数学物理学报  2018, Vol. 38 Issue (3): 527-542   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
刘功伟
刁林
具记忆和变时滞项的二阶发展方程能量衰减率的凸性估计
刘功伟1, 刁林2     
1. 河南工业大学理学院 郑州 450001;
2. 商丘学院计算机工程学院 河南商丘 476000
摘要:研究了如下一类具有记忆和变时滞项的抽象发展方程 $ \begin{array}{l} u_{tt}(t)+Au(t)-\int_{0}^{t}g(t-s)A u(s){\rm d}s+\mu_1h_1(u_t(t))+\mu_2h_2(u_t(x, t-\tau(t)))=\nabla F(u(t)). \end{array} $ 通过构造合适的能量泛函和Lyapunov泛函,利用凸函数的一些性质,得到了依赖于h1及记忆核g的能量衰减估计.此衰减估计可以应用于一些具体的模型.
关键词二阶发展方程    能量估计    时滞    记忆核    
On Convexity for Energy Decay Rates of the Second-Order Evolution Equation with Memory and Time-Varying Delay
Liu Gongwei1, Diao Lin2     
1. College of Science, Henan University of Technology, Zhengzhou 450001;
2. College of Computer Science and Technology, Shangqiu University, Henan Shangqiu 476000
Abstract: In this paper, we consider the following abstract evolution equation with memory and time-varying delay of the form $ \begin{array}{l} u_{tt}(t)+Au(t)-\int_{0}^{t}g(t-s)A u(s){\rm d}s+\mu_1h_1(u_t(t))+\mu_2h_2(u_t(x, t-\tau(t)))=\nabla F(u(t)). \end{array} $ By introducing suitable energy and Lyapunov functionals, and making use of some properties of the convex functions, we establish decay estimate for the energy, which depends on the behavior of h1 and the relaxation g. The decay estimate can be applied to various concrete models. We shall also give some applications to illustrate our result.
Key words: Second order evolution     Energy decay     Delay     Memory    
1 引言

该文主要研究了如下一类具有记忆和变时滞的抽象发展方程

$ \begin{equation}\label{1.1} \left\{\begin{array}{ll} u_{tt}(t)+Au(t)-\int_{0}^{t}g(t-s)A u(s){\rm d}s+\mu_1h_1(u_t(t)) \\ \ \;+\mu_2h_2(u_t(x, t-\tau(t)))=\nabla F(u(t)), &(x, t)\in\Omega\times(0, \infty), \\ u(x, t)=0, &(x, t)\in \partial\Omega\times(0, \infty), \\ u(x, 0)=u_0(x), \, \, \, u_t(x, 0)=u_1(x), &x\in \Omega, \\ u_t(x, t)=f_0(x, t), &(x, t)\in\Omega\times[-\tau(0), 0), \end{array}\right. \end{equation} $ (1.1)

其中$\Omega$$\mathbb{R}^n$中具有光滑边界$\partial \Omega$的有界区域, $\tau(t)>0$是变时滞, 常数$\mu_1$和实数$\mu_2$, $h_1, h_2$是给定的函数, 初始值$(u_0, u_1, f_0)$在一合适的函数空间中.这里, 下标$t$表示对时间$t$求偏导数, $A: D(A)\subset H \rightarrow H$是一个具有稠密域$D(A)$的线性自伴增生算子, $\nabla F$表示一个具有Gâteaux微分函数$F: D(\sqrt{A})\rightarrow \mathbb{R} $的梯度, $g$表示记忆核.

当方程(1.1)不含时滞($\mu_2=0$)时, 该问题包含许多模型并且在过去几十年里, 得到了广泛的研究和较好的结论.比如如下粘弹性波动方程的初边值问题(参见文献[7-8, 10, 16]及其参考文献)

$ \begin{equation}\label{1.2} u_{tt}-\triangle u+\int_0^tg(t-s)\triangle u(s){\rm d}s+h(u_t)=f(u). \end{equation} $ (1.2)

$f=-|u|^{\gamma}u$, $h(u_t)=a(x)u_t$, $a(x)$$g(s)$满足一定条件时, Cavalcanti等[6]得到了问题(1.2)的指数衰减估计. Berrimi和Messaoudi[4]构造一个新的函数减弱了对$a(x)$$g(s)$的限制条件, 进而改进了文献[6]的结论.在文献[20-21]中, Messaoudi和Mustafa研究了在部分边界上具有非线性阻尼的粘弹性方程, 给出了解的一般衰减估计式.

文献[1]研究了一类更一般的抽象方程

$ \begin{equation}\label{1.3} u_{tt}(t)+Au(t)-\int_{0}^{t}g(t-s)A u(s){\rm d}s=\nabla F(u(t)), ~~(x, t)\in\Omega\times(0, \infty), \end{equation} $ (1.3)

并且得到了解的一致稳定性估计.当方程(1.1)不含时滞($\mu_2=0$)时, Wu[34]得到了解的一般衰减估计并且给出几个具体的模型.

时滞$\mu_2h_2(u_t(x, t-\tau(t)))$的引入, 使得问题变得不同已有的文献.众所周知, 时滞出现在许多实际问题中, 它不仅依赖于现在的状态而且也依赖于过去的状态.时滞是一个不稳定源, 即使是一个较小的时滞也能使原来不具时滞的稳定系统变得不稳定, 除非在加入额外的控制项[12-13, 24-25, 33, 35].

Nicaise和Pignotti在文献[24]中研究了如下一类具有时滞的波动方程的Dirichlet问题

$ u_{tt}-\triangle u+\mu_1u_t+\mu_2u_t(x, t-\tau)=0, $

$0 <\mu_2 <\mu_1$时得到了系统的稳定性结果, 相反, 当$\mu_2\geq \mu_1$时, 存在一列时滞使得相应系统是不稳定的.当耗散和阻尼都在边界上时, 他们也得到类似的结论.在一维空间中, 利用谱分析的方法[35]也得到了类似的结论. Nicaise和Pignotti[27]将此结论推广到在边界或内部具有变时滞的波动方程.

最近, Kirane和Said-Houari在文献[15]中研究了如下粘弹性波动方程

$ u_{tt}-\triangle u+\int_0^{t}\mu(t-s)\triangle u(s){\rm d}s+\mu_1u_t+\mu_2u_t(x, t-\tau)=0. $

$0 <\mu_2\leq\mu_1$时, 他们得到了解的整体适定性和解的能量衰减估计.随后, Dai和Yang[11]改进了文献[15]的结论:在没有对$\mu_1, \mu_2>0$$\mu_2 <\mu_1$进一步限制时, 得到解的整体存在性, 以及当$\mu_1=0$时能量的指数衰减估计.

在文献[2, 30]中, Nicaise等研究了局部Kelvin-Voigt和混合边界条件的时滞波动方程.最近, Messaoudi, Fareh和Doudi[22]研究了如下一类具强阻尼和强时滞的波动方程

$ u_{tt}(x, t)-\triangle u(x, t)-\mu_1\triangle u_t(x, t)-\mu_2\triangle u_t(x, t-\tau)=0, $

证明了解的整体适定性($0 <|\mu_2|\leq \mu_1$), 能量的衰减估计($|\mu_2| <\mu_1$).同时, 他们也讨论了具有分布时滞的波动方程.

最近, 文献[29]研究了一类具有时滞和源项的抽象发展方程, 给出了当该模型的线性部分的$C_0$半群是指数稳定时, 若时滞条件足够小, 则整个系统也具有同样的性质.

这里, 我们还需指出如下方程的初边值问题

$ \begin{equation}\label{1.4} u_{tt}-\triangle u+\mu_1\sigma(t)h_1(u_t)+\mu_2\sigma(t)h_2(u_t(t-\tau(t)))=0. \end{equation} $

Benaissa等[5]利用Faedo-Galerkin方法得到了解在Sobolev空间中的整体存在性, 并且利用乘子方法得到解的稳定性行为.

最近, 在文献[31]中, Pignotti研究了如下方程

$ u_{tt}(t)+Au(t)-\int_{0}^{t}g(t-s)A u(s){\rm d}s+b(t)u_t(t-\tau)=0 $

当时滞项系数属于$L^1(0, \infty)$时, 解的稳定性行为.同时也研究了具有反向耗散时, 得到系统稳定性的结果.

