数学物理学报, 2020, 40(6): 1568-1589 doi:

论文

一类半线性波动方程的适定性

肖常旺,1, 郭飞,1,2

Global Existence and Blowup Phenomena for a Semilinear Wave Equation with Time-Dependent Damping and Mass in Exponentially Weighted Spaces

Xiao Changwang,1, Guo Fei,1,2

通讯作者: 郭飞, E-mail:guof@njnu.edu.cn

收稿日期: 2019-10-25  

基金资助: 江苏省自然科学基金.  BK20181381
江苏省高校自然科学研究重大项目.  17KJA110002
江苏省青蓝工程

Received: 2019-10-25  

Fund supported: Priority Academic Program Development of Jiangsu Higher Education Institutions, the NSF of Jiangsu Province.  BK20181381
the NSF of the Jiangsu Higher Education Institutions.  17KJA110002
the Qing Lan Projects of Jiangsu Province

作者简介 About authors

肖常旺,E-mail:15996269522@163.com , E-mail:15996269522@163.com

Abstract

We consider the global small data solutions and blowup to the Cauchy problem for a semilinear wave equation with time-dependent damping and mass term as well as power nonlinearity. On one hand, if the power of the nonlinearity $p >p_F(N)=1+ \frac 2N$, it is proved that solutions with small initial data exist for all time in exponentially weighted energy spaces. On the other hand, if the power satisfies <p\leq p_F(\alpha, n)=1+\frac{2(1+\alpha)}{N(1+\alpha)-2\alpha}~(0<\alpha<1)$, for some special chosen parameters it is shown that solutions must blow up in finite time provided that the initial data satisfy some integral sign conditions.

Keywords: Semilinear wave equation ; Global solution ; Blow up ; Fujita exponent

PDF (437KB) 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

肖常旺, 郭飞. 一类半线性波动方程的适定性. 数学物理学报[J], 2020, 40(6): 1568-1589 doi:

Xiao Changwang, Guo Fei. Global Existence and Blowup Phenomena for a Semilinear Wave Equation with Time-Dependent Damping and Mass in Exponentially Weighted Spaces. Acta Mathematica Scientia[J], 2020, 40(6): 1568-1589 doi:

1 引言

本文考虑

$\begin{equation}\label{eq considered} \left\{\begin{array}{ll} u_{tt}-\Delta u+\frac{\mu_1}{(1+t)^{\alpha}}u_t+\frac{\mu_2}{(1+t)^{2\beta}}u=|u|^p, &(t, x)\in(0, \infty)\times\mathbb{R} ^N, \\[2mm]u(0, x)=u_0(x), \quad &x\in\mathbb{R} ^N , \\u_t(0, x)=u_1(x), \quad &x\in\mathbb{R} ^N, \end{array} \right. \end{equation}$

其中$N\geq1, ~0<\alpha<1, ~0<\beta<1$, $\mu_1, ~\mu_2 \in R, \ p>1$.

如果$\mu_1=\mu_2=0$, 则方程(1.1)化为

$\begin{equation} \left\{\begin{array}{ll}u_{tt}-\Delta u=|u|^p, &(t, x)\in (0, \infty)\times\mathbb{R} ^N, \\u(0, x)=u_0(x), \quad& x\in\mathbb{R} ^N, \\u_t(0, x)=u_1(x), \quad& x\in\mathbb{R} ^N.\end{array} \right. \end{equation}$

关于问题(1.2)的整体解的存在性与否的研究称为Strauss猜想, 这是1980年以来半线性波动方程理论中的关键问题之一.我们知道, 若$1 < p \leq p_{St} (N)$, 则解在有限时间内爆破; 而当$p>p_{St} (N)$时, 则整体解存在[6, 10, 15-16, 24], 其中$p_{St}(N)$为Strauss指数, 即如下一元二次方程的正根

$\begin{eqnarray} (N-1)p^2-(N+1)p-2=0, \quad N\geq2, \end{eqnarray}$

$N=1$时, $p_{St} (1)=\infty$.

如果没有质量项, 则方程(1.1)化为阻尼型的波动方程.从物理学角度来看, 该模型可以表示输电线路上电压和电流随距离和时间的变化, 如电报方程, 阻尼与速度成正比的弹性振动和有限传播速度下的热传导.再如半导体, 其材料的电阻与温度有关, 如果温度随时间变化, 则阻尼项具有依赖时间的系数, 可以参考D'Abbicco[4]和Wirth[22]. Todorova-Yordanov[17]研究了阻尼项系数与时间无关的半线性波动方程

$\begin{equation}\label{Ik} \left\{\begin{array}{ll}u_{tt}-\Delta u+u_t=|u|^p, &(t, x)\in (0, \infty)\times\mathbb{R} ^N, \\u(0, x)=u_0(x), \quad& x\in\mathbb{R} ^N , \\u_t(0, x)=u_1(x), \quad& x\in\mathbb{R} ^N.\end{array} \right. \end{equation}$

他们确定了临界指数是$p_{F} (N) = 1 + \frac2N$, 并在空间$C([0, \infty), H^1)\cap C^1 ([0, \infty), L^2)$中研究了局部解的适定性, 爆破和小初值解的整体存在性, 其中$p_{F}(N)$称为Fujita指数[5].需要指出的是, 在$N=1, ~2$时, Li-Zhou[9]也研究了相关问题; Zhang[25]用试验函数方法, 证明了临界指数$p_F(N)$也属于爆破范畴.

Lin-Nishihara-Zhai[11]和Nishihara[13]对阻尼项系数与时间相关的半线性波动方程

$\begin{equation}\label{Ni} \left\{\begin{array}{ll} u_{tt}-\Delta u+\frac{\mu_1}{(1+t)^{\alpha}}u_t=|u|^p, \ &(t, x)\in(0, \infty)\times\mathbb{R} ^N, \\[2mm]u(0, x)=u_0(x), \quad &x\in\mathbb{R} ^N, \\u_t(0, x)=u_1(x), \quad &x\in\mathbb{R} ^N, \end{array} \right. \end{equation}$

研究了"有效的"阻尼(即$-1\leq\alpha<1$)情形下的临界指数问题.

在文献[19-21]中, Wirth系统地研究了参数$\alpha$对方程(1.5)解的影响.当$-1\leq\alpha<1$时,阻尼项称为"有效的"(effective), 这是因为解的行为类似于相应的热方程; 当$\alpha>1$时,阻尼项称为"无效的"(non-effective), 这意味着解的行为与相应的波动方程类似; 当$\alpha<-1$时, 阻尼项在文献[7]中称为"过度阻尼"(over-effective), 此时$(1+t)^{\alpha}\in L^1$.需要指出的是, 若$\alpha=1$, 则方程(1.5)在如下的变换下保持不变

在这种情形下, Wakasugi[18]研究了方程(1.5)的临界指数问题, 并指出解的渐近行为与系数$\mu_1$的大小有着密切关系.

如果没有阻尼项, 则方程(1.1)化为Klein-Gordon方程, 文献[2, 23]对它进行了研究.方程(1.1)是阻尼型波动方程和Kelin-Gordon方程的组合.如果$\alpha=\beta=1$, 则方程(1.1)化为具有尺度不变的阻尼项和质量项系数的半线性波动方程

$\begin{equation}\label{pa} \left\{\begin{array}{ll} u_{tt}-\Delta u+\frac{\mu_1}{1+t}u_t+\frac{\mu_2}{(1+t)^2}u=|u|^p, \ &(t, x)\in (0, \infty)\times\mathbb{R} ^N, \\[2mm]u(0, x)=u_0(x), \quad& x\in\mathbb{R} ^N , \\u_t(0, x)=u_1(x), \quad& x\in\mathbb{R} ^N.\end{array} \right. \end{equation}$

Palmieri[14]采用权函数$e^{2\psi (t, x)}$, 其中$\psi(t, x)=\frac{\mu_1 | x |^2}{2(1 + t)^2}$, 在指数加权空间中, 对(1.6)式的临界指数问题进行了研究.

本文意在研究方程(1.1)在$N\geq1$$p>1$, $0<\alpha <1, ~0<\beta<1$情况下临界指数问题.我们发现:当$1<p\leq p_F(\alpha, N)=1+\frac{2(1+\alpha)}{N(1+\alpha)-2\alpha}$时, 方程(1.1)的解$u$在有限时间内爆破; 当$N\geq3$, $p_F(N)<p\leq\frac {N}{N-2}$时和当$N=1, ~2$, $ p_F(N)<p<\infty$时, 方程(1.1)存在整体解.目前,我们尚未发现该问题的相关研究.我们的困难在于:由于方程(1.1)不能化为阻尼型的波动方程或质量型的波动方程, 致使无法使用文献[17]中的特殊函数来求解.但Nishihara[12]给了我们启发, 我们可以使用加权函数空间来研究该问题.实际上, 如果$0<\alpha <1, ~0 <\beta<1$, 抛物方程

$\begin{equation}\label{1.77} \left\{\begin{array}{ll} u_t-\frac{1}{b(t)}\Delta u=0, \quad&(t, x)\in(0, \infty)\times\mathbb{R} ^N, \\u(0, x)=u_0(x), \quad &x\in\mathbb{R} ^N, \end{array}\right. \end{equation}$

其中$b(t)=(1+t)^{-\alpha}$, 其解$u(t, x)$

其中

因此我们可以用权函数$\psi(t, x)=e^{\frac{|x|^2}{4(1+t)^{1+\alpha}}}$去处理柯西问题(1.1).

本文的内容安排如下:在第2节, 我们将介绍本文的主要结果:局部解-定理2.1, 整体解-定理2.2和爆破结果-定理2.3.在第3节, 基于一些先验能量估计, 将采用Banach不动点证明局部解的适定性(定理2.1).在第4节, 基于一个非线性的能量不等式, 利用反证法证明小初值的解是整体存在的(定理2.2).爆破结果-定理2.3将在第5节中采用试验函数方法证明.试验函数方法最初是由Zhang[25]提出的, 现已成为研究爆破问题的基本方法之一.应当指出的是, 本文只能对一些特殊的参数$\alpha, ~\beta, ~\mu_1$$\mu_2$来应用试验函数方法, 一般情况下是否适用, 我们目前还未知.

2 主要结果

本文中$f\lesssim g$, 意味着存在一个常数$C>0$, 使得$f\leq Cg$.本文使用的函数空间皆定义于整个空间$\mathbb{R} ^N$上, 我们一般不再强调这个事实.

选择权函数$\psi$

$\begin{equation}\label{psi}\psi(t, x)=\frac{a|x|^2}{(1+t)^{1+\alpha}}, \end{equation}$

其中

$\begin{equation}\label{2.7}a=\frac{(1+\alpha)\mu_1}{4(2+\delta)}, \quad \mbox{其中} \delta>0 \mbox{是小常数}.\end{equation}$

$\begin{equation}b(t)=\frac{\mu_1}{(1+t)^\alpha}~(\mu_1>0), ~~m(t)=\frac{\mu_2}{(1+t)^{2\beta}}\end{equation}$

为阻尼项和质量项的系数.

