数学物理学报  2016, Vol. 36 Issue (6): 1124-1136   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
张宏伟
刘功伟
呼青英
一类具对数源项波动方程的初边值问题
张宏伟, 刘功伟, 呼青英     
河南工业大学数学系 郑州 450001
摘要:该文考虑一类具对数源项波动方程的初边值问题.利用Galerkin方法结合对数Sobolev不等式和对数Gronwall不等式,对所有初始值得到了整体解的存在性.通过引入位势井,给出了解在时间无穷远处爆破(即指数增长)的充分条件.当具有对数源项的波动方程还带有线性阻尼时,通过构造适当的Lyapunov函数,得到了能量的衰减估计.
关键词对数波动方程     整体存在性     指数增长     衰减    
Initial Boundary Value Problem for a Class Wave Equation with Logarithmic Source Term
Zhang Hongwei, Liu Gongwei, Hu Qingying     
Department of Mathematics, Henan University of Technology, Zhengzhou 450001
Abstract: In this paper we consider the initial boundary value problem for a class wave equation with logarithmic source term. By using Galerkin method combining with the logarithmic Sobolev inequality and logarithmic Gronwall inequality, we obtain the existence of global solution for all initial data. By introducing potential well theory, we give the sufficient condition of the blow-up property in infinity time (i.e. exponential growth) of the solution. By constructing an appropriate Lyapunov function, we obtain the decay estimates of energy for the wave equation with logarithmic source term and linear damping term.
Key words: Logarithmic wave equation     Existence of global solution     Exponential growth     Decay    
1 引言

本文研究如下初边值问题

$ \begin{equation} u_{tt}- u_{xx} - u_{xxtt}+ u_{xxxx} +(f(u_x))_{x}=0, x \in \Omega, t > 0, \label{01.1} \end{equation} $ (1.1)
$ \begin{equation} u(0, t)= u(l, t)=0, u_{x}(0, t)=u_{x}(l, t)=0, t \ge 0, \label{01.2} \end{equation} $ (1.2)
$ \begin{equation} u(x, 0) = u_0(x), u_t (x, 0) = u_1(x), x \in \Omega, \label{01.3} \end{equation} $ (1.3)

其中$f(s)=s \ln|s|^k, k\ge 1, \Omega=(0, l), u(x, t)$是未知函数, 下标$x$$t$分别记关于$x$$t$的偏导数, $u_0(x)$$u_1(x)$为给定的初始值函数并满足边界条件(1.2).

方程(1.1)源于多个不同领域, 在过去的五十多年里, 关于初边值问题(1.1)-(1.3)解的整体存在性和不存在性已有许多结果.由文献[1-2]知, 方程(1.1)可以变换为如下拟线性波动方程(至少在单连通$(x, t)$空间)

$ \begin{equation} u_{tt}- u_{xx} - u_{xxtt}+ u_{xxxx} = f(u)_{xx}. \label{01.4} \end{equation} $ (1.4)

带有非线性项$f(u)=u \ln|u|^k$的方程(1.4)由Wazwaz[3]首次引入并称为广义对数修正(或正则) Boussinesq方程(log-RBE), 它是在基于如下的对数-KdV (log-KdV)方程

$ v_{t}+ v_{xxx} + (v\log|v|)_{xx}=0 $

而分析Gaussian孤立波解时提出的. Boussinesq -型方程(或修正Boussinesq -型方程)

$ \begin{equation} u_{tt}- u_{xx} - u_{xxtt}+a u_{xxxx} = f(u)_{xx}, a\ge 0 \end{equation} $ (1.5)

是物理学和流体力学中的一类重要模型, 在过去的几十年, 它吸引了众多数学和物理工作者(参见文献[4-16]及其参考文献).

另一方面, 在研究非线性弹性杆的强孤立波时, 引入了纵向波动方程[17-18]

$ \begin{equation} u_{tt}- [a_0+na_1(u_x)^{n-1}]u_{xx} - u_{xxtt}= 0, \end{equation} $ (1.6)

其中$a_0$是正常数, $a_1$是任意实数, $n$是自然数.对更一般方程

$ \begin{equation} u_{tt}- u_{xx} - u_{xxtt}+ au_{xxxx} +(f(u_x))_{x}=0, a\ge 0 \end{equation} $ (1.7)

的适定性有许多学者已进行了研究(见文献[7, 9, 14, 19-27]及其参考文献).

当方程(1.7) (或(1.5))带有不同的非线性项, 如$f (s) = \pm a|s|^p$, $f(s) = a|s|^ps, p > 1$或满足一定的符号性或另外的结构性增长条件时, 已有许多研究结果.在上述提到的文献中, 非线性项要满足的如下条件

$ \begin{equation} 0 < F(s) = \int ^s_0 f(t){\rm d}t\le \frac{1}{p+\theta}f(s)s, \end{equation} $ (1.8)

其中$\theta$是某一正常数, 这是非常重要的条件而且在用变分法证明方程(1.7)的弱解存在性时被广泛应用.最近, Zhang[28], Kutev等[29]和Xu[30]还研究了方程(1.5)带有组合幂型非线性项$f(s)=\sum \limits_{k=1}^m a_k|s|^ps $或指数型非线性项$f(s)={\rm e}^{ks^2}$的Cauchy问题.然而, 非常有趣的非线性项$f(s)=s\ln|s|^k$却不能包含在上述文献中, 因为它不满足条件$F(s)\ge d_1 |s|^{\nu}-d_2, s \in{\Bbb R}, d_1, d_2 > 0, \nu >p$, 这个条件是条件(1.8)的一个推论.

受文献[31-34]的启发, 我们研究问题(1.1)-(1.3).文献[31-32]的作者研究了下面对数非线性的热方程的初边值问题

$ \begin{equation} u_{t}-k \Delta u_t- \Delta u = u\log|u|, x \in \Omega, t > 0, \end{equation} $ (1.9)

其中$k\ge 0$, $\Omega$${\Bbb R}^N$中带有光滑边界的有界区域.他们得到了解的存在性, 衰减估计和无穷时间爆破.文献[33-34]的作者研究了对数Klein-Gordon方程

$ \begin{equation} u_{tt}- \Delta u+u=k u\log|u|^2, \end{equation} $ (1.10)

得到了整体弱解的存在性.