综上所述, 对问题(1.1)的研究是很有意义的.类似文献[19, 26-27], 我们引入以下新的变量

$ z(x, \rho, t)=u_t(x, t-\rho\tau(t)), ~~x\in\Omega, ~~\rho\in(0, 1), ~~t>0. $

因此, 问题(1.1)等价于

$ \begin{equation}\label{1.5} \left\{\begin{array}{ll} u_{tt}(t)+Au(t)-\int_{0}^{t}g(t-s)A u(s){\rm d}s+\mu_1h_1(u_t(t)) \\ \ \;+\mu_2h_2(z(x, 1, t))=\nabla F(u(t)), &(x, t)\in\Omega\times(0, \infty), \\ \tau(t)z_t(x, \rho, t)+(1-\rho\tau'(t))z_{\rho}(x, \rho, t)=0, &(x, \rho, t)\in \Omega\times(0, 1)\times (0, \infty), \\ u(x, t)=0, &(x, t)\in \partial\Omega\times(0, \infty), \\ u(x, 0)=u_0(x), \, \, \, u_t(x, 0)=u_1(x), &x\in \Omega, \\ z(x, \rho, 0)=f_0(x, -\rho\tau(0)), &(x, \rho)\in\Omega\times[0.1). \end{array}\right. \end{equation} $ (1.5)

本文结构如下:第2节, 我们给出一些假设条件以及局部解的存在性结果; 第3节, 给出并证明主要结论, 同时给出几个具体的例子验证本文的结论.

2 预备知识

$H$是实Hilbert空间, 其内积为$\langle \cdot \rangle$, 范数为$\|\cdot \|.$在不引起混淆的情况下, 省略变量$x$$t$, $C$$C_i$表示正常数.

我们需要如下假设条件.

(A1) $A$是定义在$X $具有稠密域$D(A)$的自伴线性算子, 满足

$ \begin{equation}\label{2.1} \langle Ax, x\rangle\geq M\|x\|, ~~~x\in D(A), ~M>0. \end{equation} $ (2.1)

(A2)记忆核$g:[0, \infty)\rightarrow [0, \infty)$是绝对连续函数满足

$ \begin{equation}\label{2.2} g(0)>0, ~~~~~~~1-\int_0^{\infty}g(s){\rm d}s=l>0, \end{equation} $ (2.2)

并且存在非增可微正函数$\xi$使得

$ \begin{equation}\label{2.3} g'(t)\leq -\xi(t)g(t), ~~~~\forall t\geq0. \end{equation} $ (2.3)

(A3)设$F: D(\sqrt{A})\rightarrow \mathbb{R} $是已知的Gâteaux微分函数$\forall x\in D(\sqrt{A}).$类似文献[1], 假设

(a) 对任意$u\in D(\sqrt{A})$, 存在正常数$c(u)$使得

$ \begin{array}{l} |DF(u)(v)|\leq c(u)\|v\|, ~~~\forall v\in D(\sqrt{A}), \end{array} $

其中$DF(u)$$F$$u$点的Gâteaux微分.因此, $DF(u)$延拓到$H$, 记作$\nabla F(u)$

$ \begin{array}{l} \langle \nabla F(u), v\rangle=DF(u)(v), ~~~~\forall v\in H. \end{array} $

(b) 对任意$c>0$, 存在$L_c>0$使得当$u, v\in D(\sqrt{A})$满足$\|\sqrt{A}u\|, \|\sqrt{A}v\|\leq c$时, 都有

$ \begin{equation}\label{2.4} \|\nabla F(u)-\nabla F(v)\|\leq L_c\|\sqrt{A}u-\sqrt{A}v\|. \end{equation} $ (2.4)

(c) 存在严格单增连续函数$\psi:[0, \infty)\rightarrow [0, \infty)$使得

$ \begin{equation}\label{2.5} |F(u)|\leq \psi(\|\sqrt{A}u\|)\|\sqrt{A}u\|^2, ~~~\forall u\in D(\sqrt{A}). \end{equation} $ (2.5)

(A4) $h_1:\mathbb{R} \rightarrow\mathbb{R} $是非减连续函数, 并且存在存在严格单增函数$h_0\in C^1([0, \infty))$, $h_0(0)=0$, 以及正常数$c_1, c_2$使得

$ \begin{equation}\label{2.6} h_0(|s|)\leq |h_1(s)|\leq h_0^{-1}(|s|), ~~~ \forall\ |s|\leq \epsilon, \end{equation} $ (2.6)
$ \begin{equation}\label{2.7} c_1|s|\leq |h_1(s)|\leq c_2|s|, ~~~ \forall\ |s|\geq \epsilon. \end{equation} $ (2.7)

并且假设$H_1(s)=\sqrt{s}h_0(\sqrt{s})$是严格凸的$C^2[0, r^2]$函数, $r>0$.

(A5) $h_2:\mathbb{R} \rightarrow\mathbb{R} $非减$C^1(\mathbb{R} )$奇函数, 存在正常数$c_3, c_4, c_5$使得

$ \begin{equation}\label{2.8} |h_2'(s)|\leq c_3, ~~~~~c_4sh_2(s)\leq H_2(s)\leq c_5sh_1(s), \end{equation} $ (2.8)

其中$s\in \mathbb{R} $, $H_2(s)=\int_0^sh_2(t){\rm d}t$.

(A6)类似文献[26-27], $\tau\in W^{2, \infty}([0, T])$, $T>0$, 存在正常数$\tau_0, \tau_1$$d$使得$0 <\tau_0\leq \tau(t)\leq\tau_1$, $\tau'(t)\leq d <1$, $ \forall t>0.$并且耗散和时滞的系数满足

$ \begin{equation}\label{2.9} 0\leq |\mu_2| < \frac{c_4(1-d)}{c_5(1-c_4d)}\mu_1. \end{equation} $ (2.9)

注2.1  由条件(A4)可知$sh(s)>0, s\neq0.$由条件(A5)可知$H_2$偶凸函数满足$H_2(s)\leq sh_2(s), $从而$c_4\leq1.$

下面定义问题(1.5)对应的能量泛函

$ \begin{array}{l}\label{2.10} E(t)=\frac{1}{2}\|u_t(t)\|^2+\frac{1}{2}\bigg(1-\int_0^tg(s){\rm d}s\bigg)\|\sqrt{A}u(t)\|^2+\frac{1}{2}(g\circ\sqrt{A}u)(t) \\ -F(u)(t)+\frac{\kappa\tau(t)}{2}\int_{\Omega}\int_0^1H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x, \end{array} $ (2.10)

其中

$ \big(g\circ\sqrt{A}u\big)(t)=\int_0^tg(t-s)\|\sqrt{A}u(t)-\sqrt{A}u(s)\|^2{\rm d}s, $
$ \begin{equation}\label{2.11} \frac{2|\mu_2|(1-c_4)}{c_4(1-d)} <\kappa <\frac{2(\mu_1-c_5|\mu_2|)}{c_5}, \end{equation} $ (2.11)

由假设条件(2.9)知(2.11)式中$\kappa$的选取是合理的.

结合文献[5, 17]建立初边值问题(1.5)解的局部存在性:

定理2.1  假设条件(A1)-(A6)和(2.11)式成立, 如果$u_0\in D(A), u_1\in D(\sqrt{A})$, $f_0\in W^{1, 2}(\Omega, (-\tau(0), 0))$满足相容性条件$f_0(\cdot, 0)=u_1.$则存在时间$T>0$使得初边值问题(1.5)存在唯一强解

$ \begin{array}{l} u\in L^{\infty}(0, T; D(A)), u_t\in L^{\infty}(0, T; D(\sqrt{A})), ~~u_{tt}\in L^{\infty}(0, T; H). \end{array} $

并且, 如果$u_0\in D(\sqrt{A})$, $u_1\in H$, 初边值问题(1.5)存在唯一弱解

$ \begin{array}{l} u\in C([0, T]; D(\sqrt{A}))\cap C^1([0, T]; H). \end{array} $
3 主要结论及其证明

本节建立问题(1.5)整体解的能量衰减估计.