定义带权$e^{\sigma\psi(t, \cdot)} (\sigma>0)$$L^2$空间和Sobolev空间$H^1$如下

$\begin{equation}\begin{array}{ll}&L^2_{\sigma\psi(t, \cdot)}=\{f\in L^2: \| e^{\sigma\psi(t, \cdot)}f\|_{L^2}<\infty\}, \\&H^1_{\sigma\psi(t, \cdot)}=\{f\in H^1: \| e^{\sigma\psi(t, \cdot)}f\|_{L^2}+\|e^{\sigma\psi(t, \cdot)}\nabla f\|_{L^2}<\infty\}, \end{array}\end{equation}$

其范数为

$\begin{equation}\label{d1}\begin{array}{ll}&\|f\|_{L^2_{\sigma\psi(t, \cdot)}}=\| e^{\sigma\psi(t, \cdot)}f\|_{L^2}, \\&\|f\|_{H^1_{\sigma\psi(t, \cdot)}}=\| e^{\sigma\psi(t, \cdot)}f\|_{L^2}+\|e^{\sigma\psi(t, \cdot)}\nabla f\|_{L^2}.\end{array}\end{equation}$

$\begin{equation}\label{D}{\cal D}=H^1_{\psi(0, \cdot)}\otimes L^2_{\psi(0, \cdot)}=H^1_{a|x|^2}\otimes L^2_{a|x|^2}.\end{equation}$

对于任意的$\sigma>0$$t>0$, 成立

$\begin{eqnarray}L^2_{\sigma\psi(t, \cdot)}\hookrightarrow L^1\cap L^2.\end{eqnarray}$

事实上, 利用Cauchy-Schwartz不等式并注意到$\psi>0$, 我们有

$\begin{equation}\begin{array}{ll}&\|f\|_{L^1}\lesssim (1+t)^{\frac{N(1+\alpha)}{2}}\|f\|_{L^2_{\sigma\psi(t, \cdot)}}, \\&\|f\|_{L^2}\lesssim \|f\|_{L^2_{\sigma\psi(t, \cdot)}}.\end{array}\end{equation}$

另外容易验证

$\begin{equation}L^2_{\sigma\psi(t, \cdot)}\hookrightarrow L^r , \end{equation}$

其中$ r\in [1, 2]$.

下面先给出局部解存在的结果, 它是本文采用反证法得到小初值的整体解的前提.

定理 2.1 (局部解)  假设$0<\alpha<1, ~0<\beta<1, ~\mu_1>0, ~\mu_2\geq0$, 且

$ \begin{eqnarray} 1<p\leq\frac{N}{N-2}, ~N\geq3;~1<p<\infty, ~N=1, ~2. \end{eqnarray}$

则对每个初值$(u_0, u_1)\in{\cal D}$ (见(2.6)式), 存在一个最大存在时间$T^*\in(0, \infty]$, 使得柯西问题(1.1)有唯一解$u\in C\left([0, T^*), H^1(\mathbb{R} ^N)\right)\cap C^1\left([0, T^*), L^2(\mathbb{R} ^N)\right)$, 且对任意的$T\in(0, T^*)$, 满足

$\begin{equation}\sup\limits_{t\in [0, T]}\left(\|u(t, \cdot)\|_{H^1_{\psi(t, \cdot)}}+\|u_t(t, \cdot)\|_{L^2_{\psi(t, \cdot)}}\right)<\infty, \end{equation}$

其中$\|\cdot\|_{H^1_{\psi(t, \cdot)}}$$\|\cdot\|_{L^2_{\psi(t, \cdot)}}$的定义见(2.5).进一步, 若$T^*<\infty$, 则

$\begin{equation}\label{T}\limsup\limits_{T\rightarrow T^*t\in [0, T]} \left( \|u(t, \cdot)\|_{H^1_{\psi(t, \cdot)}}+\|u_t(t, \cdot)\|_{L^2_{\psi(t, \cdot)}} \right)=\infty.\end{equation}$

现在给出整体解的结果, 由前文知Fujita指数$ p_F (N) = 1 + \frac2N. $对于方程(1.1), 我们定义初始能量

$\begin{equation}\label{ie} I_0^2=\int_{\mathbb{R}^N}e^{2\psi(0, x)}\left(|u_1(x)|^2+|\nabla u_0(x)|^2+|u_0(x)|^2\right){\rm d}x. \end{equation}$

定理 2.2(整体解)  假设$0<\alpha<1, ~0<\beta<1, ~\mu_1>0, ~\mu_2\geq0$, 且

$ \begin{equation}p_F(N)<p\leq\frac{N}{N-2}, ~N\geq3;~p_F(N)<p<\infty, ~N=1, ~2.\end{equation}$

如果存在一个常数$\delta >0$, 对任意的初值$(u_0, u_1)\in{\cal D}$(见(2.6)式)满足

则方程(1.1)有唯一解$u\in X_{\infty}=C\left([0, \infty), H^1(\mathbb{R} ^N)\right)\cap C^1\left([0, \infty), L^2(\mathbb{R} ^N)\right)$, 且

$\begin{equation}\|u(t, \cdot)\|_{L^2_{\psi(t, \cdot)}}\leq C(1+t)^{-\frac{(1+\alpha)N}{4}+\frac{\epsilon}{2}}, \label{2.14}\end{equation}$

$\begin{equation}\|\left( u_t(t, \cdot), \nabla u(t, \cdot) \right)\|_{L^2_{\psi(t, \cdot)}}\leq C(1+t)^{-\frac{(1+\alpha)(N+2)}{4}+\frac{\epsilon}{2}}, \label{2.15}\end{equation}$

其中$C=C(\alpha, \beta, \mu_1, \mu_2, \delta, p, N)$.

$\begin{equation}p_F(\alpha, N)=1+\frac{2(1+\alpha)}{N(1+\alpha)-2\alpha}.\end{equation}$

我们将证明, 当$1< p \leq p_F(\alpha, N)$时, 对于特定参数$\alpha, ~\beta~\mu_1, ~\mu_2$和满足一定积分符号条件且具有紧支集的初值, 方程(1.1)相应的解必在有限时间内爆破.

定理 2.3 (爆破)  对于方程(1.1), 假设

$\begin{equation}\label{3.2} \left\{\begin{array}{ll}\big(u_0(x), u_1(x)\big)\in H^1(\mathbb{R} ^N)\times L^2(\mathbb{R} ^N)\mbox{ 具有紧支集 }, \\\int_{\mathbb{R} ^N}\big( \mu_1u_0(x)+u_1(x)\big) {\rm d}x>0, \end{array} \right. \end{equation}$

其中$0<\alpha<1, ~2\beta=\alpha+1$, $\mu_1>0, ~\mu_2=-\mu_1\alpha$.如果$1<p\leq p_F(\alpha, N)$, 则方程(1.1)不存在解$u\in C\left([0, \infty), H^1(\mathbb{R} ^N)\right)\cap C^1\left([0, \infty), L^2(\mathbb{R} ^N)\right)$.若对方程(1.1)赋予初值

$\begin{eqnarray}u(0, x)=\varepsilon u_0(x), \quad u_t(0, x)=\varepsilon u_1(x), \end{eqnarray}$

其中$u_0(x)$$u_1(x)$满足条件(2.18), 则存在一个常数$C=C\left(\alpha, \mu_1, u_0(x), u_1(x), N, p\right)$, 使得对于任何次临界指数$p$, 方程(1.1)的解的最大存在时间$T^*$满足

$\begin{eqnarray}T^*\leq C\varepsilon^{-\frac{1}{\frac{1}{p-1}(1+\alpha)-\frac{N}{2}(1+\alpha)+\alpha}}.\end{eqnarray}$

3 局部解

问题

$\begin{equation}\label{3.3} \left\{\begin{array}{ll} u_{tt}-\Delta u+\frac{\mu_1}{(1+t)^{\alpha}}u_t+\frac{\mu_2}{(1+t)^{2\beta}}u=0, \ &(t, x)\in (s, \infty)\times\mathbb{R} ^N, \\[2mm]u(s, x)=u_0(x), \quad &x\in\mathbb{R} ^N , \\u_s(s, x)=u_1(x), \quad& x\in\mathbb{R} ^N\end{array} \right. \end{equation}$

的解为

其中$E_0(t, s, x)$$E_1(t, s, x)$分别是初值$(u_0, u_1)=(\delta_0, 0)$$(0, \delta_0)$时的基本解, $\delta_0$是变量$x$的Dirac分布, $\ast_{(x)}$是对变量$x$的卷积.由Duhamel原理, 方程(1.1)在$(0, T)\times\mathbb{R} ^N$上的解就是算子$R$的不动点, 其中$R$定义为

$\begin{eqnarray}u\in Y(t)\rightarrow Ru(t, x)&=&E_{0}(t, 0, x)\ast_{(x)}u_{0}(x)+E_{1}(t, 0, x)\ast_{(x)}u_{1}(x)\\&&+\int_0^t E_1(t-s, 0, x)\ast_{(x)}|u(s, x)|^p{\rm d}s, \end{eqnarray}$

这里$ Y(t)=C\left([0, t], H^1_{\psi(t, .)}\right)\cap C^1\left([0, t], L^2_{\psi(t, .)}\right). $

对于$\psi$(见(2.1)式), 容易验证

$\begin{equation}\psi_t=-(1+\alpha)\frac{a|x|^2}{(1+t)^{2+\alpha}}=-\frac{1+\alpha}{1+t}\psi, \end{equation}$

$\begin{equation}\nabla \psi=\frac{2ax}{(1+t)^{1+\alpha}}, ~\triangle\psi=\frac{2aN}{(1+t)^{1+\alpha}}, \label{2.6}\end{equation}$

$\begin{equation}\frac{|\nabla\psi|^2}{-\psi_t}=\frac{b(t)}{2+\delta}, \label{2.8}\end{equation}$

$\begin{equation}\triangle\psi=\frac{(1+\alpha)N}{2(2+\delta)}\frac{b(t)}{1+t}=\left(\frac{(1+\alpha)N}{4}-\delta_1\right)\frac{b(t)}{t+1}, \label{2.9}\end{equation}$

其中当$\delta\rightarrow0$时, $\delta_1=\frac{\delta(1+\alpha)N}{2(2+\delta)}\rightarrow0$.

在证明定理2.1时, 我们将利用下面的一些引理, 虽然它们只是对文献[8, 17]中相应结果的修改, 但为了方便起见, 我们也给出它们的证明.

引理 3.1  对$\sigma>0$$v\in H^1_{\sigma \psi(t, \cdot)}$, 成立

$\begin{eqnarray}(1+t)^{-(1+\alpha)}\|e^{\sigma\psi(t, \cdot)}v\|^2_{L^2}+\|\nabla(e^{\sigma\psi(t, \cdot)}v)\|^2_{L^2}\lesssim\|e^{\sigma\psi(t, \cdot)}\nabla v\|^2_{L^2} , \quad\forall ~t \geq0, \end{eqnarray}$

其中$\psi(t, x)$为权函数(见(2.1)式).

  令$f=e^{\sigma\psi}v$, 则

$ \begin{eqnarray} \nabla v=\nabla(e^{-\sigma\psi}f)=-\sigma e^{-\sigma\psi}f\nabla\psi+e^{-\sigma\psi}\nabla f.\end{eqnarray}$

$e^{\sigma\psi}\nabla v=\nabla f-\sigma f\nabla \psi$

$\begin{eqnarray}\label{4.11}\|e^{\sigma\psi}\nabla v\|^2_{L^2}&=&(\nabla f-\sigma f\nabla \psi, \nabla f-\sigma f\nabla \psi)\\&=&\|\nabla f||^2_{L_2}+\sigma^2\|f\nabla \psi\|^2_{L_2}-2\sigma(\nabla f, f\nabla\psi).\end{eqnarray}$

分部积分得

$\begin{eqnarray}(\nabla f, f\nabla\psi)=\int_{\mathbb{R} ^N}(\nabla f)f\nabla\psi {\rm d}x=-\frac{1}{2}\int_{\mathbb{R} ^N}f^2\triangle\psi {\rm d}x.\end{eqnarray}$

由(3.4)和(3.9)式得

$ \begin{eqnarray} \|e^{\sigma\psi(t, \cdot)}\nabla v\|^2_{L^2}\gtrsim(1+t)^{-(1+\alpha)}\|f\|^2_{L^2}+\|\nabla f\|^2_{L^2}.\end{eqnarray}$

引理3.1证毕.

