数学杂志  2020, Vol. 40 Issue (2): 219-227   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
凌征球
覃思乾
周泽文
Robin边界条件下的抛物方程爆破分析
凌征球, 覃思乾, 周泽文    
玉林师范学院数学与统计学院, 广西 玉林, 537000
摘要:本文研究了Robin边界条件下含有梯度项的一类抛物方程解的爆破问题.利用合适的Sobolve型与一阶微分不等式等方法,获得了当方程的解发生爆破时其爆破时间的下界估计与解不发生爆破的条件.推广了文献[4]的结果.
关键词爆破    下界    梯度    Robin边界条件    
BLOW-UP ANALYSIS IN A PARABOLIC EQUATION WITH GRADIENT TERM UNDER ROBIN BOUNDARY CONDITION
LING Zheng-qiu, QIN Si-qian, ZHOU Ze-wen    
School of Mathematics and Statistics, Yulin Normal University, Yulin 537000, China
Abstract: In this paper, we study the blow-up questions of solution for a parabolic equation with a gradient term under Robin boundary condition. By using Sobolev-type and first order differential inequalities, we get a lower bound on blow-up time when blow-up does occur and the conditions which ensure that blow-up does not appear, which promote the conclusion of the fourth reference.
Keywords: blow-up     lower bound     gradient     Robin boundary condition    
1 引言

在过去的几十年间, 非线性抛物方程解的爆破问题已经得到了很广泛的研究, 文献[1, 2]有较详细的介绍.其中, 关于解爆破时间$ T $的研究, 仍然有许多文章仅针对$ T $的上界, 但下界的估计却更加符合实际情况.随着Payne[3]开创性的工作以后, 这种讨论爆破时间下界估计的研究得到了许多学者的关注.需要指出的是, Li[4]在齐次Dirichlet边界条件下讨论了方程

$ u_t = \Delta u + u^p - |\nabla u|^q , \qquad (x, t) \in \Omega \times (0, T) $

的爆破问题, 获得了方程的解发生爆破时爆破时间的下界估计.这里拓展文[4]的研究范围, 将在Robin边界条件

$ \frac{\partial u}{\partial \nu} + k u = 0, \quad (x, t) \in \partial\Omega \times (0, T) $

下讨论相同的问题.在这种边界条件下, 文献[4]使用的Sobolev不等式不再可以使用.另外, 由于方程中梯度项的存在, 需要克服一些困难.因此, 当方程的解发生爆破时, 通过建立合适的Sobolev型微分不等式, 获得了爆破时间的下界估计.关于爆破时间下界估计的更多结论见文献[5-11].

2 爆破情况

$ \Omega \subset \mathbb{R}^3 $是具有足够光滑边界$ \partial \Omega $与凸性的有界区域.考虑如下具有非线性梯度项的抛物问题

$ \begin{equation} \left\{ \begin{aligned} & u_t = \Delta u + u^p - |\nabla u|^q , && (x, t)\in \Omega \times (0, T), \\ & \frac{\partial u}{\partial \nu} + k u = 0, && (x, t)\in \partial\Omega \times (0, T), \\ & u(x, 0) = u_0(x) \geq 0, && x\in \Omega, \end{aligned} \right. \end{equation} $ (2.1)

其中$ p, k > 0, q > 2 $, $ \Delta $$ \nabla $表示拉普拉斯和梯度算子, $ T $是可能的爆破时间, $ (\partial u/\partial \nu) $表示在边界向外的法向单位导数, $ u_0(x) $满足适当紧条件的连续函数.