引理3.1  如果条件(A1)-(A6)和(2.11)式成立, 则存在正常数$\beta_1$, $\beta_2$使得

$ \begin{array}{l}\label{3.1} E'(t)&\leq& -\beta_1\langle h_1(u_t), u_t\rangle-\beta_2\langle h_2(z(1, t)), z(1, t)\rangle+\frac{1}{2}(g'\circ \sqrt{A}u)(t)-\frac{1}{2}g(t)\|\sqrt{A}u\|^2 \\ &\leq& 0. \end{array} $ (3.1)

  方程$(1.5)_1$两边同乘$u_t(t)$, 然后在$\Omega$上积分, 则有

$ \begin{array}{l}\label{3.2} \frac{\rm d}{{\rm d}t}\bigg(\frac{1}{2}\|u_t\|^2+\frac{1}{2}\|\sqrt{A}u\|^2-F(u)\bigg) \\ =\int_0^tg(t-s)\int_{\Omega}\sqrt{A}u_t(t)\sqrt{A}u(s){\rm d}x{\rm d}s -\mu_1\int_{\Omega}h_1(u_t)u_t{\rm d}x-\mu_2\int_{\Omega}h_2(z(x, 1, t))u_t{\rm d}x. \end{array} $ (3.2)

等式(3.2)右端第一项为

$ \begin{array}{l} \int_0^t g(t-s)\langle \sqrt{A}u_t(t), \sqrt{A}u(s)\rangle {\rm d}s\\ =-\int_0^tg(t-s)\int_{\Omega}\big(\sqrt{A}u(t)-\sqrt{A}u(s)\big)_t\big(\sqrt{A}u(t)-\sqrt{A}u(s)\big){\rm d}x{\rm d}s\\ +\int_0^tg(t-s)\int_{\Omega}\sqrt{A}u_t(t)\sqrt{A}u(s){\rm d}x{\rm d}s\\ \frac{1}{2}\frac{\rm d}{{\rm d}t}\bigg[\int_0^tg(s){\rm d}s\|\sqrt{A}u(t)\|^2-(g\circ\sqrt{A}u)(t)\bigg]+\frac{1}{2}(g'\circ\sqrt{A}u)(t)-\frac{1}{2}g(t)\|\sqrt{A}u(t)\|^2. \end{array} $

结合上式及$(1.5)_2$式, 则有

$ \begin{array}{l}\label{3.3} E'(t)=-\mu_1\int_{\Omega}h_1(u_t)u_t{\rm d}x-\mu_2\int_{\Omega}h_2(z(x, 1, t))u_t{\rm d}x-\frac{1}{2}g(t)\|\sqrt{A}u(t)\|^2+\frac{1}{2}(g'\circ\sqrt{A}u)(t) \\ +\frac{\kappa\tau'(t)}{2}\int_{\Omega}\int_0^1H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x+\frac{\kappa\tau(t)}{2}\int_{\Omega}\int_0^1h_2(z(x, \rho, t))z_t(x, \rho, t){\rm d}\rho {\rm d}x \\ =-\mu_1\int_{\Omega}h_1(u_t)u_t{\rm d}x-\mu_2\int_{\Omega}h_2(z(x, 1, t))u_t{\rm d}x-\frac{1}{2}g(t)\|\sqrt{A}u(t)\|^2+\frac{1}{2}(g'\circ\sqrt{A}u)(t) \\ +\frac{\kappa\tau'(t)}{2}\int_{\Omega}\int_0^1H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x-\frac{\kappa}{2}\int_{\Omega}\int_0^1(1-\rho\tau'(t))\frac{\rm d}{{\rm d}\rho}H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x \\ =-\mu_1\int_{\Omega}h_1(u_t)u_t{\rm d}x-\mu_2\int_{\Omega}h_2(z(x, 1, t))u_t{\rm d}x-\frac{1}{2}g(t)\|\sqrt{A}u(t)\|^2+\frac{1}{2}(g'\circ\sqrt{A}u)(t) \\ -\frac{\kappa}{2}\int_{\Omega}\{(1-\tau'(t))H_2(z(x, \rho, t))-H_2(z(x, 0, t))\}{\rm d}x. \end{array} $ (3.3)

类似于文献[32], (3.3)式右端估计如下

$ \begin{array}{l}\label{3.4} -\frac{\kappa}{2}\int_{\Omega}\{(1-\tau'(t))H_2(z(x, \rho, t))-H_2(z(x, 0, t))\}{\rm d}x \\ \leq -\frac{\kappa c_4}{2}(1-d)\langle h_2(z(1, t)), z(1, t)\rangle+\frac{\kappa c_5}{2}\langle h_1(u_t), u_t\rangle. \end{array} $ (3.4)

现在估计$I_0=-\mu_2\int_{\Omega}h_2(z(x, 1, t))u_t{\rm d}x.$定义凸函数$G$的对偶函数$ G^*(s)=\sup_{t\geq0}(st-G(t))$, 则有$st\leq G^*(s)+G(t)$, $s, t\geq0.$由文献[3]的讨论知, $G^*(s)=s(G')^{-1}(s)-G((G')^{-1}(s))$, $s\geq0.$故由$H_2$的定义则有

$ \begin{equation}\label{3.5} H_2^*(s)=sh_2^{-1}(s)-H_2(h_2^{-1}(s)), ~~~s\geq0. \end{equation} $ (3.5)

由(3.5)式得

$ \begin{equation}\label{3.6} I_0\leq|\mu_2|\int_{\Omega}\big(h_2(z(x, 1, t))z(x, 1, t)-H_2(z(x, 1, t))+H_2(u_t)\big){\rm d}x. \end{equation} $ (3.6)

事实上, 当$h_2(z(x, 1, t))>0$$u_t\leq 0$

$ \begin{array}{l} I_0=\mu_2\langle h_2(z(1, t)), -u_t(t)\rangle\\ \leq |\mu_2|\int_{\Omega}\big(H_2^*(h_2(z(x, 1, t)))+H_2(-u_t(x, t))\big){\rm d}x \\ =|\mu_2|\int_{\Omega}\big(h_2(z(x, 1, t)) z(x, 1, t)-H_2(z(x, 1, t))+H_2(u_t(x, t))\big){\rm d}x, \end{array} $

这里用到(3.5)式和$h_2$为奇函数及$H_2$为偶函数.类似当$h_2(z(x, 1, t))\leq0, u_t(t)>0$$h_2(z(x, 1, t))u_t(t)\geq0$时, 不等式(3.6)式仍成立.

由(3.6)和(2.8)式, 得

$ \begin{equation}\label{3.7} I_0\leq|\mu_2|(1-c_4)\langle h_2(z(1, t)), z(1, t)\rangle+|\mu_2|c_5\langle h_1(u_t), u_t\rangle. \end{equation} $ (3.7)

因此, 由(3.3), (3.4)和(3.7)式, 得

$ \begin{array}{l} E'(t) \leq -\Big(\mu_1-\frac{\kappa c_5}{2}-|\mu_2|c_5\Big)\langle h_1(u_t), u_t\rangle-\Big(\frac{\kappa c_4(1-d)}{2}-(1-c_4)|\mu_2|\Big)\langle h_2(z(1, t)), z(1, t)\rangle \\ +\frac{1}{2}(g'\circ \sqrt{A}u)(t)-\frac{1}{2}g(t)\|\sqrt{A}u\|^2\leq0. \end{array} $

由(2.11)式, 可知$\beta_1\doteq \mu_1-\frac{\kappa c_5}{2}-|\mu_2|c_5$$\beta_2\doteq \frac{\kappa c_4(1-d)}{2}-(1-c_4)|\mu_2|$为正常数.

引理3.2  如果条件(A1)-(A6)成立, 则存在$\rho_0>0$使得当$u_0\in D(\sqrt{A})$, $u_1\in H$满足

$ \begin{equation}\label{3.8} \|u_1\|^2+\|\sqrt{A}u_0\|^2+\kappa\tau(0)\int_{\Omega}\int_0^1H_2(f_0(x, -\rho\tau(0))){\rm d}\rho {\rm d}x <\rho_0^2 \end{equation} $ (3.8)

时, 问题(1.5)存在唯一整体弱解$u$.进一步, 对任意$t\geq0$$E(t)>0$

$ \begin{equation}\label{3.9} E(0)\leq\rho_0^2, \end{equation} $ (3.9)
$ \begin{equation}\label{3.10} E(t)\geq \frac{1}{2}\|u_t\|^2+\frac{l}{4}\|\sqrt{A}u(t)\|^2+\frac{1}{2}(g\circ\sqrt{A}u)(t), \end{equation} $ (3.10)
$ \begin{equation}\label{3.11} \psi(\|\sqrt{A}u(t)\|)\leq \frac{l}{4}. \end{equation} $ (3.11)

并且, 若$u_0\in D(A)$, $u_1\in D(\sqrt{A})$, 则$u$是问题(1.5)的强解.