引理 3.2  对$\theta(q)=N(\frac{1}{2}-\frac{1}{q})\in[0, 1], ~\sigma\in(0, 1]$$v\in H^1_{\psi(t, \cdot)}$, 成立

$\begin{eqnarray}\|e^{\sigma\psi(t, \cdot)}v\|_{L^q}\lesssim(1+t)^{\frac{1+\alpha}{2}(1-\theta(q))}\|\nabla v\|^{1-\sigma}_{L^2}\|e^{\sigma\psi(t, \cdot)}\nabla v\|^{\sigma}_{L^2}, \quad\forall~t \geq0.\end{eqnarray}$

  令$f=e^{\sigma\psi}v$.利用经典的Gagliardo-Nirenberg插值不等式(见文献[1]), 我们得到

$\begin{eqnarray}\label{3.12}\|f\|_{L_q}\lesssim\|f\|_{L^2}^{1-\theta(q)}\|\nabla f\|_{L^2}^{\theta(q)}, \end{eqnarray}$

其中$\theta(q)=N(\frac{1}{2}-\frac{1}{q})$.由引理3.1, 有

$\begin{eqnarray}\label{3.13}\|f\|_{L^2}\lesssim(1+t)^{\frac{1+\alpha}{2}}\|e^{\sigma\psi(t, \cdot)}\nabla v\|_{L^2}, \quad\|\nabla f\|_{L^2}\lesssim\|e^{\sigma\psi(t, \cdot)}\nabla v\|_{L^2}.\end{eqnarray}$

因此

$\begin{eqnarray}\label{3.14}\|f\|_{L_q}&\lesssim&(1+t)^{\frac{1+\alpha}{2}(1-\theta(q))}\|e^{\sigma\psi(t, \cdot)}\nabla v\|^{1-\theta(q)}_{L^2}\|e^{\sigma\psi(t, \cdot)}\nabla v\|^{\theta(q)}{_{L^2}}\\& \lesssim&(1+t)^{\frac{1+\alpha}{2}(1-\theta(q))}\|e^{\sigma\psi(t, \cdot)}\nabla v\|_{L^2}\\&\lesssim&(1+t)^{\frac{1+\alpha}{2}(1-\theta(q)}\|\nabla v\|^{1-\sigma}_{L^2}\|e^{\sigma\psi(t, \cdot)}\nabla v\|^{\sigma}_{L^2}.\end{eqnarray}$

引理3.2证毕.

下面的引理是Gronwall不等式的非线性形式, 可参考文献[3].

引理 3.3  设$k(s)$是非负连续函数, $M$是正实数, $g(s)$是非负连续增函数, 有

$\begin{equation}G(u)=\int^u_0\frac{1}{g(s)}{\rm d}s.\end{equation}$

$y(t)$为连续函数使得

$\begin{equation}y(t)\leq M+\int^t_0k(s)g(y(s)){\rm d}s, ~~\forall~t\geq0, \end{equation}$

$\begin{equation}G(y(t))\leq G(M)+\int^t_0k(s){\rm d}s, ~~\forall~t\geq0.\end{equation}$

$\begin{equation}E(T)=\left\{u\in C\left([0, T], H^1\right)\cap C^1\left([0, T], L^2\right):~~\|u\|^{\psi}_{T}< \infty \right\}, \end{equation}$

其中

$\begin{equation}\|u\|^{\psi}_{T}=\sup\limits_{t\in[0, T]}\left(\|u(t, \cdot)\|_{H^1_{\psi(t, \cdot)}}+\|u_t(t, \cdot)\|_{L^1_{\psi(t, \cdot)}}\right).\end{equation}$

容易验证$\left(E(T), \|\cdot\|^{\psi}_{T}\right) $是一个Banach空间.

现在, 我们证明定理2.1.

定理 2.1 的证明  对$T, ~R>0$, 记

$\begin{eqnarray} A^\psi_{T, ~R}=\left\{v\in C\left([0, T], H^1\right)\cap C^1\left([0, T], L^2\right): ~\|v\|^{\psi}_{T}\leq R\right\}. \end{eqnarray}$

定义映射

其中$u$是下列柯西问题的唯一解

$\begin{equation}\label{5.1} \left\{\begin{array}{ll}u_{tt}-\Delta u+b(t)u_t+m(t)u=|v|^p, \quad&(t, x)\in (0, T]\times\mathbb{R} ^N, \\u(0, x)=u_0(x), \quad &x\in \mathbb{R} ^N, \\u_t(0, x)=u_1(x), \quad &x\in \mathbb{R} ^N.\\\end{array} \right. \end{equation}$

首先, 我们证明$\theta(A^\psi_{T, R})\subset A^\psi_{T, R}$.在方程(3.22)两边同乘$e^{2\psi}u_t$得到

$\begin{eqnarray}e^{2\psi}u_t\left(u_{tt}-\Delta u+b(t)u_t+m(t)u\right)=e^{2\psi}u_t|v|^p \label{5.2}.\end{eqnarray}$

把下面一些等式

代入(3.23)式, 有

$\begin{eqnarray}\label{11}&&\frac{\partial\left[{\frac{e^{2\psi}}{2}}(u_t^2+|\nabla u|^2+m(t)u^2)\right]}{\partial t}-{\rm div}(e^{2\psi}u_t\nabla u)+\frac{e^{2\psi}}{\psi_t}u^2_t\left(|\nabla\psi|^2+b(t)\psi_t\right)\\&&-\frac{e^{2\psi}}{\psi_t}|u_t\nabla\psi-\psi_t\nabla u|^2-\psi_te^{2\psi}(u_t^2+m(t)u^2)-\frac{1}{2}e^{2\psi}u^2\frac{{\rm d}m(t)}{{\rm d}t} =e^{2\psi}u_t|v|^p.\end{eqnarray}$

由于$\psi_t <0, ~\frac{{\rm d}m(t)}{{\rm d}t}<0$, 注意到(3.5)式, 可知$|\nabla\psi(t, x)|^2+b(t)\psi_t(t, x) \leq 0$, 故(3.24)式可化为

$\begin{equation}\frac{\partial\left[{\frac{e^{2\psi}}{2}}(u_t^2+|\nabla u|^2+m(t)u^2)\right]}{\partial t}-{\rm div}(e^{2\psi}u_t\nabla u)\leq e^{2\psi}u_t|v|^p.\label{5.3}\end{equation}$

$\begin{equation}\label{E}E_{\psi, u}(t)=\int_{\mathbb{R} ^N}\frac{e^{2\psi}}{2}(u_t^2+|\nabla u|^2+m(t)u^2){\rm d}x.\end{equation}$

$L^1$函数用弱导数意义下的散度定理, 有$\int_{\mathbb{R} ^N}{\rm div}(e^{2\psi}u t\nabla u){\rm d}x=0$.再将(3.25)式在$\mathbb{R} ^N$上积分并利用Cauchy-Schwartz不等式, 得到

$\begin{eqnarray}\label{s4.9}E_{\psi, u}(t)&\leq &E_{\psi, u}(0)+\int_0^t\int_{\mathbb{R} ^N}e^{2\psi(s, x)}u_s(s, x)|v(s, x)|^p{\rm d}x{\rm d}s\\&\lesssim &E_{\psi, u}(0)+\int_0^t\left(\int_{\mathbb{R} ^N}e^{2\psi(s, x)}|v(s, x)|^{2p}{\rm d}x\right)^{\frac{1}{2}} E^{\frac{1}{2}}_{\psi, u}(s){\rm d}s.\end{eqnarray}$

$ g(s)=s^{\frac{1}{2}}, ~G(u)=u^{\frac{1}{2}}$, 应用引理3.3, (3.27)式可化为

$\begin{equation}\label{s4.10}E^{\frac{1}{2}}_{\psi, u}(t)\lesssim E_{\psi, u}^{\frac{1}{2}}(0)+\int_0^t\left(\int_{\mathbb{R} ^N}e^{2\psi(s, x)}|v(s, x)|^{2p}{\rm d}x\right)^{\frac{1}{2}} {\rm d}s.\end{equation}$

由引理3.2, 取$\sigma=\frac{1}{p}\in(0, 1)$并注意到$A^\psi_{T, R}$的定义, 知

$\begin{eqnarray}\label{s4.11}\left(\int_{\mathbb{R} ^N}e^{2\psi(s, x)}|v(s, x)|^{2p}{\rm d}x\right)^{\frac{1}{2}}& \lesssim&(1+t)^{\frac{1+\alpha}{2}(1-\theta(2p))p}\|\nabla v(s, .)\|^{p-1}_{L^2}\|e^{\psi(s, .)}\nabla v(s, .)\|_{L^2}\\&\lesssim& (1+t)^{\frac{1+\alpha}{2}(1-\theta(2p))p}R^p, \end{eqnarray}$

因此

$\begin{equation}\label{s4.12} E^{\frac{1}{2}}_{\psi, u}(t)\lesssim E^{\frac{1}{2}}_{\psi, u}(0)+\int_0^t(1+s)^{\frac{1+\alpha}{2}(1-\theta(2p))p}R^p{\rm d}s.\end{equation}$

$E_{\psi, u}$的定义, 我们有

$\begin{eqnarray}&&\|e^{\psi(t, \cdot)}u_t(t, \cdot)\|_{L_2}+\|e^{\psi(t, \cdot)}\nabla u(t, \cdot)\|_{L_2}\lesssimE^{\frac{1}{2}}_{\psi, u}(0)+T(1+T)^{\frac{1+\alpha}{2}(1-\theta(2p))p}R^p, \\&&\|e^{\psi(t, \cdot)}u(t, \cdot)\|_{L_2} = \frac{1}{\sqrt{m(t)}}\|e^{\psi(t, \cdot)}\sqrt{m(t)}u(t, \cdot)\|_{L_2}\\&&\lesssim(1+T)^{\beta}E^{\frac{1}{2}}_{\psi, u}(0)+T(1+T)^{\frac{1+\alpha}{2}(1-\theta(2p))p+\beta}R^p, \end{eqnarray}$

因此

$\begin{eqnarray}\|u\|^\psi_T&=&\|u(t, \cdot)\|_{H^1_{\psi(t, \cdot)}}+\|u_t(t, \cdot)\|_{L^2_{\psi(t, \cdot)}}\\&\lesssim&(1+T)^{\beta}E^{\frac{1}{2}}_{\psi, u}(0)+T(1+T)^{\frac{1+\alpha}{2}(1-\theta(2p))p+\beta}R^p.\end{eqnarray}$

选择足够小的$T$和足够大的$R$, 使得

$ \begin{equation}(1+T)^{\beta}E^{\frac{1}{2}}_{\psi, u}(0)<\frac{R}{2}, \quad\quadT(1+T)^{\frac{1+\alpha}{2}(1-\theta(2p))p+\beta}R^p<\frac{R}{2}, \end{equation}$

所以

$\begin{equation} \|u\|^\psi_T<R, \end{equation}$

这意味着$\theta:A^{\psi}_{T, R}\rightarrow A^{\psi}_{T, R} $.