本文研究初边值问题(1.1)-(1.3).据作者所知, 这是首次研究对数波动方程整体解的存在性, 衰减估计和爆破问题.为处理对数非线性项$u\log|u|$, 我们引入对数Sobolev不等式[35-36]和对数Gronwall不等式[33], 利用Galerkin方法得到了整体解对所有初始值存在.通过引入位势井理论[37], 给出了在适当条件下解在时间无穷远处爆破的充分条件, 即当时间趋于无穷时, 解在某范数下的指数增长性.这里的主要困难是文献[37-38]中经典的位势井理论并不完全适用于对数非线性项, 因为非线性函数$f(u) =u \log|u|$不满足文献[37-38]中对$f(u)$要求的条件.因此, 我们修正经典的凸性方法得到解在时间无穷远处爆破的条件.这个结果表明半线性波动方程中多项式非线性是解在有限时间爆破的最优条件.另外, 在波动方程具有对数源项和线性阻尼时, 我们还得到了当初始值在适当的稳定集时能量的衰减估计, 其证明的关键思想源于文献Haraux和Zuazua[39]及Zuazua[40], 这个思想最近还被Benaissa和Messaoudi[41]及Gerbi和Said-Houari[42]使用过.这个方法主要是通过构造适当的Lyapunov函数以及能量的小扰动结合位势井理论得到结果.我们发现在对数源项时其证明更加困难, 而且稳定集也不同于文献[42]中所给出的稳定集.

本文的结构如下:第2节, 我们引入了本文所需要的一些数学符号和重要引理; 第3节, 利用Galerkin方法得到了整体解的存在性.在第4节中, 证明了解在时间无穷远处爆破的条件.第5节, 在方程(1.1)还具有线性阻尼时, 得到了能量的衰减估计.

2 预备知识

本文把$L^p(\Omega)$记为$L^p$, 记$H^s(\Omega)$$H^s$, 记$H_0^s(\Omega)$$H_0^s$, 记$\| .\| _p=\| .\| _{L^p(\Omega)}, ~ \| .\| =\| .\| _{L^2(\Omega)}$以及$(u, v)=\int _ {\Omega} u(x)v(x){\rm d}x$.

$u(x, t)$为问题(1.1)-(1.3)在$\Omega \times [0, T)$上的弱解, 如果$u \in L^{\infty}(0, T; H^2_0), ~u_t \in L^{\infty}(0, T ; H^1)$在分布意义下满足方程(1.1), 即

$ \begin{eqnarray}& (u_{t}, w) +(u_{xt}, w_{x})+\int ^t_0[(u_x, w_x)+(u_{xx}, w_{xx})]{\rm d}s -\int ^t_0( u_x\log|u_x|^k, w_x){\rm d}s\\ &=(u_{1}, w)+(u_{1x}, w_{x}), ~~\forall w \in H^2_0, ~\forall t \in [0, T), \end{eqnarray} $ (2.1)
$ \begin{equation} u(x, 0) = u_0(x), ~\mbox{在 $H^2_0$ 中; $u_t(x, 0) = u_1(x)$在$L^2$ 中} \end{equation} $ (2.2)

且满足如下能量不等式

$ \begin{equation} E(t) \le E(0), ~~\forall t \in [0, T ), \label{02.3} \end{equation} $ (2.3)

其中

$ E(t) =E(u)=\frac{1}{2}\bigg(\| u_{t}\| ^2+\| u_{xt}\| ^2+\| u_x\| ^2+\| u_{xx}\| ^2 -\int _0^l u^2_x\log|u_x|^k{\rm d}x\bigg)+\frac{k}{4}\| u_x\| ^2, $
$ E(0) =\frac{1}{2}\bigg(\| u_{1}\| ^2+\| u_{1x}\| ^2+\| u_{0x}\| ^2+\| u_{0xx}\| ^2-\int _0^l u^2_{0x}\log|u_{0x}|^k{\rm d}x\bigg)+\frac{k}{4}\| u_{0x}\| ^2. $

本文将多次使用如下的对数Sobolev不等式[31, 35-36].

引理2.1[31, 35-36](对数Sobolev不等式)  对$H^1_0(\Omega)$中任一函数$u$和任意实数$a > 0$, 则

$ \begin{equation} 2\int _{\Omega} |u|^2 \log \frac {|u|}{\| u\| }{\rm d}x+(1+\log a)\| u\| ^2 \le \frac{a^2}{\pi}\int _{\Omega} |u_x|^2{\rm d}x. \end{equation} $ (2.4)

引理2.2[33] (对数Gronwall不等式)  设$\phi (t)$是非负函数, $\phi (t) \in L^\infty (0, T ), \phi(0) \ge 0$, 且满足

$ \phi(t) \le \phi(0) + a \int^t_0 \phi(s) \log[a +\phi(s)]{\rm d}s, t \in [0, T], $

其中$a > 1$为正常数, 则

$ \phi(t) \le (a + \phi(0)){\rm e}^{at}, t\in[0, T]. $

现在, 引入两个泛函$J(u)$$I(u)$:

$ \begin{equation} J(u)=J(u(t))=J(t)=\frac{1}{2} \bigg(\| u_x\| ^2+\| u_{xx}\| ^2-\int _0^l u^2_x\log|u_x|^k{\rm d}x\bigg)+\frac{k}{4}\| u_x\| ^2, \end{equation} $ (2.5)
$ \begin{equation} I(u)=I(u(t))=I(t)=\| u_x\| ^2+\| u_{xx}\| ^2-\int _0^l u^2_x\log|u_x|^k{\rm d}x. \end{equation} $ (2.6)

显然有

$ \begin{equation} J(u)=\frac{1}{2}I(u)+\frac{k}{4}\| u_x\| ^2, \end{equation} $ (2.7)
$ \begin{equation} E(t)=\frac{1}{2}(\| u_t\| ^2+\| u_{xt}\| ^2)+J(u). \end{equation} $ (2.8)

按照对数Sobolev不等式引理2.1, $J(u)$$I(u)$$H^2_0$上适定.

引理2.3  对任意$u \in H^2_0\backslash \{0\}$, 则

(ⅰ) $I(\lambda u)=\lambda \frac{\rm d}{{\rm d}\lambda}J(\lambda u)$$\lim\limits_{\lambda \rightarrow 0}J(\lambda u)=0$, $\lim\limits_{\lambda \rightarrow + \infty}J(\lambda u)=- \infty$;

(ⅱ)存在唯一的$\lambda^{*}=\lambda^{*}(u)$使得$\frac{\rm d}{{\rm d}\lambda}J(\lambda u)|_{\lambda=\lambda^*}=0$$ J(\lambda u)$$0 < \lambda \le \lambda^*$上单调增加, 在$\lambda^* \le \lambda < \infty$上单调减少, 并在$\lambda =\lambda^{*}$取得极大值.换句话说, 存在唯一的$\lambda^* \in (0, +\infty)$使得

$ \begin{equation} I(\lambda u)=\lambda \frac{\rm d}{{\rm d}\lambda}J(\lambda u) > 0, 0 < \lambda < \lambda^*;I(\lambda u)=0, \lambda =\lambda^*; I(\lambda u) < 0, \lambda > \lambda^*, \label{02.9} \end{equation} $ (2.9)

这里$\lambda^*=\exp\big(\frac{\| u_x\| ^2+\| u_{xx}\| ^2-\int _0^l u^2_x\log|u_x|^k{\rm d}x}{k\| u_x\| ^2}\big)$.

