引用: 1]

Kloeden P , Marín-Rubio P , Real J .

Equivalence of invariant measures and stationary statistical solutions for the autonomous globally modified Navier-Stokes equations

Comm Pure Appl Anal, 2009, 8, 785- 802

DOI:10.3934/cpaa.2009.8.785      [本文引用: 1]

Li H , Bates P W , Lu S .

Dynamics of the 3D fractional complex Ginzburg-Landau equation

J Differential Equations, 2015, 259, 5276- 5301

DOI:10.1016/j.jde.2015.06.028      [本文引用: 1]

李永军, 桑燕苗, 赵才地.

一阶格点系统的不变测度与Liouville型方程

数学物理学报, 2020, 40A (2): 328- 339

DOI:10.3969/j.issn.1003-3998.2020.02.006      [本文引用: 1]

Li Y J , Sang Y M , Zhao C D .

Invariant measures and Liouville type equation for first-order lattice system

Acta Mathematica Scientia, 2020, 40A (2): 328- 339

DOI:10.3969/j.issn.1003-3998.2020.02.006      [本文引用: 1]

Łukaszewicz G .

Pullback attractors and statistical solutions for 2-D Navier-Stokes equations

Discrete Cont Dyn Syst-B, 2008, 9, 643- 659

DOI:10.3934/dcdsb.2008.9.643      [本文引用: 2]

Łukaszewicz G , Robinson J C .

Invariant measures for non-autonomous dissipative dynamical systems

Discrete Cont Dyn Syst, 2014, 34, 4211- 4222

DOI:10.3934/dcds.2014.34.4211      [本文引用: 5]

Rosa R. Theory and applications of statistical solutions of the Navier-Stokes equations//Robinson J C, Rodrigo J L. London Mathematical Society Lecture Note Series. Cambridge: Cambridge University Press, 2009: 228-257

[本文引用: 1]

Schmalfuss B .

Attractors for non-autonomous and random dynamical systems perturbed by impulses

Discrete Cont Dyn Syst, 2003, 9, 727- 744

DOI:10.3934/dcds.2003.9.727      [本文引用: 1]

Vishik M , Fursikov A .

Translationally homogeneous statistical solutions and individual solutions with infinite energy of a system of Navier-Stokes equations

Siberian Math J, 1978, 19, 710- 729

Wang B .

Dynamics of systems on infinite lattices

J Differential Equations, 2006, 221, 224- 245

DOI:10.1016/j.jde.2005.01.003      [本文引用: 1]

Wang C , Xue G , Zhao C .

Invariant Borel probability measures for discrete long-wave-short-wave resonance equations

Appl Math Comp, 2018, 339, 853- 865

DOI:10.1016/j.amc.2018.06.059      [本文引用: 1]

Wang J , Zhao C , Caraballo T .

Invariant measures for the 3D globally modified Navier-Stokes equations with unbounded variable delays

Comm Nonl Sci Numer Simu, 2020, 91, 105459

DOI:10.1016/j.cnsns.2020.105459      [本文引用: 1]

Xu J H , Caraballo T .

Long time behavior of fractional impulsive stochastic differential equations with infinite delay

Discrete Cont Dyn Syst-B, 2019, 24, 2719- 2743

[本文引用: 1]

Xu J H , Zhang Z , Caraballo T .

Well-posedness and dynamics of impulsive fractional stochastic evolution equations with unbounded delay

Comm Nonlin Sci Numer Simu, 2019, 75, 121- 139

DOI:10.1016/j.cnsns.2019.03.002      [本文引用: 1]

Yan X , Wu Y , Zhong C .

Uniform attractors for impulsive reaction-diffusion equations

Appl Math Comp, 2010, 216, 2534- 2543

DOI:10.1016/j.amc.2010.03.095      [本文引用: 1]

Zhao C , Yang L .

Pullback attractor and invariant measures for three dimensional globally modified Navier-Stokes equations

Comm Math Sci, 2017, 15, 1565- 1580

DOI:10.4310/CMS.2017.v15.n6.a4      [本文引用: 1]

Zhao C , Xue G , Łukaszewicz G .

Pullabck attractor and invariant measures for the discrete Klein-Gordon-Schrödinger equatios

Discrete Cont Dyn Syst-B, 2018, 23, 4021- 4044

[本文引用: 2]

Zhao C , Caraballo T .

Asymptotic regularity of trajectory attractor and trajectory statistical solution for the 3D globally modified Navier-Stokes equations

J Differential Equations, 2019, 266, 7205- 7229

DOI:10.1016/j.jde.2018.11.032     

Zhao C , Li Y , Caraballo T .

Trajectory statistical solutions and Liouville type equations for evolution equations: Abstract results and applications

J Differential Equations, 2020, 269, 467- 494

DOI:10.1016/j.jde.2019.12.011      [本文引用: 2]

Zhao C , Li Y , Łukaszewicz G .

Statistical solution and partial degenerate regularity for the 2D non-autonomous magneto-micropolar fluids

Z Angew Math Phys, 2020, 71, 1- 24

DOI:10.1007/s00033-019-1224-x      [本文引用: 1]

Zhao C , Song Z , Caraballo T .

Strong trajectory statistical solutions and Liouville type equations for dissipative Euler equations

Appl Math Lett, 2020, 99, 105981

DOI:10.1016/j.aml.2019.07.012     

Zhao C , Li Y , Song Z .

Trajectory statistical solutions for the 3D Navier-Stokes equations: The trajectory attractor approach

Nonlinear Anal RWA, 2020, 53, 103077

DOI:10.1016/j.nonrwa.2019.103077     

Zhao C , Li Y , Sang Y .

Using trajectory attractor to construct trajectory statistical solutions for 3D incompressible micropolar flows

Z Angew Math Mech, 2020, 100, e201800197

Zhao C , Caraballo T , Łukaszewicz G .

Statistical solution and Liouville type theorem for the Klein-Gordon-Schrödinger equations

J Differential Equations, 2021, 281, 1- 32

DOI:10.1016/j.jde.2021.01.039      [本文引用: 1]

Zhao C , Jiang H , Caraballo T .

Statistical solutions and piecewise Liouville theorem for the impulsive reaction-diffusion equations on infinite lattices

Appl Math Comp, 2021, 404, 126103

DOI:10.1016/j.amc.2021.126103      [本文引用: 1]

Zhou S .

Random exponential attractor for cocycle and application to non-autonomous stochastic lattice systems with multiplicative white noise

J Differential Equations, 2017, 263, 2247- 2279

DOI:10.1016/j.jde.2017.03.044      [本文引用: 1]

Zhu Z , Sang Y , Zhao C .

Pullback attractor and invariant measures for the discrete Zakharov equations

J Appl Anal Comp, 2019, 9, 2333- 2357

[本文引用: 1]

/