其次, 我们证明:当$T>0$充分小时, $\theta$是一个压缩映射.为此, 对$v, ~\overline{v}\in A^{\psi}_{T, R}$, 记$u=\theta(v), ~\overline{u}=\theta(\overline{v})$.则~$\omega=u-\overline{u}$满足

$ \begin{equation} \left\{\begin{array}{ll}\omega_{tt}-\Delta \omega+b(t)\omega_t+m(t)\omega=|v|^p-|\overline{v}|^p, \ &(t, x)\in (0, T)\times\mathbb{R} ^N, \\\omega(0, x)=\omega_t(0, x)=0, \quad& x\in \mathbb{R} ^N.\end{array} \right. \end{equation}$

$\begin{equation}\label{E2}E_{\psi, \omega}(t)=\int_{\mathbb{R} ^N}\frac{e^{2\psi}}{2}(\omega_t^2+|\nabla \omega|^2+m(t)\omega^2){\rm d}x.\end{equation}$

类似(3.27)式, 得到

$\begin{equation}\label{s4.18}E_{\psi, \omega}(t)\lesssim \int_0^t\int_{\mathbb{R} ^N}e^{2\psi(s, x)}(|v(s, x)|^p-|\overline{v}(s, x)|^p)\omega_s(s, x){\rm d}x.\end{equation}$

代入上式, 得

$\begin{eqnarray}\label{s4.19}E_{\psi, \omega}(t)&\lesssim &\int_0^t\int_{\mathbb{R} ^N}e^{2\psi(s, x)}|v(s, x)-\overline{v}(s, x)|(|v(s, x)|^{p-1}+|\overline{v}(s, x)|^{p-1})\omega_s(s, x){\rm d}x{\rm d}s\\&\lesssim &\int_0^t\left(\int_{\mathbb{R} ^N}e^{2\psi(s, x)}|v(s, x)-\overline{v}(s, x)|^2|v(s, x)|^{2(p-1)}{\rm d}x\right)^{\frac{1}{2}}E_{\psi, \omega}^{\frac{1}{2}}(s){\rm d}s\\&&+\int_0^t\left(\int_{\mathbb{R} ^N}e^{2\psi(s, x)}|v(s, x)-\overline{v}(s, x)|^2|\overline{v}(s, x)|^{2(p-1)}{\rm d}x\right)^{\frac{1}{2}}E_{\psi, \omega}^{\frac{1}{2}}(s){\rm d}s.\end{eqnarray}$

由引理3.3, 得

$\begin{eqnarray}\label{s4.21}E_{\psi, \omega}^{\frac{1}{2}}(t)&\lesssim&\int_0^t\left(\int_{\mathbb{R} ^N}e^{2\psi(s, x)}|v(s, x)-\overline{v}(s, x)|^2|v(s, x)|^{2(p-1)}{\rm d}x\right)^{\frac{1}{2}}{\rm d}s\\&& +\int_0^t\left(\int_{\mathbb{R} ^N}e^{2\psi(s, x)}|v(s, x)-\overline{v}(s, x)|^2|{\overline{v}}(s, x)|^{2(p-1)}{\rm d}x\right)^{\frac{1}{2}}{\rm d}s.\end{eqnarray}$

利用Hölder不等式和引理3.2, 并注意到$A^{\psi}_{T, R}$的定义, 有

$\begin{eqnarray}\label{s4.22}&&\|e^{\psi(s, \cdot)}|v(s, \cdot)-\overline{v}(s, \cdot)||v(s, \cdot)|^{(p-1)}\|_{L^2}\\&\lesssim&\|e^{(2-p)\psi(s, \cdot)}|v(s, \cdot)-\overline{v}(s, \cdot)|\|_{L^{2p}}\|e^{\psi(s, \cdot)}|v(s, \cdot)|\|_{L^{2p}}^{p-1}\\&\lesssim&(1+s)^{\frac{1+\alpha}{2}(1-\theta(2p))p}\|e^{\psi(s, \cdot)}\nabla(v(s, \cdot)-\overline{v}(s, \cdot))\|_{L^2}\|e^{\psi(s, \cdot)}\nabla v(s, \cdot)\|^{p-1}_{L^2}\\&\lesssim& (1+s)^{\frac{1+\alpha}{2}(1-\theta(2p))p}\|v-\overline{v}\|^{\psi}_TR^{p-1}\end{eqnarray}$

$\begin{equation}\label{s4.23}\|e^{\psi(s, \cdot)}|v(s, \cdot)-\overline{v}(s, \cdot)||\bar{v}(s, \cdot)|^{(p-1)}\|_{L^2}\lesssim(1+s)^{\frac{1+\alpha}{2}(1-\theta(2p))p}\|v-\overline{v}\|^{\psi}_TR^{p-1}.\end{equation}$

将上面两式代入(3.39)式, 得到

$\begin{eqnarray}\label{s4.24}E_{\psi, \omega}^{\frac{1}{2}}(t)&\lesssim&\int_0^t(1+s)^{\frac{1+\alpha}{2}(1-\theta(2p))p}{\rm d}s\|v-\overline{v}\|^{\psi}_TR^{p-1}\\&\lesssim&\ T(1+T)^{\frac{1+\alpha}{2}(1-\theta(2p))p}\|v-\overline{v}\|^{\psi}_TR^{p-1}.\end{eqnarray}$

$E_{\psi, \omega}$的定义(3.36), 知

$\begin{eqnarray}\label{s4.25}&&\|e^{\psi(t, \cdot)}\omega_t(t, \cdot)\|_{L^2}+\|e^{\psi(t, \cdot)}\nabla\omega(t, \cdot)\|_{L^2}\lesssim T(1+T)^{\frac{1+\alpha}{2}(1-\theta(2p))p}\|v-\overline{v}\|^{\psi}_TR^{p-1}, \\&&\|e^{\psi(t, \cdot)}\omega(t, \cdot)\|_{L^2}=\frac{1}{\sqrt{m(t)}}\|e^{\psi(t, \cdot)}\sqrt{m(t)}\omega(t, \cdot)\|_{L^2}\\&& \ \lesssimT(1+T)^{\frac{1+\alpha}{2}(1-\theta(2p))p+\beta}\|v-\overline{v}\|^{\psi}_TR^{p-1}, \end{eqnarray}$

所以

$\begin{eqnarray}\|\omega\|^{\psi}_T&=&\|\theta(v)-\theta(\overline{v})\|^{\psi}_T\lesssim T(1+T)^{\frac{1+\alpha}{2}(1-\theta(2p))p+\beta}\|v-\overline{v}\|^{\psi}_TR^{p-1}\nonumber\\&\leq &C_0T(1+T)^{\frac{1+\alpha}{2}(1-\theta(2p))p+\beta}\|v-\overline{v}\|^{\psi}_TR^{p-1}, \end{eqnarray}$

其中$C_0=C_0(\alpha, \beta, \mu_1, \mu_2, N, p, R)$.选择足够小的$T>0$, 使得

$\begin{equation}\|\omega\|^{\psi}_T=\|\theta(v)-\theta(\overline{v})\|^{\psi}_T<{\alpha}_1\|v-\overline{v}\|^{\psi}_T, \end{equation}$

即映射$\theta$是压缩的.

由于$\left(E(T), \|\cdot\|^{\psi}_T\right)$是一个Banach空间, 利用Banach不动点定理, 可知柯西问题(1.1)有唯一解$u\in C\left([0, T^*), H^1(\mathbb{R} ^N)\right) \cap C^1 \left([0, T^* ), L^2(\mathbb{R} ^N)\right)$且对任何$t\in[0, T^*)$, 相应的$E_{\psi, u}(t)$都有限, $T^*<\infty$时意味着当$T \rightarrow T^*$时, 能量将会爆破.

4 整体解

本节我们将利用文献[18]的加权能量方法证明方程(1.1)的小初值的解的整体存在性结果-定理2.2.不失一般性, 不妨假设$\mu_1=\mu_2=1$.

命题 4.1  假设

$\begin{equation}p_F(N)<p<\frac{N}{N-2}, ~N\geq3;~p_F(N)<p<\infty, ~N=1, ~2\end{equation}$

$0<\alpha<1, ~0<\beta<1$.如果存在$\delta>0$$T>0$, 使柯西问题(1.1)的解 $u\in C\left([0, T], H^1\right)\cap C^1\left([0, T], L^2\right)$, 且满足

$\begin{equation}\sup\limits_{0\leq t\leq T}\int_{\mathbb{R} ^N}e^{2\psi}(m(t)u^2+|\nabla u|^2+|u_t|^2){\rm d}x<\infty, \end{equation}$

$\begin{equation}\label{MM}M(t)\leq CI_0^2+CM^{p+1}(t), \end{equation}$

其中

$\begin{eqnarray}\label{s15}M(t)&=&\sup\limits_{0\leq \tau\leq t}\bigg\{(1+\tau)^{\frac{(1+\alpha)(N+2)}{2}-\epsilon}\int_{\mathbb{R} ^N}e^{2\psi}\left(m(\tau)u^2+|\nabla u|^2+u_{\tau}^2\right){\rm d}x\\&&+(1+\tau)^{\frac{(1+\alpha)N}{2}-\epsilon}\int_{\mathbb{R} ^N}e^{2\psi}u^2{\rm d}x\bigg\}, \end{eqnarray}$

$C=C(\alpha, \beta, p, \delta, N), ~\epsilon=\epsilon(\delta) =\frac{3\delta(1+\alpha)N}{2(2+\delta)}$, $\psi(t, x)=\frac{a|x|^2}{(1+t)^{1+\alpha}}$, $a=\frac{1+\alpha}{4(2+\delta)}$, $I_0^2$是初始能量(定义见(2.13)式), 即$ I_0^2=\int_{\mathbb{R}^N}e^{2\psi(0, x)}\left(|u_1(x)|^2+|\nabla u_0(x)|^2+|u_0(x)|^2\right){\rm d}x. $

证  方程(1.1)即

$\begin{equation}\label{BB} \left\{\begin{array}{ll}u_{tt}-\Delta u+b(t)u_t+m(t)u=|u|^p, \quad&(t, x)\in(0, \infty)\times\mathbb{R} ^N, \\u(0, x)=u_0(x), \quad &x\in\mathbb{R} ^N , \\u_t(0, x)=u_1(x), \quad &x\in\mathbb{R} ^N, \end{array}\right.\end{equation}$

其中$b(t)=\frac{1}{(1+t)^\alpha}, ~m(t)=\frac{1}{(1+t)^{2\beta}}$.$F(u)=\int_0^u|s|^p{\rm d}s, ~f(u)=|u|^p$.在(4.5)式两边同乘$e^{2\psi} u_t$, 得

$\begin{eqnarray}&&\frac{\partial\left[\frac{e^{2\psi}}{2}\left(u_t^2+|\nabla u|^2+m(t)u^2\right)\right]}{\partial t}-{\rm div}(e^{2\psi}u_t\nabla u)+e^{2\psi}u_t^2\left(b(t)-\frac{|\nabla \psi|^2}{-\psi_t}-\psi_t\right)\\&&+e^{2\psi}u^2(-\psi_tm(t))-\frac{1}{2}e^{2\psi}u^2\frac{{\rm d}m(t)}{{\rm d}t}+\underbrace{\frac{e^{2\psi}}{-\psi_t}|\psi_t\nabla u-u_t\nabla \psi|^2}_{A_1}\\&=&\frac{\partial(e^{2\psi}F(u))}{\partial t}-2e^{2\psi}F(u)\psi_t.\label{4.1}\end{eqnarray}$

由于$\frac{{\rm d}m(t)}{{\rm d}t}<;0$, $A_1=\frac{e^{2\psi}}{-\psi_t}|\psi_t\nabla u-u_t\nabla \psi|^2\geq\frac{e^{2\psi}}{-\psi_t}\big(\frac{1}{5}\psi_t^2|\nabla u|^2-\frac{1}{4} u_t^2|\nabla\psi|^2\big)$, 由(4.6)式可得