 显然有

$ \begin{eqnarray*} J(\lambda u)&=&\frac{1}{2}\lambda ^2 (\| u_x\| ^2+\| u_{xx}\| ^2)+\frac{k}{4}\lambda ^2 \| u_x\| ^2 -\frac {1}{2}\int _0^l \lambda ^2u^2_x\log|\lambda u_x|^k{\rm d}x\\ &=&\frac{1}{2}\lambda ^2 (\| u_x\| ^2+\| u_{xx}\| ^2)+\frac{k}{4}\lambda ^2 \| u_x\| ^2 -\frac {k}{2}\lambda ^2\int _0^l u^2_x\log|u_x|{\rm d}x-\frac {k}{2}\lambda ^2(\log|\lambda |)\| u_x\| ^2 \end{eqnarray*} $

以及

$ \begin{eqnarray*} \frac{\rm d}{{\rm d}\lambda }J(\lambda u)&=&\lambda (\| u_x\| ^2+\| u_{xx}\| ^2) -\lambda \int _0^l u^2_x\log| u_x|^k{\rm d}x- \lambda (\log|\lambda |^k)\| u_x\| ^2\\ &=&\lambda (\| u_x\| ^2+\| u_{xx}\| ^2) -\lambda \int _0^l u^2_x\log|\lambda u_x|^k{\rm d}x. \end{eqnarray*} $

则因为$\| u_x\| \ne 0 $$I(\lambda u)=\lambda \frac{\rm d}{{\rm d}\lambda}J(\lambda u)$$\lim\limits_{\lambda \rightarrow 0}J(\lambda u)=0$以及$\lim\limits_{\lambda \rightarrow + \infty}J(\lambda u)=- \infty$.另外, 还有

$ I(\lambda u) > 0, 0 < \lambda < \lambda^*;I(\lambda u)=0, \lambda =\lambda^*; I(\lambda u) < 0, \lambda > \lambda^*, $

这里$\lambda^*=\exp\big(\frac{\| u_x\| ^2+\| u_{xx}\| ^2-\int _0^l u^2_x\log|u_x|^k{\rm d}x}{k\| u_x\| ^2}\big)$, $J(\lambda u)$$0 < \lambda < \lambda^*$上单调增, 在$\lambda > \lambda^*$上单调减且在$\lambda =\lambda^*$取得极大值.

下面引入熟知的Nehari流形和井深$d$

$ \begin{equation} N=\{u|u \in H^2_0\backslash \{0\}, I(u)=0\}, \end{equation} $ (2.10)
$ \begin{equation} d=\inf \{\sup_{\lambda \ge 0} J(\lambda u)| u \in H^2_0\backslash \{0\}\}. \end{equation} $ (2.11)

则由引理2.3知$0 < d=\inf\limits_{u \in N}J(u)$.再在$H_0^2$中定义与问题(1.1)-(1.3)相关的两个子集.令

$ \begin{equation} W=\{u \in H^2_0|~J(u) < d, I(u) > 0\}\cup \{0\}, \end{equation} $ (2.12)
$ \begin{equation} V=\{u \in H^2_0|~J(u) < d, I(u) < 0\}. \end{equation} $ (2.13)

引理2.4  令$u \in H^2_0 $$r=(\frac {2\pi}{k})^{\frac{1}{4}}e^\frac {1}{2}$,

(ⅰ)  若$0 < \| u_x\| < r$, 则$I(u) > 0$;

(ⅱ)  若$I(u) < 0$, 则$\| u_x\| > r$;

(ⅲ)  若$I(u)=0$$\| u_x\| \ne 0$, 即$v \in N$, 则$\| u_x\| > r$.

 由对数Sobolev不等式, 对任意$a > 0$, 得

$ \begin{eqnarray} I(u)&=& (\| u_x\| ^2+\| u_{xx}\| ^2)-k \int _0^l u^2_x \Big(\log \frac {|u_x|}{\| u_x\| }+\log\| u_x\| \Big){\rm d}x \\ & \ge &\Big( 1-\frac {ka^2}{2\pi}\Big)(\| u_x\| ^2+\| u_{xx}\| ^2)+\frac{k(1+\log a)}{2}\| u_x\| ^2-k\| u_x\| ^2\log\| u_x\| .\label{02.14} \end{eqnarray} $ (2.14)

在(2.14)式中取$a=\sqrt\frac {2\pi }{k}$, 得

$ \begin{equation} I(u) \ge k\Big(\frac{2+\log\frac {2\pi }{k}}{4}-\log\| u_x\|\Big )\| u_x\| ^2. \end{equation} $ (2.15)

$0 < \| u_x\| < r$, 则由(2.15)式得$I(u) > 0$.若$I(u) < 0$, 则利用(2.15)式知

$ \frac{2+\log\frac {2\pi }{k}}{4}-\log\| u_x\| > 0, $

这意味着$\| u_x\| >(\frac {2\pi }{k})^{\frac{1}{4}}e^\frac {1}{2}=r$.若$I(u)=0$$\| u_x\| \ne 0$, 则利用(2.15)式, 得

$ \begin{equation} \log\| u_x\| \ge \frac{2+\log(\frac {2\pi }{k})}{4}, \end{equation} $ (2.16)

$\| u_x\| >r$.

引理2.5  (ⅰ) $ d \ge \frac{k}{4}( \frac {2\pi }{k})^ \frac{1}{2} e=\frac{k}{4}r^2$;

(ⅱ)  若$u \in H^2_0$$I(u) < 0$, 则

$ \begin{equation} I(u) < 2(J(u)-d).\label{02.17} \end{equation} $ (2.17)

(ⅰ)  若$I(u)=0$$\| u_x\| \ne 0$, 则由引理2.4得$\| u_x\| \ge r=(\frac {2\pi }{k})^ \frac{1}{4} e^\frac {1}{2}$.结合(2.7)式, 得

$ J(u)=\frac{1}{2}I(u)+\frac{k}{4}\| u_x\| ^2\ge \frac{k}{4} \Big( \frac {2\pi}{k}\Big)^ \frac{1}{2} e, $

则有$ d \ge \frac{k}{4}(\frac {2\pi }{k})^ \frac{1}{2} e$.