  设$[0, T)$是问题(1.5)弱解$u$的最大存在区间.如果$\psi(\|\sqrt{A}u_0\|)<\frac{l}{4}, $则由(2.5)式得

$ \begin{array}{l} E(0)&= &\frac{1}{2}\|u_1\|^2+\frac{1}{2}\|\sqrt{A}u_0\|^2-F(u_0)+\frac{\kappa\tau(0)}{2}\int_{\Omega}\int_0^1H_2(f_0(x, -\rho\tau(0))){\rm d}\rho {\rm d}x \\ &\geq&\frac{1}{2}\|u_1\|^2+\frac{2-l}{4}\|\sqrt{A}u_0\|^2\geq0. \end{array} $

首先, 可以证明若

$ \begin{equation}\label{3.12} \psi(\|\sqrt{A}u_0\|) <\frac{l}{4}, ~~ \psi \bigg(\frac{2}{\sqrt{l}}E(0)^\frac{1}{2}\bigg) <\frac{l}{4}, \end{equation} $ (3.12)

$ \begin{equation}\label{3.13} E(t)\geq \frac{1}{2}\|u_t(t)\|^2+\frac{l}{4}\|\sqrt{A}u(t)\|^2, ~~~~\forall t\in[0, T). \end{equation} $ (3.13)

事实上, 设$\tau$是(3.13)式对任意$t\in[0, s)$成立的所有$s\in [0, T)$的上确界.若$\tau <T$, 则

$ E(\tau)\geq \frac{1}{2}\|u_t(\tau)\|^2+\frac{l}{4}\|\sqrt{A}u(\tau)\|^2\geq0. $

联立上式及(3.1)式, 则有

$ \psi(\|\sqrt{A}u(\tau)\|)\leq\psi \bigg(\frac{2}{\sqrt{l}}E(\tau)^\frac{1}{2}\bigg)\leq \psi \bigg(\frac{2}{\sqrt{l}}E(0)^\frac{1}{2}\bigg) <\frac{l}{4}, $
$ \begin{array}{l} E(\tau)&\geq &\frac{1}{2}\|u_t(\tau)\|^2+\frac{l}{4}\|\sqrt{A}u(\tau)\|^2+\frac{l}{4}\|\sqrt{A}u(\tau)\|^2-F(u(\tau))\\ &>&\frac{1}{2}\|u_t(\tau)\|^2+\frac{l}{4}\|\sqrt{A}u(\tau)\|^2, \end{array} $

最后一个不等式利用了(2.5)式.这与$\tau$的最大值性矛盾.因此$\tau=T.$

$\rho_0=\frac{\sqrt{l}}{2}\psi^{-1}(\frac{l}{4})>0.$我们将证明:对任意$u_0\in D(\sqrt{A})$, $u_1\in H$, $f_0\in L^2(\Omega\times(-\tau(0), 0))$满足

$ \begin{equation}\label{3.14} \|\sqrt{A}u_0\|^2+\|u_1\|^2+\kappa\tau(0)\int_{\Omega}\int_0^1H_2(f(x, -\rho\tau(0))){\rm d}\rho {\rm d}x <\rho_0^2 \end{equation} $ (3.14)

时, (3.12)式成立.事实上, 注意到(3.14)式表明$\|\sqrt{A}u_0\| <\rho_0$, 得

$ \psi(\|\sqrt{A}u_0\|) <\psi{(\rho_0)}=\psi\bigg(\frac{\sqrt{l}}{2}\psi^{-1} \Big(\frac{l}{4}\Big)\bigg) <\frac{l}{4}. $

因此, 由(2.5)式, 得

$ E(0)\leq \frac{1}{2}\|u_1\|^2+\frac{2+l}{4}\|\sqrt{A}u_0\|^2+\frac{\kappa\tau(0)}{2}\int_{\Omega}\int_0^1 H_2(f_0(x, -\rho\tau(0))){\rm d}\rho {\rm d}x <\rho_0^2. $

联立上式及$\rho_0$的定义, 得

$ \psi\bigg(\frac{2}{\sqrt{l}}E(0)^{\frac{1}{2}}\bigg) <\psi \bigg(\frac{2}{\sqrt{l}}\rho_0\bigg) <\psi\bigg(\psi^{-1}\Big(\frac{l}{4}\Big)\bigg)=\frac{l}{4}. $

总之, 当条件(3.14)成立时, (3.12)式成立.特别地

$ 0\leq \frac{1}{2}\|u_t(t)\|^2+\frac{l}{4}\|\sqrt{A}u(t)\|^2\leq E(t)\leq E(0)\leq \rho_0^2. $

因此, 系统的能量在$[0, T)$上非负有界, 从而是整体存在的.并且(3.13)式意味着(3.10)式成立, 从而由下式知(3.11)式成立

$ \psi(\|\sqrt{A}u(t)\|)\leq \psi \bigg(\frac{2}{\sqrt{l}}E(t)^{\frac{1}{2}}\bigg)\leq \psi \bigg(\frac{2}{\sqrt{l}}E(0)^{\frac{1}{2}}\bigg) <\frac{l}{4}. $

证毕.

现定义能量泛函的扰动函数

$ \begin{equation}\label{3.15} {\cal L}(t)=N_1E(t)+N_2\Psi(t)+\Phi(t)+\chi(t), \end{equation} $ (3.15)

其中

$ \begin{equation}\label{3.16} \Phi(t)=\int_\Omega u_tu{\rm d}x, \end{equation} $ (3.16)
$ \begin{equation}\label{3.17} \Psi(t)=\int_{\Omega}u_t\int_0^tg(t-s)(u(s)-u(t)){\rm d}s{\rm d}x, \end{equation} $ (3.17)
$ \begin{equation}\label{3.18} \chi(t)=\tau(t)\int_{\Omega}\int_0^1{\rm e}^{-\rho\tau(t)}H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x, \end{equation} $ (3.18)

$N_1$, $N_2$是将在后面确定的正常数.

类似于文献[34], 对于$u\in D(\sqrt{A})$, 由(2.1)和(2.2)式, 得

$ \begin{equation}\label{3.19} \int_{\Omega}\bigg(\int_0^tg(t-s)(u(t)-u(s)){\rm d}s\bigg)^2{\rm d}x\leq\frac{1-l}{M}(g\circ\sqrt{A}u)(t). \end{equation} $ (3.19)

引理3.3  当$N_1$$N_2$足够大时, 存在正常数$\alpha_1$$\alpha_2$使得

$ \begin{equation}\label{3.20} \alpha_1E(t)\leq {\cal L}(t)\leq\alpha_2E(t). \end{equation} $ (3.20)

  由Young不等式得

$ |\Phi(t)|\leq \|u_t\|^2+\frac{1}{2M}\|\sqrt{A}u\|^2, ~~~ |\Psi(t)|\leq \frac{1}{2}\|u_t\|^2+\frac{1-l}{2M}(g\circ\sqrt{A}u)(t). $

并且, 存在正常数$c <1$使得

$ |\chi(t)|\leq c\tau(t)\int_{\Omega}\int_0^1H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x. $

因此, 选取$N_2>0$后再取$N_1$足够大, 则(3.20)式成立.

为了得到系统能量的衰减估计, 需要对$F$进一步假设.

(A7) $F(0)=0, \nabla F(0)=0$, 存在严格增函数$\psi:[0, \infty)\rightarrow[0, \infty)$使得

$ \begin{equation}\label{3.21} |\langle \nabla F(u), u\rangle|\leq \psi(\|\sqrt{A}u\|)\|\sqrt{A}u\|^2, ~~~u\in D(\sqrt{A}). \end{equation} $ (3.21)

注3.1  类似文献[1, 34], 由假设条件(A7)可得(2.5)式.事实上, 对$u\in D(\sqrt{A})$, 有

$ \begin{array}{l} |F(u)|\leq \int_0^1|\langle \nabla F(tu), u\rangle|{\rm d}t \leq \|\sqrt{A}u\|^2\int_0^1\psi(t\|\sqrt{A}u\|)t{\rm d}t \leq \frac{1}{2}\psi(\|\sqrt{A}u\|)\|\sqrt{A}u\|^2. \end{array} $

定理3.1  假设条件(A1)-(A7)和(3.8)式成立, $u_0\in D(\sqrt{A})$, $u_1\in H, $则对$t_0>0$, 存在正常数$k_1, ~k_2$, $k_3$$\varepsilon_0$, 使得系统能量满足

$ \begin{equation}\label{3.22} E(t)\leq k_1H^{-1}\bigg(k_2\int_{t_0}^t\xi(s){\rm d}s+k_3\bigg), ~~~~~t\geq t_0, \end{equation} $ (3.22)

其中$ H(t)=\int_t^1\frac{1}{H_0(s)}{\rm d}s, $ $H_0(t)=t$, 若$H_1$在[0, r]是线性的; $H_0(t)=tH_1'(\varepsilon_0t)$, 若$H_1'(0)=0$, $H_1''(t)>0, (0, r]$.这里$H$是在$(0, 1]$上的严格递减凸函数满足$ \mathop {\lim }\limits_{t \to 0} H(t)=+\infty$.