$\begin{eqnarray}\label{4.7}&&\frac{\partial\left[\frac{e^{2\psi}}{2}\left(u_t^2+|\nabla u|^2+m(t)u^2\right)\right]}{\partial t}-{\rm div}(e^{2\psi}u_t\nabla u)+e^{2\psi}u_t^2\bigg(b(t)-\frac{|\nabla \psi|^2}{-\psi_t}-\psi_t-\frac{1}{4}\frac{|\nabla\psi|^2}{-\psi_t}\bigg)\\&&+e^{2\psi}u^2\left(-\psi_tm(t)\right)+\frac{1}{5}e^{2\psi}(-\psi_t)|\nabla u|^2\leq \frac{\partial(e^{2\psi}F(u))}{\partial t}-2e^{2\psi}F(u)\psi_t.\end{eqnarray}$

将上式在$\mathbb{R} ^N$上积分, 有

$\begin{eqnarray}&&\frac {{\rm d}E(t)}{{\rm d}t}+\int_{\mathbb{R} ^N}e^{2\psi}\bigg\{\left(\frac{3+4\delta}{4(2+\delta)}b(t)-\psi_t\right)u_t^2+\frac{1}{5}(-\psi_t)|\nabla u|^2+(-\psi_t)m(t)u^2\bigg\}{\rm d}x\\&\leq& \frac{\rm d}{{\rm d}t}\int_{\mathbb{R} ^N}e^{2\psi}F(u){\rm d}x-\int_{\mathbb{R} ^N}e^{2\psi}\psi_tF(u){\rm d}x\label{4.13}, \end{eqnarray}$

其中

$\begin{equation}E(t)=\int_{\mathbb{R} ^N}e^{2\psi}(u_t^2+|\nabla u|^2+m(t)u^2){\rm d}x.\label{E10}\end{equation}$

在(4.5)式两边同乘$e^{2\psi} u$, 有

$\begin{eqnarray}&&\frac{\partial\left[e^{2\psi}\left(uu_t+\frac{b(t)}{2}u^2\right)\right]}{\partial t}-{\rm div}(e^{2\psi}u\nabla u)\\&&+e^{2\psi}\bigg\{|\nabla u|^2+\bigg[\bigg(-\psi_t+\frac{\alpha}{2(1+t)}\bigg)b(t)+m(t)\bigg]u^2+\underbrace{2u\nabla u\nabla\psi}_{B_1}-2\psi_tuu_t-u_t^2\bigg\}\\&=& e^{2\psi}uf(u).\label{4.4}\end{eqnarray}$

注意到

$\begin{eqnarray}\label{4.5}B_1e^{2\psi}&=&4e^{2\psi}u\nabla u\nabla\psi-e^{2\psi}\nabla(u^2)\nabla\psi\\&=& 4e^{2\psi}u\nabla u\nabla\psi-{\rm div}(e^{2\psi}u^2\nabla\psi)+2e^{2\psi}u^2|\nabla\psi|^2+e^{2\psi}u^2\triangle\psi, \end{eqnarray}$

将(3.4)-(3.6)式和(4.11)式代入(4.10)式, 得到

$\begin{eqnarray}\label{4.10}&&\frac{\partial\left[e^{2\psi}\left(uu_t+\frac{b(t)}{2}u^2\right)\right]}{\partial t}-{\rm div}(e^{2\psi}(u\nabla u+u^2\nabla\psi))\\&&+e^{2\psi}\bigg\{\underbrace{|\nabla u|^2+4u\nabla u\nabla\psi+(-\psi_tb(t)+2|\nabla\psi|^2)}_{B_2}u^2\\&&+\bigg[N_{\alpha}\frac{b(t)}{2(1+t)}+m(t)\bigg]u^2-2\psi_tuu_t-u^2_t\bigg\}= e^{2\psi}uf(u), \end{eqnarray}$

其中$N_{\alpha}=\alpha+\frac{(1+\alpha)N}{2}-2\delta_1.$由于$\frac{|\nabla\psi|^2}{-\psi_t}=\frac{b(t)}{2+\delta}$, 故

$\begin{eqnarray}\label{ss1}B_2&=&|\nabla u|^2+4u\nabla u\nabla\psi+(4+\delta)|\nabla\psi|^2u^2\\&=&\left(1-\frac{4}{4+\frac{\delta}{2}}\right)|\nabla u|^2+\frac{\delta}{2}|\nabla\psi|^2u^2+\bigg|\frac{2}{\sqrt{4+\frac{\delta}{2}}}\nabla u+\sqrt{4+\frac{\delta}{2}}u\nabla\psi\bigg|^2\\&\geq&\delta_2\left(|\nabla u|^2+b(t)(-\psi_t)u^2\right), \end{eqnarray}$

其中当$\delta\rightarrow0$时, $\delta_2={\rm min}\left\{1-\frac{4}{4+\frac{\delta}{2}}, \frac{\delta}{2+\delta}\right\}\rightarrow0$.把(4.13)式代入(4.12)式并在$\mathbb{R} ^N$积分, 得到

$\begin{eqnarray}&&\frac{\rm d}{{\rm d}t}\left[\int_{\mathbb{R} ^N}e^{2\psi}\left(uu_t+\frac{b(t)}{2}u^2\right){\rm d}x\right]+\int_{\mathbb{R} ^N}e^{2\psi}\bigg\{\delta_2\left(|\nabla u|^2+b(t)(-\psi_t)u^2\right)\\&&+\bigg[N_{\alpha}\frac{b(t)}{2(1+t)}+m(t)\bigg]u^2-2\psi_tuu_t-u^2_t\bigg\}{\rm d}x\leq \int_{\mathbb{R} ^N}e^{2\psi}uf(u){\rm d}x\label{4.12}.\end{eqnarray}$

为了处理(4.14)式中的$-u_t^2$项, 将(4.8)式乘以$(t+t_0)^{\alpha}~(t_0\gg1)$, 得

$\begin{eqnarray}\label{4.14}&&\frac{\rm d}{2{\rm d}t}\left[(t+t_0)^{\alpha}E(t)\right]-\frac{\alpha}{2(t+t_0)^{1-\alpha}}E(t)\\&&+\int_{\mathbb{R} ^N}e^{2\psi}\bigg\{\bigg(\frac{3+4\delta}{4(2+\delta)}+(-\psi_t)(t+t_0)^{\alpha}\bigg)u_t^2\\&&+\frac{1}{5}(-\psi_t)(t+t_0)^{\alpha}|\nabla u|^2+(-\psi_t)(t+t_0)^{\alpha}m(t)u^2\bigg\}{\rm d}x\\&\leq& (t+t_0)^{\alpha}\left[\frac{\rm d}{{\rm d}t}\int_{\mathbb{R} ^N}e^{2\psi}F(u){\rm d}x-2\int_{\mathbb{R} ^N}e^{2\psi}\psi_tF(u){\rm d}x\right]=R_1.\end{eqnarray}$

将(4.14)式乘以$\lambda$ ($0<\lambda\ll1$)并与(4.15)式相加, 得到

$\begin{eqnarray}\frac{{\rm d}\widetilde{E}(t)}{{\rm d}t}+\widetilde{H}(t)&=& \frac{\rm d}{{\rm d}t}\bigg(\int_{\mathbb{R} ^N}e^{2\psi}\bigg\{\frac{(t+t_0)^{\alpha}}{2}(u_t^2+|\nabla u|^2+m(t)u^2)+\underbrace{\lambda uu_t}_{D_1}+\frac{\lambda b(t)}{2}u^2\bigg\}{\rm d}x\bigg)\\&&+\int_{\mathbb{R} ^N}e^{2\psi}\bigg\{\bigg(\frac{3+4\delta}{4(2+\delta)}-\lambda+(-\psi_t)(t+t_0)^{\alpha}-\frac{\alpha}{2(t+t_0)^{1-\alpha}}\bigg)u_t^2\\&&+\bigg(\frac{-\psi_t(t+t_0)^{\alpha}}{5}+\lambda\delta_2-\frac{\alpha}{2(t+t_0)^{1-\alpha}}\bigg)|\nabla u|^2\\&&+\bigg(\lambda m(t)+(-\psi_t)(t+t_0)^{\alpha}m(t)-\frac{\alpha}{2(t+t_0)^{1-\alpha}}m(t)\\&&+\lambda\delta_2b(t)(-\psi_t)\bigg)u^2+N_{\alpha}\frac{\lambda b(t)}{2(1+t)}u^2-\underbrace{2\lambda\psi_tuu_t}_{D_2}\bigg\}{\rm d}x\\&\leq& R_1+\int_{\mathbb{R} ^N}e^{2\psi}\lambda uf(u){\rm d}x=R_2\label{4.111}, \end{eqnarray}$

其中

$\begin{eqnarray}\widetilde{E}(t)&=&\int_{\mathbb{R} ^N}e^{2\psi}\bigg\{\frac{(t+t_0)^{\alpha}}{2}(u_t^2+|\nabla u|^2+m(t)u^2)+\underbrace{\lambda uu_t}_{D_1}+\frac{\lambda b(t)}{2}u^2\bigg\}{\rm d}x, \\\widetilde{H}(t)&=&\int_{\mathbb{R} ^N}e^{2\psi}\bigg\{\bigg(\frac{3+4\delta}{4(2+\delta)}-\lambda+(-\psi_t)(t+t_0)^{\alpha}-\frac{\alpha}{2(t+t_0)^{1-\alpha}}\bigg)u_t^2\\&&+\bigg(\frac{-\psi_t(t+t_0)^{\alpha}}{5}+\lambda\delta_2-\frac{\alpha}{2(t+t_0)^{1-\alpha}}\bigg)|\nabla u|^2\\&&+\bigg(\lambda m(t)+(-\psi_t)(t+t_0)^{\alpha}m(t)-\frac{\alpha}{2(t+t_0)^{1-\alpha}}m(t)\\&&+\lambda\delta_2b(t)(-\psi_t)\bigg)u^2+N_{\alpha}\frac{\lambda b(t)}{2(1+t)}u^2-\underbrace{2\lambda\psi_tuu_t}_{D_2}\bigg\}{\rm d}x.\end{eqnarray}$

下面我们估计$\widetilde{E}(t)$$\widetilde{H}(t)$.$E(t)$的定义(4.9), 注意到$|D_1|\leq\frac{\delta_2\lambda b(t)}{2}u^2+\frac{\lambda(t+t_0)^{\alpha}}{2\delta_2}u_t^2$, 容易验证

$\begin{eqnarray}\label{4.17}\widetilde{E}(t)&\leq&\int_{\mathbb{R} ^N}\bigg(1+\frac{\lambda}{\delta_2}\bigg)e^{2\psi}\bigg[\frac{(t+t_0)^{\alpha}}{2}(u_t^2+|\nabla u|^2+m(t)u^2)\bigg]{\rm d}x+\frac{\lambda}{2}(\delta_2+1)b(t)\int_{\mathbb{R} ^N}e^{2\psi}u^2{\rm d}x\\&\leq& C_0(t+t_0)^{\alpha}E(t)+\frac{\lambda}{2}(\delta_2+1)b(t)J(t, u^2)\end{eqnarray}$

$\begin{equation}\widetilde{E}(t)\geq c_0(t+t_0)^{\alpha}E(t)+\frac{\lambda}{2}(-\delta_2+1)b(t)J(t, u^2), \label{4.18}\end{equation}$

其中$C_0=1+\frac{\lambda}{\delta_2}$, $c_0=1-\frac{\lambda}{\delta_2}$,

$\begin{equation}\label{J10}J(t, g)=\int_{\mathbb{R} ^N}e^{2\psi}g(t, x){\rm d}x.\end{equation}$