(ⅱ)  若$u \in H^2_0$$I(u)< 0$, 则利用引理2.2可得, 必存在$\lambda^* $使得$0 < \lambda^* < 1$$I(\lambda ^* u)=0$.由$d$的定义(2.11)知

$ d \le J(\lambda ^* u)=\frac{1}{2}I(\lambda^* u)+\frac{k}{4}\| \lambda^* u_x\| ^2 = \frac{k}{4}\| \lambda^* u_x\| ^2 < \frac{k}{4}\| u_x\| ^2. $

由此和$J(u)$$I(u)$的定义, 得

$ d < \frac{k}{4}\| u_x\| ^2 = J(u)- \frac {1}{2}I(u), $

$I(u)< 2(J(u)-d)$.

引理2.6  若$u_0 \in H^2_0, u_1 \in H^1$$0 < E(0) < d$, $u$是问题(1.1)-(1.3)的弱解, 则

(ⅰ)  若$I(u_0) > 0$, 则$u \in W$;

(ⅱ)  若$I(u_0) < 0$, 则$u \in V$.

 由弱解定义和(2.3)式, 有

$ \begin{eqnarray} & \frac{1}{2}(\| u_t\| ^2+\| u_{xt}\| ^2)+J(u)\le \frac{1}{2}(\| u_1\| ^2+\| u_{1x}\| ^2)+J(u_0) < d, ~\forall t \in [0, T). \end{eqnarray} $ (2.18)

(ⅰ)  我们断言$u(t)\in W$对所有$t \in [0, T)$成立.否则, 必存在$t_0 \in (0, T)$使得$u(t_0) \in \partial W$, 于是

(a) $I(u(t_0)) =0$$\| u_x(t_0)\| \neq 0$, 或(b) $J(u(t_0)) =d$.

由(2.18)式, (b)是不可能的, 于是只可能$I(u(t_0)) =0$$\| u_x(t_0)\| \neq 0$.然而, 由$d$的定义知$J(u(t_0)) \ge d$, 这与(2.18)式矛盾.从而, 对任意$t \in [0, T)$$u(t)\in W$.

(ⅱ)  若存在$t_0 \in (0, T)$使得对任意的$0\le t < t_0$$u(x, t) \in V$$u(x, t_0) \in \partial V$, 则

(c) $I(u(t_0)) =0$, 或(d) $J(u(t_0)) =d$.

由(2.18)式, (d)不成立.若$I(u(t_0)) =0$且对$0 < t < t_0$$I(u(t)) < 0$, 则由引理2.4(ⅱ), 对$0 < t < t_0$, 有$\| u_x\| > r$$0 < t\le t_0$成立.于是, 由$d$的定义得$J(u(t_0)) \ge d$, 这与(2.18)式矛盾.这意味着对所有$t \in [0, T)$, 有$u(t)\in V$成立.

3 整体解的存在性

本节研究问题(1.1)-(1.3)整体解的存在性, 主要是基于Galerkin方法, 证明过程中要用到对数Sobolev不等式和对数Gronwall不等式.

定理3.1  设$u_0 \in H^2_0, u_1 \in H^1$, 则问题(1.1)-(1.3)存在整体解$u \in L^{\infty}(0, \infty; H^2_0)$$u_t \in L^{\infty} (0, \infty; L^2)$.

 用Galerkin方法来证明.设$\{w_j\}^{\infty}_{j=1}$是空间$H^2_0$的一组标准正交基.构造问题(1.1)-(1.3)的近似解如下

$ \begin{equation} u_m(x, t) =\sum_{j=1}^m g_{jm}(t)w_j(x), ~~m = 1, 2, \cdots, \end{equation} $ (3.1)

它满足

$ \begin{eqnarray}& (u_{mtt}, w_s)+(u_{mx}, w_{sx})+(u_{mxtt}, w_{sx})+ (u_{mxx}, w_{sxx})-(u_{mx}\log|u_{mx}|^k, w_{sx})=0, \\ &s=1, 2, \cdots, m, \end{eqnarray} $ (3.2)
$ \begin{equation} u _m(0)=u_m(x, 0) =\sum_{j=1}^m a_{jm}(t)w_j(x) \rightarrow u_0(x), ~~\mbox{在}~~H^2_0\mbox{中}, \end{equation} $ (3.3)
$ \begin{equation} u_{mt}(0)=u_{mt}(x, 0) =\sum_{j=1}^m b_{jm}(t)w_j(x) \rightarrow u_1(x), ~~\mbox{在}~~L^2\mbox{中}. \end{equation} $ (3.4)

方程(3.2)两边同乘$g'_{sm}(t)$然后关于$s$求和, 得

$ \begin{equation} \frac{\rm d}{{\rm d}t}E_m(t)=0 \end{equation} $ (3.5)

$ \begin{equation} E_m(t)= E_m(0), \end{equation} $ (3.6)

这里

$ \begin{eqnarray*} E_m(t)&=&E(u_m)\\ &=&\frac{1}{2}\Big(\| u_{mt}\| ^2+\| u_{mxt}\| ^2+\| u_{mx}\| ^2+\| u_{mxx}\| ^2 -\int _0^l u^2_{mx}\log|u_{mx}|^k{\rm d}x\Big)+\frac{k}{4}\| u_{mx}\| ^2. \end{eqnarray*} $

因在空间$H^2_0$$u_m(0)\rightarrow u_0$, 得序列$u_m(0)$$H^2_0$中有界.由基本不等式$|t^2\log t|\le C(1+t^3), $ $t>0$, 利用Sobolev嵌入定理$H^2_0 \hookrightarrow L^\infty$, 有

$ \int _0^l u^2_{0mx}\log|u_{0mx}|^k{\rm d}x \le C(1+\| u_{0m}\| ^3_{H^2_0})\le C_0, $

这里$C_0 = C(\| u_0\| _{H^2}, \| u_1\| )$是正常数.从而, 由(3.6)式得

$ \begin{equation} \| u_{mt}\| ^2+\| u_{mxt}\| ^2+\| u_{mx}\| ^2+\| u_{mxx}\| ^2 \le C+2\int _0^l u^2_{mx}\log|u_{mx}|^k{\rm d}x. \end{equation} $ (3.7)

现在利用Gross对数Sobolev不等式引理2.1估计(3.7)式的右边最后一项如下

$ \begin{eqnarray} 2\int _0^l u^2_{mx}\log|u_{mx}|^k{\rm d}x&=&2k \int _0^l u^2_{mx} \Big(\log \frac {|u_x|}{\| u_{mx}\| }+\log\| u_{mx}\|\Big ){\rm d}x\\ &\le& k\Big[\frac {a^2}{\pi}\| u_{mxx}\| ^2-(1+\log a)\| u_{mx}\| ^2+\| u_{mx}\| ^2\log\| u_{mx}\| ^2\Big]. \end{eqnarray} $ (3.8)