  为了得到能量$E(t)$的衰减估计, 首先需要估计${\cal L}'(t).$

第一步  估计$\Phi'(t).$利用(1.5)式, 由(2.1)和(2.2)式, 得

$ \begin{array}{l}\label{3.23} \Phi'(t)&\leq &\|u_t\|^2-l\|\sqrt{A}u\|^2+\int_{\Omega}\sqrt{A}u(t)\int_0^tg(t-s) \big(\sqrt{A}u(s)-\sqrt{A}u(t)\big){\rm d}s{\rm d}x \\ & &-\mu_1\int_{\Omega}h_1(u_t)u{\rm d}x-\mu_2\int_{\Omega}h_2(z(x, 1, t))u{\rm d}x+\langle\nabla F(u(t)), u(t)\rangle. \end{array} $ (3.23)

由Young不等式, 立即可得

$ \int_{\Omega}\sqrt{A}u(t)\int_0^tg(t-s)(\sqrt{A}u(s) -\sqrt{A}u(t)){\rm d}s{\rm d}x\leq \eta \|\sqrt{A}u\|^2+\frac{1-l}{4\eta}(g\circ\sqrt{A}u)(t), $
$ |\mu_1\int_{\Omega}h_1(u_t)u{\rm d}x|\leq \eta\|\sqrt{A}u\|^2+\frac{\mu_1^2}{4\eta M}\|h_1(u_t)\|^2, $
$ |\mu_2\int_{\Omega}h_2(z(x, 1, t))u{\rm d}x|\leq \eta\|\sqrt{A}u\|^2+\frac{\mu_2^2}{4\eta M}\|h_2(z(1, t))\|^2, $
$ |\langle\nabla F(u(t)), u(t)\rangle|\leq \frac{l}{4}\|\sqrt{A}u\|^2. $

将这些式子代入(3.23)式, 令$\eta=\frac{l}{12}$, 则有

$ \begin{equation}\label{3.24} \Phi'(t)\leq \|u_t\|^2-\frac{l}{2}\|\sqrt{A}u\|+C_1\big((g\circ\sqrt{A}u)(t)+\|h_1(u(t))\|^2+\|h_2(z(1, t))\|^2\big). \end{equation} $ (3.24)

第二步  估计$\Psi'(t).$由(1.5)式, 得

$ \begin{array}{l}\label{3.25} \Psi'(t)=\int_{\Omega}\sqrt{A}u(t)\int_0^tg(t-s)\big(\sqrt{A}u(t)-\sqrt{A}u(s) \big){\rm d}s{\rm d}x \\ +\int_{\Omega}\nabla F(u)\int_0^tg(t-s)(u(t)-u(s)){\rm d}s{\rm d}x \\ -\mu_1\int_{\Omega}h_1(u_t)\int_0^tg(t-s)(u(t)-u(s)){\rm d}s{\rm d}x \\ -\mu_2\int_{\Omega}h_2(z(x, 1, t))\int_0^tg(t-s)(u(t)-u(s)){\rm d}s{\rm d}x \\ -\int_{\Omega}\bigg(\int_0^tg(t-s)\sqrt{A}u(s){\rm d}s\bigg)\bigg(\int_0^tg(t-s)(\sqrt{A}u(t)-\sqrt{A}u(s)){\rm d}s\bigg){\rm d}x \\ -\int_{\Omega}u_t\int_0^tg'(t-s)(u(t)-u(s)){\rm d}s{\rm d}x-\bigg(\int_0^tg(s)\bigg)\|u_t\|^2 \\ \doteq \sum\limits_{n=1}^{7}I_n. \end{array} $ (3.25)

下面需要估计(3.25)式中的$I_n$.类似于文献[34]的讨论, 我们有如下估计

$ I_1\leq \delta\|\sqrt{A}u\|^2+\frac{1-l}{4\delta}(g\circ\sqrt{A}u)(t), $
$ I_3\leq \delta\|h_1(u_t)\|^2+\frac{(1-l)\mu_1^2}{4\delta M}(g\circ \sqrt{A}u)(t), $
$ I_4\leq \delta\|h_2(z(1, t))\|^2+\frac{(1-l)\mu_2^2}{4\delta M}(g\circ \sqrt{A}u)(t), $
$ I_5\leq2\delta (1-l)^2\|\sqrt{A}u\|^2+(2\delta+\frac{1}{4\delta})(g\circ\sqrt{A}u)(t), $
$ I_6\leq \delta\|u_t\|^2-\frac{g(0)}{4\delta M}(g'\circ\sqrt{A}u)(t). $

$I_2$在文献[1, 34]中已证, 为完备见, 简证如下:事实上, 由(3.1), (3.9)和(3.10)式得$\|\sqrt{A}u\|\leq \frac{2E(t)^\frac{1}{2}}{\sqrt{l}}\leq \frac{2\rho_0}{\sqrt{l}}, $注意到$\nabla F(0)=0$和(2.4)式, 可知$\|\nabla F\|\leq C\|\sqrt{A}u\|$, 其中正常数$C=C(\rho_0, l)>0$.因此, 由(3.19)式, 得

$ I_2\leq C^2\delta \|\sqrt{A}u\|^2+\frac{1-l}{4\delta M}(g\circ\sqrt{A}u)(t). $

将这些估计式代入(3.25)式, 得

$ \begin{array}{l}\label{3.26} \Psi'(t)&\leq &-\bigg(\int_0^tg(s){\rm d}s-\delta\bigg)\|u_t\|^2+\delta\big(\|h_1(u_t)\|^2+\|h_2(z(1, t))\|^2\big)+\delta C_2\|\sqrt{A}u\|^2 \\ & &+\frac{C_3}{\delta}\big(g\circ\sqrt{A}u\big)(t)-\frac{g(0)}{4\delta M}(g'\circ\sqrt{A}u)(t). \end{array} $ (3.26)

第三步  估计$\chi'(t)$.类似于文献[32], 注意到

$ \begin{array}{l} \chi'(t)=\tau'(t)\int_{\Omega}\int_0^1{\rm e}^{-\rho\tau(t)}H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x-\tau(t)\int_{\Omega}\int_0^1\rho \tau'(t){\rm e}^{-\rho\tau(t)}H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x \\ +\tau(t)\int_{\Omega}\int_0^1{\rm e}^{-\rho\tau(t)}h_2(z(x, \rho, t))z_t(x, \rho, t){\rm d}\rho {\rm d}x. \end{array} $

利用方程$(1.5)_2$, 分部积分得

$ \begin{array}{l} \tau(t)\int_{\Omega}\int_0^1{\rm e}^{-\rho\tau(t)}h_2(z(\rho, t))z_t(x, \rho, t) {\rm d}\rho {\rm d}x\\ =-\int_{\Omega}\int_0^1{\rm e}^{-\rho\tau(t)}(1-\rho\tau'(t))\frac{\rm d}{{\rm d}\rho}H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x \\ =-\int_{\Omega}{\rm e}^{-\tau(t)}(1-\tau'(t))H_2(z(x, 1, t)){\rm d}x+\int_{\Omega}H_2(z(x, 0, t)){\rm d}x \\ -\int_{\Omega}\int_0^1\big(\tau(t)(1-\rho\tau'(t))+\tau'(t)\big){\rm e}^{-\rho\tau(t)}H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x. \end{array} $

将上式代入$\chi'(t)$, 由条件(A5)和(A6), 得

$ \begin{array}{l}\label{3.27} \chi'(t)=-{\rm e}^{\tau(t)}\int_{\Omega}\big(1-\tau'(t)\big)H_2(z(x, \rho, t)){\rm d}x+\int_{\Omega}H_2(u_t){\rm d}x-\chi(t) \\ \leq -\chi(t)-c_4{\rm e}^{-\tau_1}(1-d)\langle h_2(z(1, t)), z(1, t)\rangle+c_5\langle h_1(u_t), u_t\rangle. \end{array} $ (3.27)