根据最大值原理有$ u\geq 0 $.另外, 除了在某些时刻会发生爆破之外, 还假设问题存在古典正解.当解发生爆破时, 本文的目的是要确定爆破时间的下界估计.为此, 定义下面的辅助函数

$ \begin{equation} \varphi(t) = \int_\Omega u^{8p+1} \mathrm{d}x. \end{equation} $ (2.2)

利用分部积分与边界条件, 简单的计算就可以得到

$ \begin{align} && \frac{1}{8p+1}\varphi^\prime (t) & = \int_\Omega u^{8p} u_t \mathrm{d}x = \int_\Omega u^{8p} \Delta u \mathrm{d}x + \int_\Omega u^{9p}\mathrm{d}x - \int_\Omega u^{8p} |\nabla u|^q \mathrm{d}x \\ && & \leq - 8p \int_\Omega u^{8p-1} |\nabla u|^2 \mathrm{d}x + \int_\Omega u^{9p}\mathrm{d}x - \int_\Omega u^{8p} |\nabla u|^q \mathrm{d}x. \end{align} $ (2.3)

另外, 由于$ | \nabla u^{\frac{8p+1}{2}}|^2 = \frac{(8p+1)^2}{4} u^{8p-1} |\nabla u|^2, \ \quad | \nabla u^{\frac{8p+q}{q}}|^q = (\frac{8p+q}{q})^q u^{8p} |\nabla u|^q. $因此(2.3)式可以写成

$ \begin{eqnarray} \frac{1}{8p+1}\varphi^\prime (t) &\leq & ( \frac{1}{2} \int_\Omega u^{9p} \mathrm{d}x - (\frac{q}{8p+q})^q \int_\Omega |\nabla u^{\frac{8p+q}{q}}|^q \mathrm{d}x ) \\ && +( \frac{1}{2} \int_\Omega u^{9p} \mathrm{d}x - \frac{32p}{(8p+1)^2} \int_\Omega |\nabla u^{\frac{8p+1}{2}}|^2 \mathrm{d}x ) = J_1 + J_2. \end{eqnarray} $ (2.4)

下面考虑$ J_1 $.由于$ |\nabla u^{\frac{8p+q}{q}}| $的指数是$ q $而不是$ 2 $, 因此需要一点技巧去克服这个困难使得其指数变成$ 2 $.为了方便, 令$ v = u^{\frac{8p+q}{q}} $并通过Hölder不等式就有

$ \begin{equation} \int_\Omega |\nabla v^{\frac{q}{2}}|^2 \mathrm{d}x = \frac{q^2}{4} \int_\Omega v^{q-2} |\nabla v|^2 \mathrm{d}x \leq \frac{q^2}{4} ( \int_\Omega v^q \mathrm{d}x )^{\frac{q-2}{q}} ( \int_\Omega |\nabla v|^q \mathrm{d}x )^{\frac{2}{q}}. \end{equation} $ (2.5)

再从Poincare不等式, $ \lambda_1 \int_\Omega v^q \mathrm{d}x \leq \int_\Omega |\nabla v^{\frac{q}{2}}|^2 \mathrm{d}x - \int_{\partial \Omega} v^{\frac{q}{2}} \frac{\partial v^{\frac{q}{2}} }{\partial \nu} \mathrm{d}S, $这里$ \lambda_1 $是下列弹性膜问题

$ \Delta w + \lambda w = 0, \quad x \in \Omega; \quad \frac{\partial w}{\partial \nu} + k w = 0, \quad x \in \partial \Omega $

的第一正特征值.利用Robin边界条件, 又有

$ \begin{equation} \lambda_1 \int_\Omega v^q \mathrm{d}x \leq \int_\Omega |\nabla v^{\frac{q}{2}}|^2 \mathrm{d}x + \frac{k(8p+q)}{2} \int_{\partial \Omega} v^{q} \mathrm{d}S. \end{equation} $ (2.6)

再根据散度定理

$ \begin{align} && \int_{\partial\Omega} x_i \nu_i v^q \mathrm{d}S & \leq \int_\Omega x_{i, i} v^q \mathrm{d}x + q \int_\Omega x_i v^{q-1} |v_{, i}| \mathrm{d}x \\ && & \leq 3 \int_\Omega v^q \mathrm{d}x + q ( \int_\Omega x_i x_i v^q \mathrm{d}x )^{\frac{1}{2}}( \int_\Omega v^{q-2}|\nabla v|^2 \mathrm{d}x )^{\frac{1}{2}}. \end{align} $ (2.7)