把({3.8})式带入到(3.7)式, 得

$ \begin{eqnarray} &&\| u_{mt}\| ^2+\| u_{mxt}\| ^2+(1+k+k\log a)\| u_{mx}\| ^2+ \Big(1-\frac {ka^2}{\pi}\Big)\| u_{mxx}\| ^2\\ &\le& C(1+\| u_{mx}\| ^2\log\| u_{mx}\| ^2). \end{eqnarray} $ (3.9)

在({3.9})式中取$a = \sqrt\frac {\pi }{2k}$

$ \begin{equation} \| u_{mt}\| ^2+\| u_{mxt}\| ^2+\| u_{mx}\| ^2+\| u_{mxx}\| ^2\le C(1+\| u_x\| ^2\log\| u_{mx}\| ^2). \end{equation} $ (3.10)

注意到

$ u_{mx}(t)=u_{mx}(0)+\int^t_0u_{mxt}(s){\rm d}s, $

$ \begin{eqnarray} \| u_{mx}\| ^2&\le& 2\| u_{mx}(0)\| ^2+2T \int^t_0\| u_{mxt}\| ^2(s){\rm d}s\\ &\le& 2\| u_{mx}(0)\| ^2+\max\{1, 2T\}\frac{1+C}{C} \int^t_0\| u_{mxt}\| ^2(s){\rm d}s, \end{eqnarray} $

则由(3.10)式得

$ \| u_{mx}\| ^2\le A+B \int^t_0\| u_{mx}\| ^2 \log\| u_{mx}\| ^2{\rm d}s, $

其中$A=2\| u_{mx}(0)\| ^2+\max\{1, 2T\}(1+C)T, B=\max\{1, 2T\}(1+C)$.注意到$B \ge 1$, 则由对数Gronwall不等式引理2.2, 得

$ \begin{equation} \| u_{mx}\| ^2\le (A+B){\rm e}^{Bt}\le C_T. \end{equation} $ (3.11)

于是, 从(3.10)和(3.11)式得出结论

$ \begin{equation} \| u_{mt}\| ^2+\| u_{mxt}\| ^2+\| u_{mx}\| ^2+\| u_{mxx}\| ^2\le C. \end{equation} $ (3.12)

由估计({3.12})知$T_{\max}=T$, 再由经典方法知$u_{mtt}$$L^\infty(0, T; H^{-2})$中一致有界.于是, $\{u_m \}$中存在子序列, 仍记为$\{u_m \}$, 使得

$L^{\infty}(0, T; H^2_0)$中弱$*$意义下$u_m \rightarrow u$;

$L^\infty(0, T; H^1)$中弱$*$意义下$u_{m t} \rightarrow u_t$;

$L^\infty(0, T; H^{-2})$中弱$*$意义下$u_{m tt} \rightarrow u_{tt}$.

进一步, 类似文献[34]中过程, 由Aubin-Lions引理和Lebesgue控制收敛定理, 得

$u_{mx}\log|u_{mx}|^k \rightarrow u_{x}\log|u_{x}|^k$$L^\infty(0, T; L^2)$中弱$*$意义下.

(3.2)式关于$t$从0到$t$积分, 对固定的$s$$m \rightarrow \infty$, 得

$ \begin{eqnarray*}& &(u_{t}, w_{s})+(u_{xt}, w_{sx} )+\int ^t_0 ((u_x, w_{sx}) +(u_{xx}, w_{sxx} )){\rm d}s-\int ^t_0( u_x\log|u_x|^k, w_{sx}){\rm d}s\\ &=&(u_{1}, w_s)+(u_{1x}, w_{sx} ) \end{eqnarray*} $

且解定义中的(2.1)式成立.另一方面, 由(3.3)和(3.4)式知$u(x, 0) = u_0(x), u_t(x, 0) = u_1(x)$, 即(2.2)式成立.由(3.6)式知上述$u$满足(2.3)式.因而上述$u(x, t)$是问题(1.1)-(1.3)的整体解.

定理3.2  设$u_0 \in H^2_0, u_1 \in H^1$, $E(0) < d$, 且$I(u_0)> 0$$\| u_0\| =0$, 则问题(1.1)-(1.3)存在整体解$u(t) \in L^{\infty}(0, \infty; H^2_0)$$u_t (t) \in L^{\infty} (0, \infty; L^2)$$u(.) \in W$$0 \le t < + \infty$成立.

 由定理3.1知问题(1.1)-(1.3)存在整体解.由引理2.6可得$u(t) \in W$$0 \le t > T$成立.

4 解在无穷远处的爆破性

这一节, 我们将证明问题(1.1)-(1.3)的解在无穷远处的爆破性.事实上, 我们将证明在适当的初始条件下解的$H^1_0$范数在时间趋于无穷时将指数增长.

定理4.1  设$u_0 \in H^2_0, ~u_1 \in H^1$.进一步假设$u_0 \in V, 0 < E(0) < d$, $(u_0, u_1)+(u_{0x}, u_{1x}) > 0$, 则问题(1.1)-(1.3)的解当$t\rightarrow + \infty$时爆破.

 设$u(t, x)$是问题(1.1)-(1.3)的弱解且$J(u_0 ) < E(0) < d, I(u_0) < 0$.考虑如下定义的函数$\phi(t):[0, +\infty) \to{\Bbb R}^+$

$ \begin{eqnarray} \phi(t)= \| u\| ^{2}+\| u_x\| ^{2}. \end{eqnarray} $ (4.1)

直接计算表明

$ \begin{equation} \phi'(t)= 2(u, u_t) +2(u_x, u_{xt}). \end{equation} $ (4.2)

因而, 利用方程(1.1)和$I(u)$的定义, 得

$ \begin{eqnarray} \phi''(t)&=&2\| u_t\| ^2+2\| u_{xt}\| ^2+2(u, u_{tt})-2(u, u_{xxtt})\\ &=&2\| u_t\| ^2+2\| u_{xt}\| ^2-2 \Big(\| u_x\| ^2+\| u_{xx}\| ^2-\int _0^l u^2_x\log| u_x|^k{\rm d}x\Big)\\ &=&2\| u_t\| ^2 +2\| u_{xt}\| ^2-2I(u). \end{eqnarray} $ (4.3)

利用Cauchy-Schwarz不等式, 根据(4.2)式对$t \in [0, \infty)$

$ |\phi'(t)|^2 \le 4 \phi(t)(\| u_t\| ^2+\| u_{xt}\| ^2). $

则, 对任意$t \in [0, \infty)$, 由(2.8)和(2.3)式得

$ \begin{eqnarray} \phi''(t)\phi (t)-[\phi'(t)]^2&\ge& 2\phi (t)[\| u_t\| ^2 +\| u_{xt}\| ^2-I(u(t))]-4\phi (t)[\| u_t\| ^2+\| u_{xt}\| ^2] \\ &=&-2\phi (t)[\| u_t\| ^2 +\| u_{xt}\| ^2+I(u(t))] \\ &\ge& -2\phi (t)[E(0)-J(u(t))+I(u(t))]. \end{eqnarray} $ (4.4)