第四步  联立(3.1), (3.24), (3.26)和(3.27)式, 则有

$ \begin{array}{l} {\cal L}'(t)=N_1E'(t)+N_2\Psi'(t)+\Phi'(t)+\chi'(t)\\ \leq \bigg(\frac{N_1}{2}-\frac{g(0)N_2}{4\delta}\bigg)(g'\circ\sqrt{A}u)(t) -\bigg(N_2\bigg(\int_0^tg(s){\rm d}s-\delta\bigg)-1\bigg)\|u_t\|^2\\ +\bigg(\delta C_2N_2-\frac{l}{2}\bigg)\|\sqrt{A}u\|^2 +\bigg(\frac{N_2C_3}{\delta}+C_2\bigg)(g\circ \sqrt{A}u)\\ +(C_1+N_2\delta)(\|h_1(u_t)\|^2+\|h_2(z(1, t))\|^2) -(N_1\beta_1-c_5)\langle h_1(u_t), u_t\rangle \\ +\big(N_1\beta_2+c_4{\rm e}^{-\tau_1}(1-d)\big)\langle h_2(z(1, t)), z(1, t)\rangle-\chi(t). \end{array} $

$g(0)>0$, $g$是正的连续函数, 则对任何$t_0>0$, $\int_0^tg(s){\rm d}s \geq\int_0^{t_0}g(s){\rm d}s=g_0, \forall t\geq t_0$.由(2.8)式知$\|h_2(z(1, t))\|^2\leq c_3\langle h_2(z(1, t)), z(1, t)\rangle.$并且由条件(A6)可知

$ \begin{array}{l} -\chi(t)\leq-\tau(t){\rm e}^{-\tau_1}\int_{\Omega}\int_0^1H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x. \end{array} $

联立这些估计式, 则有

$ \begin{array}{l} {\cal L}'(t) \leq \bigg(\frac{N_1}{2}-\frac{g(0)N_2}{4\delta}\bigg) (g'\circ\sqrt{A}u)(t)-\big(N_2(g_0-\delta)-1\big)\|u_t\|^2 \\ -\tau(t){\rm e}^{-\tau_1}\int_{\Omega}\int_0^1H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x +\bigg(\delta C_2N_2-\frac{l}{2}\bigg)\|\sqrt{A}u\|^2\\ +\bigg(\frac{N_2C_3}{\delta}+C_2\bigg)(g\circ \sqrt{A}u) +(C_1+N_2\delta)\|h_1(u_t)\|^2 -(N_1\beta_1-c_5)\langle h_1(u_t), u_t\rangle \\ +\big(N_1\beta_2+c_4{\rm e}^{-\tau_1}(1-d)-c_3(C_1+N_2\delta)\big)\langle h_2(z(1, t)), z(1, t)\rangle. \end{array} $

$\delta=\frac{l}{4C_2N_2}$, 取$N_2$足够大使得$N_2g_0-\frac{l}{4C_2}>\frac{3}{2}.$然后取$N_1$足够大使得$\frac{N_1}{2}-\frac{g(0)N_2}{4\delta}>0$, $N_1\beta_2+c_4{\rm e}^{-\tau_1}(1-d)-c_3(C_1+N_2\delta)>0$$N_1\beta_1-c_5>0$.

另一方面, 由(2.5), (2.10)和(3.11)式, 得

$ \begin{array}{l} E(t)\leq C(\|u_t\|^2+\|\sqrt{A}u\|^2+(g\circ\sqrt{A}u)(t)+\int_{\Omega}\int_0^1H_2(z(x, \rho, t)){\rm d}\rho {\rm d}x). \end{array} $

因此, 存在正常数$\alpha>0$, 使得

$ \begin{equation}\label{3.28} {\cal L}'(t)\leq -\alpha E(t)+C_4(g\circ\sqrt{A}u)(t)+C_5\|h_1(u_t)\|^2. \end{equation} $ (3.28)

不等式(3.28)两端同乘$\xi(t)$, 由(2.3)式, 得

$ \begin{array}{l} \xi(t){\cal L}'(t)&\leq &-\alpha\xi(t)E(t) +C_4\xi(t)(g\circ\sqrt{A}u)(t)+C_5\xi(t)\|h_1(u_t)\|^2\\ &\leq &-\alpha\xi(t)E(t) -C_4\xi(t)(g'\circ\sqrt{A}u)(t)+C_5\xi(t)\|h_1(u_t)\|^2\\ &\leq &-\alpha\xi(t)E(t) -C\xi(t)E'(t)+C_5\xi(t)\|h_1(u_t)\|^2. \end{array} $

利用(3.20)式, 注意到$\xi'(t)\leq0$, 则函数$M(t)=\xi(t){\cal L}(t)+CE(t)$满足

$ \begin{equation}\label{3.29} M(t)\sim E(t) \end{equation} $ (3.29)

$ \begin{equation}\label{3.30} M'(t)\leq -\alpha\xi(t)E(t)+C_5\xi(t)\|h_1(u_t)\|^2, ~~~~~ \forall t\geq t_0. \end{equation} $ (3.30)

情形1   $h_0$线性:由条件(A4)得$c_1'|s|\leq|h_1(s)|\leq c_2'|s|$$ h_1^2(s)\leq c_2'sh_1(s)$, $\forall s$.则(3.30)式为

$ \begin{array}{l} M'(t)\leq -\alpha\xi(t)E(t)+C\xi(t)\int_{\Omega}h_1(u_t)u_t{\rm d}x\leq-\alpha \xi(t)E(t)-CE'(t), \end{array} $

$(M+CE)'(t)\leq -\alpha\xi(t)E(t)$ $\forall t\geq t_0.$注意到$M+CE\sim E, $$ E(t)\leq C'{\rm e}^{-C\int_0^t\xi(s){\rm d}s}. $此时, $H_1(s)=\sqrt{s}h_0(\sqrt{s})=Cs, $因而$E(t)\leq k_1H^{-1}\big(\int_0^t\xi(s){\rm d}s\big).$

情形2   $h_0$$[0, \epsilon]$上非线性.假设$\epsilon_1=\min\{r, h_0(r)\} <\epsilon;$否则取$r$足够小.则对$\epsilon_1\leq |s|\leq\epsilon$, 由(A4)得

$ |h(s)|\leq \frac{h_0^{-1}(|s|)}{|s|}|s|\leq\frac{h_0^{-1}(\epsilon)}{\epsilon_1}|s|, \quad |h(s)|\geq\frac{h_0(|s|)}{|s|}|s|\geq\frac{h_0(\epsilon_1)}{\epsilon}|s|. $

因此

$ \begin{equation}\label{3.31} h_0(|s|)\leq |h_1(s)|\leq h_0^{-1}(|s|), ~~|s|\leq \epsilon_1; \quad c_1'|s|\leq |h_1(s)|\leq c_2'|s|, ~~ |s|\geq \epsilon_1. \end{equation} $

现取$\Omega_1=\{x\in\Omega:|u_t|>\epsilon_1\}$$\Omega_2=\{x\in\Omega:|u_t|\leq\epsilon_1\}$, 则由(3.31)式及$\epsilon_1$的定义得:在$\Omega_2$上有

$ \begin{equation}\label{3.32} u_th_1(u_t)\leq \epsilon_1h_0^{-1}(\epsilon_1)\leq h_0(r)r=H_1(r^2), \quad u_th_1(u_t)\leq \epsilon_1h_0^{-1}(\epsilon_1)\leq rh_0^{-1}(h_0(r))(r)=r^2. \end{equation} $ (3.32)

由Jensen's不等式得

$ \begin{equation}\label{3.33} H_1^{-1}(\lambda(t))\geq \frac{1}{|\Omega_2|}\int_{\Omega_2}H_1^{-1}(u_th_1(u_t)){\rm d}x, \end{equation} $ (3.33)

其中$\lambda(t)=\frac{1}{|\Omega_2|}\int_{\Omega_2}u_th_1(u_t){\rm d}x\in [0, \epsilon_1^2].$注意到在$\Omega_2$上满足, $H_1(s^2)=|s|h_0(|s|)$, 故由(3.31)式得

$ \begin{array}{l} H_1(h_1^2(u_t))=|h_1(u_t)|h_0(|h_1(u_t)|)\leq |h_1(u_t)|h_0(h_0^{-1}(u_t))=h_1(u_t)u_t. \end{array} $

注意到$H_1^{-1}$递增, 由(3.1)和(3.33)式, 得

$ \begin{array}{l} \int_{\Omega}h_1^2(u_t){\rm d}x = \int_{\Omega_2}h_1^2(u_t){\rm d}x+ \int_{\Omega_1}h_1^2(u_t){\rm d}x\\ \leq \int_{\Omega_2}H_1^{-1}(h_1(u_t)u_t){\rm d}x+\int_{\Omega_1}h_1(u_t)u_t{\rm d}x\\ \leq |\Omega_2|H^{-1}_1(\lambda(t))-CE'(t). \end{array} $

因此, (3.30)式即为

$ \begin{equation}\label{3.34} M_0(t)\leq-\alpha \xi(t)E(t)+|\Omega_2|\xi(t)H_1^{-1}(\lambda(t)), \quad t\geq t_0, \end{equation} $ (3.34)

其中$M_0(t)=M(t)+CE(t)$是和$E(t)$等价的.