利用$ \Omega $的凸性可以令

$ l_0 = \min\limits_{\partial\Omega}(\mathbf{x}\cdot { {{ ν}}}), \quad d^2 = \max\limits_{\Omega}|\mathbf{x}| $

以及Young不等式得到

$ \begin{equation} \int_{\partial\Omega} v^q \mathrm{d}S \leq ( \frac{3}{l_0} + \frac{d}{l_0} \theta ) \int_\Omega v^q \mathrm{d}x + \frac{d}{l_0 \theta} \int_\Omega |\nabla v^{\frac{q}{2}}|^2 \mathrm{d}x, \end{equation} $ (2.8)

这里$ \theta $是任意正的常数.这样从(2.8)式和(2.6)式就可以获得

$ \begin{equation} ( \lambda_1 - \frac{3k(8p+q)}{2 l_0} - \frac{kd(8p+q)}{2 l_0} \theta)\int_\Omega v^q \mathrm{d}x \leq ( 1+\frac{kd(8p+q)}{2 l_0 \theta} )\int_\Omega |\nabla v^{\frac{q}{2}}|^2 \mathrm{d}x. \end{equation} $ (2.9)

如果$ \Omega $, 也就是$ \lambda_1 $满足

$ \begin{equation} \lambda_1 - \frac{3k(8p+q)}{2 l_0} >0, \end{equation} $ (2.10)

然后选择充分小的$ \theta $, 使得

$ C_1 = \lambda_1 - \frac{3k(8p+q)}{2 l_0} - \frac{kd(8p+q)}{2 l_0} \theta >0. $

这样(2.9)式就可以写成:

$ \begin{equation} C_1 \int_\Omega v^q \mathrm{d}x \leq C_0 \int_\Omega |\nabla v^{\frac{q}{2}}|^2 \mathrm{d}x, \end{equation} $ (2.11)

其中$ C_0 = 1+\frac{kd(8p+q)}{2 l_0 \theta} $.把(2.11)式代入(2.5)式得到

$ \begin{equation} \int_\Omega |\nabla v^{\frac{q}{2}}|^2 \mathrm{d}x \leq (\frac{q}{2} )^q (\frac{C_0}{C_1} )^{\frac{q-2}{2}}\int_\Omega |\nabla v|^q \mathrm{d}x. \end{equation} $ (2.12)

因此根据$ J_1 $的定义就有

$ \begin{equation} J_1 \leq \frac{1}{2}\int_\Omega u^{9p} \mathrm{d}x - ( \frac{2}{8p+q} )^q ( \frac{C_1}{C_0} )^{\frac{q-2}{2}}\int_\Omega |\nabla u^{\frac{8p+q}{2}}|^2 \mathrm{d}x. \end{equation} $ (2.13)

下面考虑(2.13)式右边的第一项.如果$ p > 1 $, 利用Hölder不等式和$ a^r b^q \leq ra + qb, r+q = 1, a, b > 0 $得到

$ \begin{align} && \int_\Omega u^{9p}\mathrm{d}x & \leq ( \int_\Omega u^{8p+1}\mathrm{d}x )^{\frac{8p+1-\sigma_1}{8p+1}} (\int_\Omega u^{\frac{9(8p+q)}{8}}\mathrm{d}x )^{\frac{\sigma_1}{8p+1}} \\ && & \leq \frac{8p+1-\sigma_1}{8p+1} \int_\Omega u^{8p+1} \mathrm{d}x + \frac{\sigma_1}{8p+1} \int_\Omega u^{\frac{9(8p+q)}{8}}\mathrm{d}x \\ && & = \frac{8p+1-\sigma_1}{8p+1} \varphi(t) + \frac{\sigma_1}{8p+1} \int_\Omega u^{\frac{9(8p+q)}{8}}\mathrm{d}x, \end{align} $ (2.14)