因为$u_0 \in V, E(0) < d$, 根据引理2.6得$I(u(t)) < 0$.于是, 利用引理2.4(2), 我们得到

$ E(0)-J(u(t))+I(u(t)) < d-J(u(t))+2(J(u(t))-d)=J(u(t))-d < 0 $

以及

$ \begin{equation} \phi''(t)\phi (t)-[\phi'(t)]^2 > 0. \end{equation} $ (4.5)

另一方面, 直接计算可得

$ \begin{equation} (\log|\phi(t)|)'=\frac {\phi '(t)}{\phi(t)}, \end{equation} $ (4.6)
$ \begin{equation} (\log|\phi(t)|)''=\Big(\frac {\phi '(t)}{\phi(t)}\Big)' =\frac {\phi''(t)\phi (t)-[\phi'(t)]^2}{\phi ^2(t)} > 0. \end{equation} $ (4.7)

由({4.7})式知, $(\log|\phi(t)|)'=\frac {\phi '(t)}{\phi(t)}$关于$t$是单调增的, 利用这个事实, 把(4.6)式从$t_0$$t$积分, 得

$ \log|\phi(t)|-\log|\phi(t_0)|=\int ^t_{t_0} (\log|\phi(\tau)|)'{\rm d}\tau=\int ^t_{t_0}\frac {\phi '(\tau)}{\phi(\tau)}{\rm d}\tau \ge \frac {\phi'(t_0)}{\phi (t_0)}(t-t_0), $

其中$0\le t_0 < t$.则

$ \begin{equation} \phi(t)\ge \phi(t_0)\exp\Big(\frac {\phi'(t_0)}{\phi (t_0)}(t-t_0)\Big). \end{equation} $ (4.8)

如果能取到$t_0$充分小使得$ \phi'(t_0) > 0, \phi (t_0) > 0$, 则利用({4.8})式得$\lim\limits_{t\to {+\infty}} \phi(t)=+\infty$, 这也就是说问题(1.1)-(1.3)的解当$t\rightarrow + \infty$时爆破.

事实上, 因为$\phi(0)=\| u_0\| ^2+\| u_{0x}\| ^2 > 0, \phi'(0)=2(u_0, u_1)+2(u_{0x}, u_{1x}) > 0$, 则可取$t_0$充分小使得$ \phi'(t_0) > 0, \phi (t_0) > 0$.定理证毕.

5 问题(1.1)-(1.3)带有线性阻尼时的能量衰减估计

这一节, 我们将考虑问题(1.1)-(1.3)带有线性阻尼时的能量衰减估计.也就是说, 我们考虑如下方程

$ \begin{equation} u_{tt}- u_{xx} - u_{xxt}- u_{xxtt}+ u_{xxxx} +(u_x \ln|u_x|^k)_{x}=0, x \in \Omega, t > 0, \end{equation} $ (5.1)

带有初边值条件(1.2) (1.3).这时, 第二节和第三节的所有结果成立, 同时有

$ \begin{equation} E'(t)=-\| u_{xt}\| ^2, \label{5.2} \end{equation} $ (5.2)

进一步, $E(t)$是单调非增的.

引理5.1  设$u_0 \in W, ~u_1 \in H^1$$0 < (0) < \frac{\alpha k}{4}r^2 < d$, 其中$\alpha$是正常数满足$0 < k^{4}\alpha^ {4}\frac{2\pi}{k}e^2 < 1$, 则存在不依赖于$t$的正常数$K$$\xi$使得

$ 0 < E(t)\le K {\rm e}^{-\xi t}, t\ge 0. $

 设$u(t, x)$是问题(5.1), (1.2)和(1.3)的弱解.因为$u_0 \in W, ~u_1 \in H^1$, 利用定理3.2知$u \in W$对所有$t \in [0, +\infty)$, 且$0 < E(t) < d$以及$I(u) > 0$.

现在, 构造Lyapunov泛函如下, 它是能量的一个小扰动

$ \begin{equation} L(t)= E(t)+\epsilon [(u, u_t)+(u_x, u_{xt})], \label{5.3} \end{equation} $ (5.3)

其中$\epsilon > 0$将在后面确定.因为

$ |(u, u_t)|\le \frac{1}{2}(\| u\| ^2+\| u_t\| ^2), |(u_x, u_{xt})|\le \frac{1}{2}(\| u_x\| ^2+\| u_{xt}\| ^2), $

以及$E(t)$的定义(2.8), 则$L(t)$$E(t)$是等价的, 即存在两个仅依赖$\epsilon$的正常数$\beta _1$$\beta _2$使得对$t\ge 0$

$ \begin{equation} \beta _1 E(t)\le L(t)\le \beta _2 E(t). \label{5.4} \end{equation} $ (5.4)

$L(t)$关于时间求导并利用方程(5.1)以及分部积分, 得

$ \begin{equation} L'(t)= (\epsilon -1)\| u_{xt}\| ^2 +\epsilon \| u_t\| ^2- \epsilon \| u_{xx}\| ^2 -\epsilon \| u_x\| ^2 - \epsilon (u_x, u_{xt}) + \epsilon \int^l _0 u_x^2\ln|u_x|^k{\rm d}x.\label{5.5} \end{equation} $ (5.5)

对(5.5)式右边的第五项利用Young不等式, 则对任意$\delta > 0$

$ \begin{equation} |(u_x, u_{xt})|\le \frac{1}{4 \delta }\| u_{xt}\| ^2+ \delta \| u_x\| ^2, \label{5.6} \end{equation} $ (5.6)

进而, 把(5.6)式带入到(5.5)式, 得

$ \begin{equation} L'(t)\le \Big(\epsilon +\frac{\epsilon}{4 \delta }-1\Big)\| u_{xt}\| ^2+\epsilon \| u_t\| ^2- \epsilon \| u_{xx}\| ^2 +\epsilon (\delta-1) \| u_x\| ^2 +\epsilon \int^l _0 u_x^2\ln|u_x|^k{\rm d}x.\label{5.7} \end{equation} $ (5.7)