类似于文献[21, 34], 定义

$ M_1(t)=H_1'\Big(\epsilon_0\frac{E(t)}{E(0)}\Big)M_0(t)+c_0E(t), $

其中$\epsilon_0 <r^2$, $c_0>0.$$H'(t)\geq0$ $t\in (0, r^2]$, 知存在$\alpha_1, \alpha_2>0$使得

$ \begin{equation}\label{3.35} \alpha_1M_1(t)\leq E(t)\leq \alpha_2M_1(t), \end{equation} $ (3.35)

注意到$E'(t)\leq 0$, $H_1$$(0, r^2]$上严格凸, 由(3.34)式, 得

$ \begin{array}{l}\label{3.36} M_1'(t)=\epsilon_0\frac{E'(t)}{E(0)}H_1'' \Big(\epsilon_0\frac{E(t)}{E(0)}\Big)M_0(t)+H_1' \Big(\epsilon_0\frac{E(t)}{E(0)}\Big)M_0'(t)+c_0E'(t) \\ \leq -\alpha \xi(t)E(t)H_1'\Big(\epsilon_0\frac{E(t)}{E(0)}\Big) +|\Omega_2|\xi(t)H_1'\Big(\epsilon_0\frac{E(t)}{E(0)}\Big)H_1^{-1}(\lambda(t)) +c_0E'(t). \end{array} $ (3.36)

$H_1^*$$H_1$的凸对偶, 类似于引理3.1, 得

$ \begin{equation}\label{3.37}\begin{array}{ll} H_1^*(s)=s(H_1')^{-1}(s)-H_1[(H_1')^{-1}(s)], ~~s\in (0, H_1'(r^2)];\\ AB\leq H_1^*(A)+ H_1(B), ~~ A\in (0, H_1'(r^2), ~~B\in (0, r^2]. \end{array}\end{equation} $ (3.37)

因此, 取$A=H_1'(\epsilon_0\frac{E(t)}{E(0)})$, $B=H_1^{-1}(\lambda(t))$, 联立(3.1), (3.32)和(3.37)式, 得

$ \begin{array}{l} M_1(t)&\leq& -\alpha \xi(t)E(t)H_1' \Big(\epsilon_0\frac{E(t)}{E(0)}\Big)+|\Omega_2|\xi(t)H_1^* \Big(H_1'\Big(\epsilon_0\frac{E(t)}{E(0)}\Big)\Big) +|\Omega_2|\xi(t)\lambda(t)+c_0E'(t)\\ &\leq &-\alpha \xi(t)E(t)H_1' \Big(\epsilon_0\frac{E(t)}{E(0)}\Big)+|\Omega_2|\xi(t)\epsilon_0 \frac{E(t)}{E(0)}H_1'\Big(\epsilon_0\frac{E(t)}{E(0)}\Big) -\xi(0)E'(t)+c_0E'(t). \end{array} $

现在, 令$c_0>\xi(0)$$\epsilon_0 <\min\{r^2, \frac{\alpha E(0)}{|\Omega_2|}\}$, 则对$\forall t\geq t_0$, 存在$k>0$使得

$ \begin{equation}\label{3.38} M_1'(t)\leq -k\xi(t)\Big(\frac{E(t)}{E(0)}\Big)H_1' \Big(\epsilon_0\frac{E(t)}{E(0)}\Big)=-k\xi(t)H_0\Big(\frac{E(t)}{E(0)}\Big), \end{equation} $ (3.38)

其中$H_0(t)=tH_1'(\epsilon_0t).$$H_1$的定义及$h_0\in C^1([0, \infty)$$H_1'(0)=h_0'(0), $这表明$H_0(0)=0.$并且, 由$H_0'(t)=H_1(\epsilon_0t) +\epsilon_0tH_1''(\epsilon_0t)$$H_1$$(0, r^2]$上严格凸, 得$H_0'(t), H_0(t)>0$, $t\in (0, 1]$.

$M_2(t)=\frac{\alpha_1 M_1(t)}{E(0)}$, 联立(3.35)式, 则有$M_2(t)\sim E(t).$由(3.38)式知存在$k_2>0$使得

$ \begin{array}{l} M_2'(t)\leq -k_2\xi(t)H_0(M_2(t)), ~~~\forall t\geq t_0. \end{array} $

积分得存在$k_3>0$使得

$ \begin{equation}\label{3.39} M_2(t)\leq H^{-1}\bigg(k_2\int_{t_0}^t\xi(s){\rm d}s+k_3\bigg), ~~~\forall t\geq t_0, \end{equation} $ (3.39)

其中$H(t)=\int_t^1\frac{1}{H_0(s)}{\rm d}s.$这里用到$H_0$的性质及$H$$(0, 1]$上严格递减.联立(3.39)式及$E$$\xi$的有界性和连续性, 结论(3.22)得证.

注3.2   (1)举出如下例子来说明定理3.1给出的能量衰减估计.

$h_0(s)=cs^p$, $c>0$, $p\geq1, $$H_1(s)=cs^{\frac{p+1}{2}}$是凸函数.因此, 估计式(3.22)表明

$ \begin{array}{l} E(t)\leq k_1{\rm e}^{-k_2\int_{t_0}^t\xi(s){\rm d}s}, ~~~p=1; \quad E(t)\leq k_1\bigg(k_2\int_{t_0}^t\xi(s){\rm d}s+k_3\bigg)^{-\frac{2}{p-1}}, p>1. \end{array} $

$h_0(s)=\frac{1}{s}{\rm e}^{-1/s^2}$, 则$H_1(s)={\rm e}^{-1/s}$是一个凸函数.因此, 估计式(3.22)表明

$ E(t)\leq k_1\bigg(\ln\Big(k_2\int_0^t\xi(s){\rm d}s+k_2\Big)\bigg)^{-1}. $

关于$h_0(s)$更多的例子, 参见文献[1, 5, 21, 34].

(2) 当$\mu_2=0$时, 问题(1.1)类似于文献[34].条件(2.9)表明当时滞项被耗散项控制时, 我们具有相同的能量衰减估计.特别地, Yang[36]考虑了如下方程

$ u_{tt}(x, t)+\triangle^2u(x, t)-\int_0^tg(t-s)\triangle^2u(x, s){\rm d}s +\mu_1u_t(x, t)+\mu_2u_t(x, t-\tau)=0. $

(3) 若不考虑记忆项(即(A2)中$ l=1$), 类似可得到相同的能量衰减估计式.比如文献[5]中讨论了如下方程

$ \begin{array}{l} u_{tt}(x, t)-\triangle u(x, t)+\mu_1\sigma(t)g_1(u_t(x, t))+\mu_2\sigma(t)g_2(u_t(x, t-\tau(t)))=0. \end{array} $

利用乘子技巧他们得到了解的渐近估计.

(4) 定理3.1的结论可以运用到具体的实例中.设$g$满足条件(A2), 耗散和时滞项满足条件(A4)-(A6).