$D_2$$ |D_2|\leq\frac{1}{2}(-\psi_t)(t+t_0)^{\alpha}u_t^2+2\lambda^2b(t)(-\psi_t)u^2, $

$\begin{eqnarray}\label{4.19}\widetilde{H}(t)&\geq&\int_{\mathbb{R} ^N}e^{2\psi}\bigg\{\bigg[\frac{3+4\delta}{4(2+\delta)}-\lambda+\frac{1}{2}(-\psi_t)(t+t_0)^{\alpha}-\frac{\alpha}{2(t_0+t)^{1-\alpha}}\bigg]u_t^2\\&&+\bigg[\frac{1}{5}(-\psi_t)(t+t_0)^{\alpha}+\lambda\delta_2-\frac{\alpha}{2(t_0+t)^{1-\alpha}}\bigg]|\nabla u|^2\\&&+\bigg[\lambda m(t)+(-\psi_t)(t+t_0)^{\alpha}m(t)-\frac{\alpha}{2(t_0+t)^{1-\alpha}}m(t)\\&&+\lambda\delta_2b(t)(-\psi_t)-2\lambda^2 b(t)(-\psi_t)\bigg]u^2+N_{\alpha}\frac{\lambda b(t)}{2(1+t)}u^2 \bigg\}{\rm d}x, \end{eqnarray}$

所以

$\begin{eqnarray}\widetilde{H}(t)\geq C_1\left(E(t)+(t+t_0)^{\alpha}E_{\psi}(t)+b(t)J_{\psi}(t, u^2)\right)+\frac{N_{\alpha}}{t+1}\cdot\frac{\lambda b(t)}{2}J(t, u^2), \label{4.20}\end{eqnarray}$

其中$C_1=\min\left\{\frac{3+4\delta}{4(2+\delta)}-\lambda-\frac{\alpha}{2t_0^{1-\alpha}}, ~\lambda\delta_2-\frac{\alpha}{2t_0^{1-\alpha}}, ~\lambda-\frac{\alpha}{2t_0^{1-\alpha}}, \ \frac{1}{5}, ~\delta_2\lambda-2\lambda^2\right\}$,

$\begin{equation}E_{\psi}(t)=\int_{\mathbb{R} ^N}e^{2\psi}(-\psi_t)(u_t^2+|\nabla u|^2+m(t)u^2){\rm d}x, \label{E11}\end{equation}$

$\begin{equation}J_{\psi}(t, g)=\int_{\mathbb{R} ^N}e^{2\psi}(-\psi_t)g(t, x){\rm d}x.\label{J11}\end{equation}$

将(4.16)式乘以$(t_0+t)^{\gamma-3\delta_1}$并利用(4.22)式, 得到

$\begin{eqnarray}\label{s}&&(t_0+t)^{\gamma-3\delta_1}\frac{{\rm d}\widetilde{E}(t)}{{\rm d}t}+ C_1\bigg(E(t)+(t+t_0)^{\alpha}E_{\psi}(t)+b(t)J_{\psi}(t, u^2)\bigg)(t_0+t)^{\gamma-3\delta_1}\\&&+(t_0+t)^{\gamma-3\delta_1}\frac{N_{\alpha}}{t+1}\cdot\frac{\lambda b(t)}{2}J(t, u^2)\leq R_2(t+t_0)^{\gamma-3\delta_1}=R_3, \end{eqnarray}$

其中$\gamma=\alpha+\frac{1+\alpha}{2}N$.我们用(4.18)和(4.19)式去处理$(t_0+t)^{\gamma-3\delta_1}\frac{d\widetilde{E}(t)}{{\rm d}t}$.注意到

其中$C_2=C_2(\delta, \lambda)$, 由(4.25)式可得

$\begin{eqnarray}&&\frac{\rm d}{{\rm d}t}\left[C_2(t_0+t)^{\gamma-3\delta_1}\left((t+t_0)^{\alpha}E(t)+b(t)J(t, u^2)\right)\right]\\&&+(t+t_0)^{\gamma-3\delta_1}\bigg(C_1-\frac{C_0(\gamma-3\delta_1)}{(t+t_0)^{1-\alpha}}\bigg)E(t)\\&&+C_1(t+t_0)^{\gamma+\alpha-3\delta}E_{\psi}(t)+C_1(t+t_0)^{\gamma-\alpha-3\delta}J_{\psi}(t, u^2)\\&&+(t+t_0)^{\gamma-3\delta_1}\bigg(\frac{ N_{\alpha}}{t+1}-\frac{(\gamma-3\delta_1)(\delta_2+1)}{t+t_0}\bigg) \frac{\lambda}{2}b(t)J(t, u^2)\leq R_3.\label{4.22}\end{eqnarray}$

对于足够大的$t$, 存在两个正常数$a_0, ~a_1$, 使得$C_1-\frac{C_0(\gamma-3\delta_1)}{(t+t_0)^{1-\alpha}}\geq a_0$$\frac{N_{\alpha}}{t+1} -\frac{(\gamma-3\delta_1)(\delta_2+1)}{t+t_0}\geq\frac{a_1}{1+t}$.$3\delta_1=\epsilon$, 将(4.26)式在$[0, t]$上积分, 得到

$\begin{eqnarray}&&(1+t)^{\gamma+\alpha-\epsilon}E(t)+(t+1)^{\gamma-\alpha-\epsilon}J(t, u^2)\\&&+\int_0^t\bigg\{(\tau+1)^{\gamma-\epsilon}E(\tau)+(\tau+1)^{\gamma+\alpha-\epsilon}E_{\psi}(\tau)\\&&+(\tau+1)^{\gamma-1-\alpha-\epsilon}J(\tau, u^2)+(\tau+1)^{\gamma-\alpha-\epsilon}J_{\psi}(\tau, u^2)\bigg\}{\rm d}\tau\\&\leq& C_3I_0^2+C_3\int_0^tR_3{\rm d}\tau, \label{4.23}\end{eqnarray}$

其中$C_3=C_3(\alpha, \delta, \lambda, t_0, a_1, a_2)$.

$M(t)$的定义(4.4), 知$ M(t)=\sup\limits_{0\leq\tau\leq t}\{(1+\tau)^{\gamma+1 -\varepsilon}E(\tau)+(1+\tau)^{\gamma-\alpha-\varepsilon}J(\tau, u^2)\} $.为了估计$M(t)$中的$(1+t)^{\gamma+1-\varepsilon}E(t)$, 将(4.15)式乘以$(t+t_0)^{\gamma+1-\alpha-\epsilon} $, 得

$\begin{eqnarray}\label{4.44}&&\frac{\rm d}{{\rm d}t}\left[(t+t_0)^{\gamma+1-\epsilon}\frac{E(t)}{2}\right]-(\gamma+1-\epsilon)\frac{E(t)}{2}(t+t_0)^{\gamma-\epsilon}-\alpha\frac{E(t)}{2}(t+t_0)^{\gamma-\epsilon}\\&&+\frac{3+4\delta}{4(2+\delta)}(t+t_0)^{\gamma-\alpha+1-\epsilon}J(t, u_t^2)+\frac{1}{5}(t+t_0)^{\gamma+1-\epsilon}E_{\psi}(t)\leq R_1(t+t_0)^{\gamma+1-\alpha-\epsilon}.\end{eqnarray}$

$[0, t]$积分得

$\begin{eqnarray}&&(t+1)^{\gamma+1-\epsilon}E(t)+\int_0^t\bigg\{-(\tau+1)^{\gamma-\epsilon}E(\tau)+(\tau+1)^{\gamma-\alpha+1-\epsilon}J(\tau, u_{\tau}^2)\\&&+(\tau+1)^{\gamma+1-\epsilon}E_{\psi}(\tau){\rm d}\tau\bigg\}\leq C_4I_0^2+C_4\int_0^t(\tau+1)^{\gamma+1-\alpha-\epsilon}R_1{\rm d}\tau, \label{4.25}\end{eqnarray}$

其中$C_4=C_4(\alpha, \gamma, \lambda, t_0, \epsilon)$.将(4.29)式乘以$\mu$ ($0<\mu\leq1$)并和(4.27)式相加, 得

$\begin{eqnarray}\label{4.50}&&\left((1+t)^{\gamma+\alpha-\epsilon}+\mu(1+t)^{\gamma+1-\epsilon}\right)E(t)+(1+t)^{\gamma-\alpha-\epsilon}J(t, u^2)\\&&+\int_0^t\bigg\{\left((\tau+1)^{\gamma-\epsilon}-\mu(\tau+1)^{\gamma-\epsilon}\right)E(\tau)\\&&+(\tau+1)^{\gamma-1-\alpha-\epsilon}J(\tau, u^2)+(1+\tau)^{\gamma+1-\alpha-\epsilon}J(\tau, u^2_{\tau})\\&&+\mu(1+\tau)^{\gamma+1-\epsilon}E_{\psi}(\tau)+(1+\tau)^{\gamma-\alpha-\epsilon}J_{\psi}(\tau, u^2)\bigg\}{\rm d}\tau\\&\leq&C_5I_0^2+C_5\int_0^t\bigg\{R_2(\tau+1)^{\gamma-\epsilon}+R_1\mu(1+\tau)^{\gamma-\alpha+1-\epsilon}\bigg\}{\rm d}\tau\\&=& C_5I_0^2+C_5R_4, \end{eqnarray}$

其中$C_5=C_5(\alpha, \delta, \epsilon, \lambda, \mu, t_0, a_0, a_1)$, $R_4=\int_0^tR_2(\tau+1)^{\gamma-\epsilon} +R_1\mu(1+\tau)^{\gamma-\alpha+1-\epsilon}{\rm d}\tau$.由(4.30)式知

$\begin{eqnarray}\label{4.51}&&(1+t)^{\frac{(1+\alpha)(N+2)}{2}-\epsilon}E(t)+(1+t)^{\frac{(1+\alpha)N}{2}-\epsilon}J(t, u^2)\\&&+\int_0^t\bigg\{(\tau+1)^{\frac{(1+\alpha)N}{2}+1-\epsilon}J(\tau, u^2_{\tau})+(\tau+1)^{\frac{(1+\alpha)N}{2}+\alpha-\epsilon}J(\tau, |\nabla u|^2)\\&&+\left((1+\tau)^{\frac{(1+\alpha)N}{2}-1-\epsilon}+(1+\tau)^{\frac{(1+\alpha)N}{2}+\alpha-2\beta-\epsilon}\right)J(\tau, u^2)\bigg\}{\rm d}\tau\\&\leq& C_5I_0^2+C_5R_4.\end{eqnarray}$

注意到

$\begin{equation}\label{M}M(t)=\sup\limits_{0\leq\tau\leq t}\left\{(1+\tau)^{\frac{(1+\alpha)(N+2)}{2}-\epsilon}E(\tau)+(1+\tau)^{\frac{(1+\alpha)N}{2}-\epsilon}J(\tau, u^2)\right\}, \end{equation}$

由(4.31)式得

$\begin{equation}\label{s16}M(t)\leq C_5I_0^2+C_5R_4.\end{equation}$

为得到(4.3)式, 我们只需要估计(4.33)中的$R_4$即可.利用(3.4)-(3.6)和(4.30)式, 可得

$\begin{eqnarray}\label{6.10}R_4&\lesssim&(1+t)^{\frac{(1+\alpha)(N+2)}{2}-\epsilon}J(t, |u|^{p+1})+\int_0^t\bigg\{(1+\tau)^{\frac{1+\alpha}{2}N+\alpha-\epsilon}J(\tau, |u|^{p+1})\\&&+(1+\tau)^{\frac{(1+\alpha)(N+2)}{2}-\alpha-\epsilon}J_{\psi}(\tau, |u|^{p+1})\bigg\}{\rm d}\tau.\end{eqnarray}$