$E(t)$的定义知, 对任意正常数$M$, 有

$ \begin{eqnarray} L'(t)&\le& -M\epsilon E(t)+ \Big (\frac{M \epsilon}{2}+\epsilon +\frac{\epsilon}{4 \delta }-1\Big) \| u_{xt}\| ^2+\epsilon \Big(1+\frac{M}{2}\Big)\| u_t\| ^2 + \epsilon \Big(\frac{M}{2}-1 \Big)\| u_{xx}\| ^2\\ &&+\epsilon \Big(\frac{M}{2}+\delta-1\Big) \| u_x\| ^2+\epsilon \Big(1-\frac{M}{2}\Big) \int^l _0 u_x^2\ln|u_x|^k{\rm d}x.\label{5.8} \end{eqnarray} $ (5.8)

$0 < M < 1$, 由对数Sobolev不等式引理2.1和Sobolev嵌入定理, 得

$ \begin{eqnarray} L'(t)&\le& -M\epsilon E(t)+ \Big[\frac{M \epsilon}{2}+\epsilon +\frac{\epsilon}{4 \delta }-1+\epsilon \Big(1+\frac{M}{2}\Big)C_0\Big]\| u_{xt}\| ^2\\ &&+ \epsilon \Big(\frac{M}{2}-1\Big )\| u_{xx}\| ^2+\epsilon\Big (1-\frac{M}{2}\Big) \frac{ka^2}{2\pi}\| u_{xx}\| ^2+\epsilon \Big(\frac{M}{2}+\delta-1\Big) \| u_x\| ^2\\ &&+\Big[k\epsilon \Big(1-\frac{M}{2}\Big)\ln\| u_x\| ^2\Big]\| u_x\| ^2 -\epsilon k\Big(1-\frac{M}{2}\Big)\frac{1+\ln a}{2} \| u_x\| ^2\\ &=&-M\epsilon E(t)+\Big [\frac{M \epsilon}{2}+\epsilon +\frac{\epsilon}{4 \delta }-1+\epsilon \Big(1+\frac{M}{2}\Big)C_0\Big]\| u_{xt}\| ^2\\ &&- \epsilon \Big(1-\frac{M}{2}\Big)\Big(1-\frac{ka^2}{2\pi}\Big)\| u_{xx}\| ^2\\ &&+\epsilon \Big[\frac{M}{2}+\delta-1+k\Big(1-\frac{M}{2}\Big)\ln\| u_x\| ^2 -k\Big(1-\frac{M}{2}\Big)\frac{1+\ln a}{2}\Big]\| u_x\| ^2 .\label{5.9} \end{eqnarray} $ (5.9)

注意到因为$0 < M < 1$$1-\frac{M}{2} > 0$以及$\ln\| u_x\| ^2 < \ln (4J(u))$ (因为(2.7)式), 得

$ \begin{eqnarray} L'(t)&\le& -M\epsilon E(t)+ \Big[\frac{M \epsilon}{2}+\epsilon +\frac{\epsilon}{4 \delta }-1+\epsilon \Big(1+\frac{M}{2}\Big)C_0\Big]\| u_{xt}\| ^2\\ &&- \epsilon\Big (1-\frac{M}{2}\Big)\Big(1-\frac{ka^2}{2\pi}\Big)\| u_{xx}\| ^2\\ &&+\epsilon \Big[\frac{M}{2}+\delta-1+k\Big(1-\frac{M}{2}\Big)\ln(4J(u)) -k\Big(1-\frac{M}{2}\Big)\frac{1+\ln a}{2}\Big]\| u_x\| ^2\\ &=& -M\epsilon E(t)+ \Big[\frac{M \epsilon}{2}+\epsilon +\frac{\epsilon}{4 \delta }-1 +\epsilon \Big(1+\frac{M}{2}\Big)C_0\Big]\| u_{xt}\| ^2\\ &&- \epsilon\Big (1-\frac{M}{2}\Big)\Big(1-\frac{ka^2}{2\pi}\Big)\| u_{xx}\| ^2\\ &&+\epsilon\Big [\frac{M}{2}+\delta-1+k\Big(1-\frac{M}{2}\Big) \Big(\ln(4J(u)) -\frac{1+\ln a}{2}\Big)\Big]\| u_x\| ^2 .\label{5.10} \end{eqnarray} $ (5.10)

$0 < M < 1, J(u) < E(0) < \frac{k\alpha r^2}{4}$, 取$a$满足$k^{4}\alpha^ {4}(\frac{2\pi}{k})^2 e^2 < a^2 < \frac{2\pi}{k}$, 这里定理5.1中关于$\alpha$的条件保证$a$能够取到, 再取$\delta > 0$充分小使得

$ 1-\frac{ka^2}{2\pi}>0, $
$ \frac{M}{2}+\delta-1+k\Big(1-\frac{M}{2}\Big)\Big(\ln(4J(u)) -\frac{(1+\ln a)}{2}\Big) < 0, $

$ \begin{equation} L'(t)\le -M\epsilon E(t) +\Big [\frac{M \epsilon}{2}+\epsilon +\frac{\epsilon}{4 \delta }-1+\epsilon \Big(1+\frac{M}{2}\Big)C_0\Big]\| u_{xt}\| ^2 .\label{5.11} \end{equation} $ (5.11)

再取$\epsilon > 0$充分小使得

$ \frac{M \epsilon}{2}+\epsilon +\frac{\epsilon}{4 \delta }-1+\epsilon \Big (1+\frac{M}{2}\Big)C_0 < 0, $

不等式({5.11})变为

$ L'(t)\le -M\epsilon E(t). $

由(5.4)式, 得

$ \begin{eqnarray} L'(t)\le -M\beta_2\epsilon L(t).\label{5.12} \end{eqnarray} $ (5.12)

$K=M\epsilon \beta_2 > 0$, 并对(5.12)式从$0$$t$积分得到关于$L$的如下估计

$ L(t)\le C{\rm e}^{-Kt}. $

再次利用(5.4)式, 即得结果.定理证毕.