$ \begin{array}{l} \left\{\begin{array}{ll} u_{tt}(t)-\triangle u(t)+\int_{0}^{t}g(t-s)\triangle u(s){\rm d}s+\mu_1h_1(u_t(t)) \\ \ \;+\mu_2h_2(u_t(x, t-\tau(t)))=|u|^pu, &(x, t)\in\Omega\times(0, \infty), \\ u(x, t)=0, &(x, t)\in \partial\Omega\times(0, \infty), \\ u(x, 0)=u_0(x), \, \, \, u_t(x, 0)=u_1(x), &x\in \Omega, \\ u_t(x, t)=f_0(x, t), &(x, t)\in\Omega\times[-\tau(0), 0), \end{array}\right. \end{array} $

其中$0 <p\leq \frac{4}{n-2}$, $n\geq3$; $p>0$, $n=1, 2$.此问题可以改写成抽象方程(1.1), 其中$A=-\triangle$, $D(A)=H^2(\Omega)\cap H_0^1(\Omega)$, $D(\sqrt{A})=H_0^1(\Omega)$.令

$ F(x)=\frac{1}{p+2}\int_{\Omega}|x(s)|^{p+2}{\rm d}s, $

其中$x\in H_0^1(\Omega)$, 类似于文献[1], 可知定理3.1关于$F$的假设满足.因此, 能量估计(3.22)式满足.

参考文献
[1] Alabau-Boussouira F, Cannarsa P., Sforza D. Decay estimates for second order evolution equations with memory. J Funct Anal, 2008, 254(5): 1342–1372. DOI:10.1016/j.jfa.2007.09.012
[2] Ammari K, Nicaise S, Pignotti C. Stability of an abstract-wave equation with delay and a Kelvin-Voigt damping. Asymptot Anal, 2015, 95: 21–38. DOI:10.3233/ASY-151317
[3] Arnold V L. Mathematical Methods of Classical Mechanics. New York: Springer, 1989.
[4] Berrimi S, Messaoudi S A. Existence and decay of solutions of a viscoelastic equation with a nonlinear source. Nonlinear Anal, 2006, 64(10): 2314–2331. DOI:10.1016/j.na.2005.08.015
[5] Benaissa A, Benaissa A, Messaoudi S A. Global existence and energy decay of solutions for the wave equation with a time varying delay term in the weakly nonlinear internal feedbacks. J Math Phys, 2012, 53(12): 123514. DOI:10.1063/1.4765046
[6] Cavalcanti M M, Domingos Cavalcanti V N, Soriano J A. Exponential decay for the solution of semilinear viscoelastic wave equations with localized damping. Electron J Differ Eq, 2002, 2002(44): 1–14.
[7] Cavalcanti M M, Domingos Cavalcanti V N, Martinez P. General decay rate estimates for viscoelastic dissipative systems. Nonlinear Anal TMA, 2008, 68(1): 177–193. DOI:10.1016/j.na.2006.10.040
[8] Chen H, Liu G W. Well-posedness for a class of Kirchhoff equations with damping and memory terms. IMA J Appl Math, 2015, 80(6): 1808–1836.
[9] Chen M M, Liu W J, Zhou W C. Existence and general stabilization of the Timoshenko system of thermoviscoelasticity of type Ⅲ with frictional damping and delay terms. Advances in Nonlinear Analysis, 2016. DOI:10.1515/anona-2016-0085
[10] Dafermos C M. Asymptotic stability in viscoelasticity. Arch Ration Mech Anal, 1970, 37(4): 297–308. DOI:10.1007/BF00251609
[11] Dai Q Y, Yang Z F. Global existence and exponential decay of the solution for a viscoelastic wave equation with a delay. Z Angew Math Phys, 2014, 65(5): 885–903. DOI:10.1007/s00033-013-0365-6
[12] Datko R, Lagnese J, Polis M P. An example on the effect of time delays in boundary feedback stabilization of wave equations. SIAM J Control Optim, 1986, 24(1): 152–156. DOI:10.1137/0324007
[13] Datko R. Not all feedback stabilized hyperbolic systems are robust with respect to small time delays in their feedbacks. SIAM J Control Optim, 1988, 26(3): 697–713. DOI:10.1137/0326040
[14] Guesmia A. Well-posedness and exponential stability of an abstract evolution equation with infinity memory and time delay. IMA J Math Control Inform, 2013, 30(4): 507–526. DOI:10.1093/imamci/dns039
[15] Kirane M, Said-Houari B. Existence and asymptotic stability of a viscoelastic wave equation with a delay. Z Angew Math Phys, 2011, 62(6): 1065–1082. DOI:10.1007/s00033-011-0145-0
[16] Liu G W, Zhang H W. Blow up at infinity of solutions for integro-differential equation. Appl Math Comput, 2014, 230(2): 303–314.
[17] Liu G W, Zhang H W. Well-posedness for a class of wave equation with past history and a delay. Z Angew Math Phys, 2016, 67(1): 1–14.
[18] Liu W J. General decay of the solution for a viscoelastic wave equation with a time-varying delay term in the internal feedback. J Math Phys, 2013, 54(4): 043504. DOI:10.1063/1.4799929
[19] Liu W J. General decay rate estimate for the energy of a weak viscoelastic equation with an internal time-varying delay term. Taiwanese J Math, 2013, 17(6): 2101–2115. DOI:10.11650/tjm.17.2013.2968
[20] Messaoudi S A, Mustafa M I. On the control of solutions of viscoelastic equations with boundary feedback. Nonlinear Anal RWA, 2009, 10(5): 3132–3140. DOI:10.1016/j.nonrwa.2008.10.026
[21] Messaoudi S A, Mustafa M I. On convexity for energy decay rates of a viscoelastic equation with boundary feedback. Nonlinear Anal TMA, 2010, 72(9/10): 3602–3611.
[22] Messaoudi S A, Fareh A, Doudi N. Well posedness and exponential satbility in a wave equation with a strong damping and a strong delay. J Math Phys, 2016, 57(11): 111501. DOI:10.1063/1.4966551
[23] Mustafa M I. Asymptotic behavior of second sound thermoelasticity with internal time-varying dealy. Z Angew Math Phys, 2013, 64(4): 1353–1362. DOI:10.1007/s00033-012-0268-y
[24] Nicaise S, Pignotti C. Stability and instability results of the wave equation with a delay term in the boundary or internal feedbacks. SIAM J Control Optim, 2006, 45(5): 1561–1585. DOI:10.1137/060648891
[25] Nicaise S, Pignotti C. Stabilization of the wave equation with boundary or internal distributed delay. Differential Integral Equations, 2008, 21(9/10): 935–958.
[26] Nicaise S, Valein J, Fridman E. Stabilization of the heat and the wave equations with boundary timevarying delays. Discrete Contin Dyn Syst Ser S, 2009, 2(3): 559–581. DOI:10.3934/dcdss
[27] Nicaise S, Pignotti C. Interior feedback stabilization of wave equations with time dependent delay. Electron J Differ Eq, 2011, 2011(41): 1–20.
[28] Nicaise S, Pignotti C. Stability for second-order evolution equations with switching time-delay. J Dyn Diff Eq, 2014, 26(3): 781–803. DOI:10.1007/s10884-014-9382-1
[29] Nicaise S, Pignotti C. Exponential stability of abstract evolution equations with time delay. J Evol Eq, 2015, 15(1): 107–129. DOI:10.1007/s00028-014-0251-5
[30] Nicaise S, Pignotti C. Stability of the wave equation with localized Kelvin-Voigt damping and boundary delay feedback. Discrete Contin Dyn Syst Ser S, 2016, 9(3): 791–813. DOI:10.3934/dcdss
[31] Pignotti C. Stability for second-order evolution equations with memory and switching time-delay. J Dyn Diff Eq, 2016. DOI:10.1007/s10884-016-9545-3
[32] Park S H. Energy decay for a von Karman equation with time-varying delay. Applied Mathematics Letters, 2016, 55(5): 10–17.
[33] Peralta G R. Stabilization of viscoelastic wave equations with distributed or boundary delay. Z Anal Anwend, 2016, 35(3): 359–381. DOI:10.4171/ZAA
[34] Wu S T. Energy decay rates via convexity for some second-order evolution equation with memory and nonlinear time-dependent dissipation. Nonlinear Anal TMA, 2011, 74(2): 532–543. DOI:10.1016/j.na.2010.09.007
[35] Xu G Q, Yung S P, Li L K. Stabilization of wave systems with input delay in the boundary control. ESAIM Control Optim Calc Var, 2006, 12(4): 770–785. DOI:10.1051/cocv:2006021
[36] Yang Z F. Existence and energy decay of solutions for the Euler-Bernoulli viscoelastic equation with a delay. Z Angew Math Phys, 2015, 66(3): 727–745. DOI:10.1007/s00033-014-0429-2