由于$p>1, ~\frac{4}{p+1}<2$, 易知

$\begin{equation}\label{s1000}\begin{array}{ll} e^{\frac{4}{p+1}\psi}|\nabla\psi|^2=4a^2\psi e^{\frac{4}{p+1}-2}\frac{1}{(1+t)^{1+\alpha}}e^{2\psi}\lesssim\frac{1}{(1+t)^{1+\alpha}}e^{2\psi}, \\[2mm]e^{\frac{4}{p+1}\psi}\lesssim e^{2\psi}, \end{array}\end{equation}$

$\begin{eqnarray}J(t, |u|^{p+1})&\lesssim& \bigg(\int_{\mathbb{R} ^N}e^{\frac{4}{p+1}\psi}|u|^2{\rm d}x\bigg)^{\frac{1-\sigma}{2}(p+1)}\\&&\times\bigg(\int_{\mathbb{R} ^N}e^{\frac{4}{p+1}\psi}|\nabla\psi|^2 u^2{\rm d}x+\int_{\mathbb{R} ^N}e^{\frac{4}{p+1}\psi}|\nabla u|^2{\rm d}x\bigg)^{\frac{\sigma}{2}(p+1)}\\&\lesssim& (1+t)^{-\frac{(1+\alpha)(p+1)\sigma}{2}}J^{\frac{p+1}{2}}(t, u^2)+ J^{\frac{(p+1)(1-\sigma)}{2}}(t, u^2)J^{\frac{(p+1)\sigma}{2}}(t, |\nabla u|^2), \end{eqnarray}$

其中$\sigma=\sigma(p+1)=N(\frac{1}{2}-\frac{1}{p+1})$.注意到$p>1+\frac{2}{N}$, 利用Gagliardo-Nirenberg插值不等式, 得

$\begin{eqnarray}\label{s21}&&(1+t)^{\frac{(1+\alpha)(N+2)}{2}-\epsilon}J(t, |u|^{p+1})\\&\leq& C_6(1+t)^{-\frac{(1+\alpha)(p+1)\sigma}{2}+\frac{(1+\alpha)(N+2)}{2}-\epsilon-\frac{p+1}{2}(\frac{1+\alpha}{2}N-\epsilon)}M^{\frac{p+1}{2}}(t)\\&&+C_6(1+t)^{\frac{(1+\alpha)(N+2)}{2}-\epsilon-\frac{p+1}{2}\sigma(\frac{(1+\alpha)(N+2)}{2}-\epsilon)-\frac{p+1}{2}(1-\sigma)(\frac{1+\alpha}{2}N-\epsilon)}M^{\frac{p+1}{2}}(t)\\&\leq& 2C_6(1+t)^{-N\frac{1+\alpha}{2}(p-1-\frac{2}{N})+\frac{p-1}{2}\epsilon}M^{\frac{p+1}{2}}(t)\\&\leq& 2C_6M^{\frac{p+1}{2}}(t)\end{eqnarray}$

$\begin{eqnarray}\label{s22}\int_0^t(1+\tau)^{\frac{1+\alpha}{2}N+\alpha-\epsilon}J(\tau, |u|^{p+1}){\rm d}\tau&\leq&C_7M^{\frac{p+1}{2}}(t)\int_0^t(1+\tau)^{-1-\frac{1+\alpha}{2}(p-1-\frac{2}{N})+\frac{p-1}{2}\epsilon}{\rm d}\tau\\&\leq& C_7M^{\frac{p+1}{2}}(t), \end{eqnarray}$

其中$C_6=C_6(\alpha, \gamma, \delta, \epsilon, \lambda, \mu, p, N)$, $C_7=C_7(\alpha, \gamma, \delta, \epsilon, \lambda, \mu, p, N)$.易知

$\begin{eqnarray} J_{\psi}(\tau, |u|^{p+1})&\lesssim& (1+\tau)^{-1}\int_{\mathbb{R} ^N}e^{2\psi}\psi|u|^{p+1}{\rm d}x\\& \lesssim & (1+\tau)^{-1}\int_{\mathbb{R} ^N}e^{(2+\nu)\psi}|u|^{p+1}{\rm d}x~(0<\nu\ll1), \end{eqnarray}$

所以

$\begin{eqnarray}\label{s23}\int_0^t(1+\tau)^{\frac{(1+\alpha)(N+2)}{2}-\alpha-\epsilon}J_{\psi}(\tau, |u|^{p+1}){\rm d}\tau\leq C_8M^{\frac{p+1}{2}}(t), \end{eqnarray}$

其中$C_8=C_8(\alpha, \gamma, \delta, \epsilon, \lambda, \mu, p, N)$.将(4.37), (4.38)和(4.40)式代入(4.34)式, 并利用(4.33)式, 得到

$ \begin{eqnarray}\label{s31} M(t)\lesssim I_0^2+M^{\frac{p+1}{2}}(t). \end{eqnarray}$

命题4.1得证.

接下来证明定理2.2.

定理 2.2 的证明  因初始能量$I_0$足够小, 这意味着存在一个足够小的$\epsilon$, 使得$I_0^2<\epsilon<\epsilon_0, ~\epsilon_0=\min \big\{1, s_0, \frac{s_0}{2C_1}\big\}$, 其中$s_0$是下面确定的一个常数.由命题4.1, 我们有

$\begin{eqnarray}\label{4.26}M(t)\leq C_1\epsilon+C_2M^{\frac{p+1}{2}}(t), \end{eqnarray}$

其中$C_1=C_1(\alpha, \beta, \delta, p, N)$, $C_2=C_2(\alpha, \beta, \delta, p, N)$.

$\begin{eqnarray}\phi(s)=s-C_2s^{\frac{p+1}{2}}, \end{eqnarray}$

$\phi(0)=0, ~\phi'(0)=1$, 知存在一个正常数$s_0$, 使得对任何$s\in[0, s_0]$, 都有$\phi'(s)\geq\frac{1}{2}$, 从而

$\begin{eqnarray}\label{4.27}\phi(s)\leq s\leq2\phi(s), \quad\quad\forall~s\in[0, s_0].\end{eqnarray}$

由(4.42)式知$\phi(M(t))\leq C_1 \epsilon_0\leq\frac{s_0}{2}$, 再利用(4.44)式就得到

$\begin{eqnarray}\label{4.28}\phi(M(t))\leq\frac{s_0}{2}\leq\phi(s_0), \quad \forall~t\in[0, T^*).\end{eqnarray}$

由于$M(t)$$[0, T^*)$上连续且$\phi(s)$$[0, s_0]$上严格单调递增, 故$M(t)\leq s_0$

$\begin{eqnarray}M(t)\leq2\phi(M(t))\leq2C_1\epsilon_0\quad \forall~t\in [0, T^*), \end{eqnarray}$

从而

$\begin{eqnarray}\limsup\limits_{T\rightarrow T^*t\in [0, T]}\left(\|u(t, \cdot)\|_{H^1_{\psi(t, \cdot)}}+\|u_t(t, \cdot)\|_{L^2_{\psi(t, \cdot)}}\right)\lesssim\limsup\limits_{T\rightarrow T^*t\in [0, T]}M(t)\leq2C_1\epsilon_0, \end{eqnarray}$

这和(2.12)式矛盾, 所以$T^*=\infty$, 即$u$是一个整体解.由$M(t)\lesssim\epsilon_0$易得(2.15)式和(2.16)式.定理2.2得证.

5 爆破

本节我们将采用Zhang[25]的试验函数方法证明爆破结果-定理2.3.

定理 2.3 的证明  令(参见文献[25])

$\begin{eqnarray}\psi_R(t, x)=\eta_R(t)\phi_R(r)=\eta\left(\frac{t}{R^{\frac{2}{1+\alpha}}}\right)\phi\left(\frac{r}{R}\right), \quad\quad|x|=r, \end{eqnarray}$

其中$\eta(t), \phi(r)\in C_0^{\infty}$满足

$C$是一个正常数.由$\mu_2=-\mu_1\alpha$$\alpha=2\beta-1$, 方程(1.1)可化为

$\begin{eqnarray}\label{6.4}u_{tt}-\Delta u+\left(b(t)u\right)_t=|u|^p, \end{eqnarray}$

其中$b(t)=\frac{\mu_1}{(1+t)^{\alpha}}$.

$\begin{eqnarray}I_R=\int_{Q_R}|u(t, x)|^p\psi^q_R(t, x){\rm d}x{\rm d}t, \end{eqnarray}$

其中$Q_R=\left[0, R^{\frac{2}{1+\alpha}}\right]\times B_R(0), ~B_R(0)=\{x:\ |x|\leq R\}$, $\frac{1}{p}+\frac{1}{q}=1$.

$\begin{equation}I_R=\int_{Q_R}\left[u_{tt}-\Delta u+\left(b(t)u\right)_t\right]\psi_R^q{\rm d}x{\rm d}t.\end{equation}$

由分部积分得到

所以

$\begin{eqnarray}\label{5.50}I_R&=&-\int_{B_R(0)}\left(\mu_1u_0(x)+u_1(x)\right)\phi_R^q(r){\rm d}x +\int_{Q_R}u(t, x)\left(\psi_R^q(t, x)\right)_{tt}{\rm d}x{\rm d}t\\&&-\int_{Q_R}b(t)u(t, x)\left(\psi_R^q(t, x)\right)_t{\rm d}x{\rm d}t-\int_{Q_R}u(t, x)\Delta\left(\psi_R^q(t, x)\right){\rm d}x{\rm d}t\\&=&-\int_{B_R(0)}\big( \mu_1 u_0(x)+u_1(x)\big) \phi_R^q(r){\rm d}x+J_1+J_2+J_3, \end{eqnarray}$

其中

由假设(2.18)知

$\begin{equation}\label{s10}I_R<\sum\limits_{j=1}^3J_j.\end{equation}$

下面我们估计$J_1, ~J_2$$J_3$.利用Hölder不等式和$\eta(t)$, $\phi(r)$的假设, 有

$\begin{eqnarray}\label{s11}|J_2|&=& \left|\int_{Q_R}b(t)u(t, x)(\psi_R^q(t, x))_t{\rm d}x{\rm d}t\right|\\&=& \left|qR^{-\frac{2}{1+\alpha}}\int_{Q_R}b(t)u(t, x)\phi_R^q(r)\eta_R^{q-1}(t)\eta'_R(t){\rm d}x{\rm d}t\right|\\&\lesssim& R^{-\frac{2}{1+\alpha}}\int_{\hat{Q}_R}b(t)|u(t, x)|\psi_R^{q-1}(t, x){\rm d}x{\rm d}t\\&\lesssim& R^{-\frac{2}{1+\alpha}}\left(\int_{\hat{Q}_R}|u(t, x)|^p\psi_R^{q}(t, x){\rm d}x{\rm d}t\right)^{\frac{1}{p}}\left(\int_{\hat{Q}_R}b^q(t){\rm d}x{\rm d}t\right)^{\frac{1}{q}}\\&\lesssim& \hat{I}_R^{\frac{1}{p}}R^{-{\frac{2}{1+\alpha}}+\frac{N}{q}}\left(\int_{\frac{R^{\frac{2}{1+\alpha}}}{4}}^{R^{\frac{2}{1+\alpha}}}(1+t)^{-q\alpha}{\rm d}t\right)^{\frac{1}{q}}\\&\lesssim& \hat{I}_R^{\frac{1}{p}}R^{-2+\frac{N(1+\alpha)+2}{q(1+\alpha)}}, \end{eqnarray}$

其中

类似地

$\begin{equation}|J_1|\lesssim R^{-{\frac{4}{1+\alpha}}}\int_{\hat{Q}_R}|u(t, x)|\psi_R^{q-1}(t, x){\rm d}x{\rm d}t\lesssim R^{\frac{N(1+\alpha)+2}{q(1+\alpha)}-{\frac{4}{1+\alpha}}}\hat{I}^{\frac{1}{p}}_R\label{s12}\end{equation}$

$\begin{equation}|J_3|\lesssim R^{-2}\int_{\tilde{Q}_R}|u(t, x)|\psi^{q-1}_R(t, x){\rm d}x{\rm d}t\lesssim R^{\frac{N(1+\alpha)+2}{q(1+\alpha)}-2}\tilde{I}^{\frac{1}{p}}_R\label{s13}.\end{equation}$

将(5.7)-(5.9)式代入(5.6)式, 可得

$\begin{eqnarray}&I_R\lesssim\left(\hat{I}_R^{\frac{1}{p}}+\hat{I}_R^{\frac{1}{p}}+\tilde{I}_R^{\frac{1}{p}}\right)R^{\frac{N(1+\alpha)+2}{q(1+\alpha)}-2}\lesssim I_R^{\frac{1}{p}}R^{\frac{N(1+\alpha)+2}{q(1+\alpha)}-2}, \end{eqnarray}$

$\begin{equation}I_R^{1-\frac{1}{p}}\lesssim R^{\frac{N(1+\alpha)+2}{q(1+\alpha)}-2}.\end{equation}$

如果$0<p<p_F(\alpha, N)$, 则当$R\rightarrow\infty$时, $I_R\rightarrow0$, 因此$u\equiv0$.所以

这与假设(2.18)矛盾.