参考文献
[1] Bloom F. Asymptotic behavior of solutions to the damped nonlinear equation $\frac{{{\partial ^2}}}{{\partial {t^2}}}u(x,{\rm{ }}t) + r\frac{\partial }{{\partial t}}u(x,{\rm{ }}t) - \frac{\partial }{{\partial x}}\sigma (\frac{\partial }{{\partial x}}u(x,{\rm{ }}t)) = 0$. J Math Anal Appl, 1982, 87: 551–559.
[2] Yang Z J. Existence and nonexistence of global solutions to a generalized modification of the improved Boussinesq equation. Math Meth Appl Sci, 1988, 21: 1467–1477.
[3] Wazwaz A M. Gaussian solitary waves for the logarithmic Boussinesq equation and the logarithmic regularized Boussinesq equation. Ocean Engineering, 2015, 94: 111–115.
[4] Liu Y. Decay and scattering of small solutions of a generalized Boussinesq equation. J Funct Anal, 1997, 147: 51–68.
[5] Lin Q, Wu Y H, Loxton R. On the Cauchy problem for a generalized Boussinesq equation. J Math Anal Appl, 2009, 353: 186–195.
[6] Bona J, Sachs R. Global existence of smooth solutions and stability of solitary waves for a generalized Boussinesq equation. Comm Math Phys, 1988, 118: 15–29.
[7] Yang Z J. Existence and nonexistence of global solutions to a generalized modification of the improved Boussinesq equation. Math Meth Appl Sci, 1998, 21: 1467–1477.
[8] Chen G W, Zhang H W. Initial boundary value problem for a system of generalized IMBq equation. Math Meth Appl Sci, 2004, 27: 497–518.
[9] Chen G W, Wang Y P, Wang S B. Initial boundary value problem of the generalized cubic double dispersion equation. J Math Anal Appl, 2004, 299: 563–577.
[10] You Y C. Nonlinear exponential stabilization of damped Boussinesq equation//Bensoussa A, Lions J L. Analysis and Optimization of Systems. New York:Springer, 1990:642-651
[11] Varlamov V V. On the initial-boundary value for the damped Boussinesq equation. Discrete and Continuous Dynamical Systems, 1998, 4(3): 431–444.
[12] Lai S Y, Wu Y H, Xu Y. The global solution of an initial boundary value problem for the damped Boussinesq equation. Commun Pure Appl Anal, 2004, 3(2): 319–328.
[13] Xue R Y. The initial boundary value problem for the goog Boussinesq equation on the bounded domain. J Math Anal Appl, 2008, 343: 975–995.
[14] Liu Y C, Xu R Z. Potential well method for initial boundary value problem of the generalized double dispersion equations. Commun Pure Appl Anal, 2008, 7: 63–81.
[15] Li K, Fu S H. Asymptotic behavior for the damped Boussinesq equation with critical nonlinearity. Applied Mathematics Letters, 2014, 30: 44–50.
[16] Yang Z J. Longtime dynamics of the damped Boussinesq equation. J Math Anal Appl, 2013, 399: 180–190.
[17] Zhuang W, Yang G T. Propagation of solitary waves in the nonlinear rods. Applied Mathematics and Mechanics, 1986, 7: 571–581.
[18] Zhang S Y, Zhuang W. Strain solitary waves in the nonlinear elastic rods. Acta Mechanica sinica, 1988, 20: 58–66.
[19] Chen G W, Lu B. The initial-boundary value problems for a class of nonlinear wave equations with damping term. J Math Anal Appl, 2009, 351: 1–15.
[20] Xu R Z, Wang S, Yang Y B, Din Y H. Initial boundary value problem for a class of fourth-order wave equation with viscous damping term. Applicable Analysis, 2013, 92(7): 1403–1416.
[21] 张宏伟, 陈国旺. 一类非线性四阶波动方程的位势井方法. 数学物理学报 , 2003, 23A (6): 758–768.
Zhang H W, Chen G W. Potential well method for a class of nonlinear wave equations of fourth-order. Acta Math Sci, 2003, 23A(6): 758–768.
[22] 杨志坚, 陈国旺. 具有阻尼项的非线性波动方程的初值问题. 应用数学学报 , 2000, 23 : 45–54.
Yang Z J, Chen G W. Initial value problem for a nonlinear wave equation with damping term. Acta Math Appl Sci, 2000, 23: 45–54.
[23] Chen G, Yang Z J. Existence and non-existence of global solutions for a class of nonlinear wave equations. Math Meth Appl Sci, 2000, 23: 615–631.
[24] Yang Z J. Global existence, asymptotic behavior and blowup of solutions for a class of nonlinear wave equations with dissipative term. J Differ Equas, 2003, 187: 520–540.
[25] Liu Y C, Xu R Z. Fourth order wave equations with nonlinear strain and source terms. J Math Anal Appl, 2007, 331: 585–607.
[26] Liu Y C, Xu R Z. A class of fourth order wave equations with dissipative and nonlinear strain terms. J Differ Equas, 2008, 244: 200–228.
[27] Yang Z J. Cauchy problem for a class of nonlinear dispersive wave equations arising in elasto-plastic flow. J Math Anal Appl, 2006, 313: 197–217.
[28] Zhang H W, Hu Q Y. Global existence and nonexistence of solution for Cauchy problem of two-dimensional generalized Boussinesq equation. J Math Anal Appl, 2015, 422: 1116–1130.
[29] Kutev N, Kolkovska N, Dimova M. Global existence of Cauchy problem for Boussinesq equation with combined power-type nonlinearities. J Math Anal Appl, 2014, 410: 427–444.
[30] Xu R Z. Cauchy problem for generalized Boussinesq equation with combined power-type nonlinearities. Math Meth Appl Sci, 2011, 34: 2318–2328.
[31] Chen H, Luo P, Liu G W. Global solution and blow-up of a semilinear heat equation with logarithmic nonlinearity. J Math Anal Appl, 2015, 422: 84–98.
[32] Chen H, Tian S Y. Initial boundary value problem for a class of semilinear pseudo-parabolic equations with logarithmic nonlinearity. J Differ Equas, 2015, 258: 4424–4442.
[33] Cazenave T, Haraux A. Equations devolution avec non-linearite logarithmique. Ann Fac Sci Toulouse Math, 1980, 2(1): 21–51.
[34] Gorka P. Logarithmic Klein-Gordon equation. Acta Phys Polon B, 2009, 40(1): 59–66.
[35] Lieb E, Loss M. Analysis (second ed). Providence:Amer Math Soc, 2001
[36] Cotsiolis A, Tavoularis N K. On logarithmic Sobolev inequalities for higher order fractional derivatives. C R Acad Sci Paris, 2005, 340: 205–208.
[37] Payne L E, Sattinger D H. Saddle points and instability of nonlinear hyperbolic equations. Israel J Math, 1975, 22: 273–303.
[38] Liu Y C. On potential wells and applications to semilinear hyperbolic equations and parabolic equations. Nonlinear Analysis, 2006, 64: 2665–2687.
[39] Haraux A, Zuazua E. Decay estimates for some semilinear damped hyperbolic problems. Arch Ration Mech Anal, 1988, 150: 191–206.
[40] Zuazua E. Exponential decay for the semilinear wave equation with locally distributed damping. Comm Partial Differ Equas, 1990, 15(2): 205–235.
[41] Benaissa A, Messaoudi S. Exponential decay of solutions of a nonlinearly damped wave equation. Nonlinear Differential Equations Applications, 2005, 12(4): 391–399.
[42] Gerbi S, Said-Houari B. Exponential decay for solutions to semilinear damped wave equation. Discrete and Continuous Dynamical Systems, 2012, 5(3): 559–566.