如果$p=p_F(\alpha, N)$, 则有$I_R\leq C$, 其中$C$是一个与$R$无关的常数, 从而

所以$u\equiv0$, 这也与假设(2.18)矛盾.

为得到次临界情况下生命跨度的估计, 令

$\begin{eqnarray}M_R=\int_{Q_R}|u(t, x)|^p\psi_R(t, x){\rm d}x{\rm d}t.\end{eqnarray}$

重复上面的过程, 我们有

$\begin{eqnarray}\label{5.13}M_R&=&-\int_{B_R(0)}\varepsilon\left(\mu_1u_0(x)+u_1(x)\right)\phi_R(r){\rm d}x +\int_{Q_R}u(t, x)\left(\psi_R(t, x)\right)_{tt}{\rm d}x{\rm d}t\\&&-\int_{Q_R}b(t)u(t, x)\left(\psi_R(t, x)\right)_t{\rm d}x{\rm d}t-\int_{Q_R}u(t, x)\Delta\left(\psi_R(t, x)\right){\rm d}x{\rm d}t\\&=&-\int_{B_R(0)}\varepsilon\left(\mu_1u_0(x)+u_1(x)\right)\phi_R(r){\rm d}x\\&&+\int_{Q_R}u(t, x)\bigg(\left(\psi_R(t, x)\right)_{tt}-b(t)\left(\psi_R(t, x)\right)_t-\Delta\left(\psi_R(t, x)\right)\bigg){\rm d}x{\rm d}t.\end{eqnarray}$

由Young不等式知

$\begin{eqnarray}\label{5.134}M_R&\leq&-\int_{B_R(0)}\varepsilon\left(\mu_1u_0(x)+u_1(x)\right)\phi_R(r){\rm d}x+\frac{1}{p}M_R\\&& +\frac{1}{q}\int_{Q_R}\psi^{-\frac{q}{p}}_R(t, x)\bigg|\left(\psi_R(t, x)\right)_{tt}-b(t)\left(\psi_R(t, x)\right)_t-\Delta\left(\psi_R(t, x)\right)\bigg|^{q}{\rm d}x{\rm d}t\\&\leq&-\int_{B_R(0)}\varepsilon\left(\mu_1u_0(x)+u_1(x)\right)\phi_R(r){\rm d}x+\frac{1}{p}M_R+\frac{C(q)}{q}\sum\limits_{i=1}^{3}N_i, \end{eqnarray}$

其中

利用$\eta(t)$$\phi(r)$的假设,我们有

$\begin{eqnarray}\label{5.14}N_2&=& R^{-{\frac{2}{1+\alpha}}q}\int_{Q_R}(\phi_R(r)\eta_R(t))^{-\frac{q}{p}}\left|b(t)\phi_R(r)\eta'_R(t)\right|^{q}{\rm d}x{\rm d}t\\&\leq& CR^{-{\frac{2}{1+\alpha}}q+N}\int_{\frac{R^{\frac{2}{1+\alpha}}}{4}}^{R^{\frac{2}{1+\alpha}}}(1+t)^{-q\alpha}{\rm d}t\\&\leq& CR^{-2q+N+{\frac{2}{1+\alpha}}}, \end{eqnarray}$

其中$C=C(\mu_1, N, p)$.类似地, 可得

$\begin{equation}\label{5.15}N_1\leq CR^{-{\frac{4}{1+\alpha}}q+N+{\frac{2}{1+\alpha}}}\end{equation}$

$\begin{equation}\label{5.16}N_3\leq CR^{-2q+N+\frac{2}{1+\alpha}}, \end{equation}$

其中$C=C(N, p)$.由(5.14)-(5.17)式知

$\begin{equation}\frac{1}{q}M_R\leq 3CR^{-2q+N+{\frac{2}{1+\alpha}}}-\int_{B_R(0)}\varepsilon\left(\mu_1u_0(x)+u_1(x)\right)\phi_R(r){\rm d}x, \end{equation}$

其中$C=C(\mu_1, N, p)$.注意到假设(2.18), 知存在$\overline{R}>0$, 使得

$\begin{eqnarray}\int_{B_R(0)}\big( \mu_1u_0(x)+ u_1(x)\big)\phi_R(r){\rm d}x\geq C_0>0, \quad\forall ~R\geq \overline{R}, \end{eqnarray}$

其中$C_0=C(\alpha, N, u_0, u_1))$.假设$u$是在$ [0, T]\times\mathbb{R} ^N$上的局部解且$T\geq\overline{R}$, 设最大存在时间$T^*=R^{\frac{2}{1+\alpha}}$, 将(5.18)式代入(5.19)式, 有

$\begin{eqnarray}0<\frac{1}{q}M_R&\leq& 3CR^{N+\frac{2}{1+\alpha}-2q}-\int_{B_R(0)}\varepsilon \big( \mu_1u_0(x)+ u_1(x) \big) \phi_R(r){\rm d}x\\&\leq &3CT^{*^{\frac{N(1+\alpha)+2}{2}-q(1+\alpha)}}-C_0\varepsilon.\end{eqnarray}$

所以

$\begin{equation}T^*\leq C\varepsilon^{-\frac{1}{\frac{1}{p-1}(1+\alpha)-\frac{(1+\alpha)N}{2}+\alpha}}, \end{equation}$

其中$C=C\left(\alpha, \mu_1, u_0(x), u_1(x), N, p\right)$.

参考文献

Adams A , John F . Sobolev Spaces. Second Edition. New York: Academic Press, 1975

[本文引用: 1]

Böhme C , Reissig M .

A scale-invariant Klein-Gordon model with time-dependent potential

Ann Univ Ferrara Sez Ⅶ Sci Mat, 2012, 58 (2): 229- 250

URL     [本文引用: 1]

Corduneanu C . Principles of Differential and Integral Equations. Boston: Allyn and Bacon Inc, 1971

[本文引用: 1]

D'Abbicco M .

The threshold of effective damping for semilinear wave equation

Math Methods Appl Sci, 2015, 38 (10): 1032- 1045

URL     [本文引用: 1]

Fujita H .

On the blowing up of solutions of the Cauchy problem for ut=△u+u1+α

J Fac Sci Univ Tokyo Sect I, 1966, 13: 109- 124

[本文引用: 1]

Georgiev V , Lindblad H , Sogge C D .

Weighted Strichartz estimates and global existence for semilinear wave equations

Amer J Math, 1997, 119 (6): 1291- 1319

URL     [本文引用: 1]

Ikeda M, Wakasugi Y. Global well-posedness for the semilinear wave equation with time dependent damping in the overdamping case. 2017, arXiv: 1708.08044

[本文引用: 1]

Ikehata R , Tanizawa K .

Global existence of solutions for semilinear damped-wave equations in ${\Bbb R}$n with noncompactly supported initial data

Nonlinear Anal, 2015, 61 (7): 1189- 1208

[本文引用: 1]

Li T T , Zhou Y .

Breakdown of solutions to □u+ut=|u|1+α

Discrete Contin Dyn Syst, 1995, 1 (4): 503- 520

URL     [本文引用: 1]

Lindblad H , Sogge C D .

On existence and scattering with minimal regularity for semilinear wave equations

J Funct Anal, 1995, 130 (2): 357- 426

URL     [本文引用: 1]

Lin J , Nishihara K , Zhai J .

Critical exponent for the semilinear wave equation with time-dependent damping

Discrete Contin Dyn Syst, 2012, 32 (12): 4307- 4320

URL     [本文引用: 1]

Nishihara K , Zhai J .

Asymptotic behavior of solutions for time dependent damped wave equations

J Math Anal Appl, 2009, 360 (2): 412- 421

URL     [本文引用: 1]

Nishihara K .

Asymptotic behavior of solutions to the semilinear wave equation with time-dependent damping

Tokyo J Math, 2011, 34 (2): 327- 343

URL     [本文引用: 1]

Palmieri A .

Global existence of solutions for semilinear wave equation with scaleinvariant damping and mass in exponentially weighted spaces

J Math Anal Appl, 2018, 461 (2): 1215- 1240

URL     [本文引用: 1]

Sideris T C .

Nonexistence of global solutions to semilinear wave equations in high dimensions

J Differential Equations, 1984, 52 (3): 378- 406

URL     [本文引用: 1]

Strauss W A .

Nonlinear scattering theory at low energy

J Funct Anal, 1981, 41 (1): 110- 133

URL     [本文引用: 1]

Todorova G , Yordanov B .

Critical exponent for a nonlinear wave equation with damping

J Differential Equations, 2001, 174 (2): 464- 489

URL     [本文引用: 3]

Wakasugi Y. Critical Exponent for the Semilinear Wave Equation with Scale Invariant Damping//Ruzhansky M, Turunen V. Fourier Analysis. Boston: Birkhauser, 2014: 375-390

[本文引用: 2]

Wirth J .

Solution representations for a wave equation with weak dissipation

Math Meth Appl Sci, 2004, 27 (1): 101- 124

URL     [本文引用: 1]

Wirth J .

Wave equations with time-dependent dissipation I. Non-effective dissipation

J Differential Equations, 2006, 222 (2): 487- 514

URL    

Wirth J .

Wave equations with time-dependent dissipation Ⅱ. Effective dissipation

J Differential Equations, 2007, 232 (1): 74- 103

URL     [本文引用: 1]

Wirth J. Asymptotic Properties of Solutions to Wave Equations with Time-Dependent Dissipation[D]. Munich: TU Bergakademie Freiberg, 2005

[本文引用: 1]

Yagdjian K , Galstian A .

Fundamental solutions for the Klein-Gordon equation in de Sitter spacetime

Comm Math Phys, 2009, 285 (1): 293- 344

URL     [本文引用: 1]

Yordanov B , Zhang Q S .

Finite time blow up for critical wave equations in high dimensions

J Funct Anal, 2006, 231 (2): 361- 374

URL     [本文引用: 1]

Zhang Q S .

A blow-up result for a nonlinear wave equation with damping:the critical case

C R Acad Sci Paris Sér I Math, 2001, 333 (2): 109- 114

URL     [本文引用: 4]

/