这里

$ \sigma_1 = \frac{8(p - 1)(8p+1)}{8(p-1)+9q} \in (0, 8p+1). $

借助文献[3]的思想, 下面寻找一个合适的Sobolev型不等式来计算$ u^{9(8p+q)/8} $的积分.首先, 令$ x_{im} $$ x_{iM} $分别表示$ \Omega $在坐标$ x_i $上的最小值与最大值, $ D_z $表示$ \Omega $与平面$ x_3 = z $相交的截面.其次, 为了方面计算, 设$ \omega: = u^{(8p+q)/4} $.根据Schwarz不等式得到

$ \begin{equation} \int_\Omega \omega^{\frac{9}{2}} \mathrm{d}x = \int_{x_{3m}}^{x_{3M}} (\int_{D_z} \omega^{\frac{9}{2}} \mathrm{d}A ) \mathrm{d}x_3 \leq \int_{x_{3m}}^{x_{3M}} (\int_{D_z} \omega^3 \mathrm{d}A \int_{D_z} \omega^6 \mathrm{d}A )^{\frac{1}{2}} \mathrm{d}x_3. \end{equation} $ (2.15)

$ P = (\bar{x}_1, \bar{x}_2, z) $是截面$ D_z $上的一个点, $ P_1 $$ P_2 $表示截面$ D_z $的边界与直线$ x_2 = \bar{x}_2 $的两个交点.同样, $ Q_1 $$ Q_2 $表示$ D_z $的边界与直线$ x_1 = \bar{x}_1 $的两个交点, 那么

$ \begin{aligned} \omega^3(P) = \omega^3(P_1) + 3 \int^P_{P_1} \omega ^2 \omega_{x_1} \mathrm{d} x_1, \; \; \; \; \omega^3(P) = \omega^3(P_2) - 3 \int^P_{P_2} \omega ^2 \omega_{x_1} \mathrm{d} x_1. \end{aligned} $

这样可以获得

$ \begin{equation} \omega^3(P) \leq \frac{1}{2} [ \omega^3(P_1) + \omega^3(P_2) ] + \frac{3}{2} \int^{P_2}_{P_1} \omega ^2 |\omega_{x_1}| \mathrm{d} x_1. \end{equation} $ (2.16)

类似的,

$ \begin{equation} \omega^3(P) \leq \frac{1}{2} [ \omega^3(Q_1) + \omega^3(Q_2) ] + \frac{3}{2} \int^{Q_2}_{Q_1} \omega ^2 |\omega_{x_2}| \mathrm{d} x_2. \end{equation} $ (2.17)

把上面两式左右相乘并在$ D_z $上积分, 即有

$ \begin{eqnarray*} \int_{D_z} \omega^6(P) \mathrm{d}A & \leq& ( \frac{1}{2} \int_{x_{2m}}^{x_{2M}} [ \omega^3(P_1) + \omega^3(P_2) ]\mathrm{d} x_2 + \frac{3}{2} \int_{D_z}\omega^2 |\omega_{x_1}| \mathrm{d}A ) \\ &&\times ( \frac{1}{2} \int_{x_{1m}}^{x_{1M}} [ \omega^3(Q_1) + \omega^3(Q_2) ]\mathrm{d} x_1 + \frac{3}{2} \int_{D_z}\omega^2 |\omega_{x_2}| \mathrm{d}A ) \\ &\leq& ( \frac{1}{2} \int_{\partial D_z} \omega^3 |\nu_1| \mathrm{d} s + \frac{3}{2} \int_{D_z}\omega^2 |\omega_{x_1}| \mathrm{d}A )\\ &&\times ( \frac{1}{2} \int_{\partial D_z} \omega^3 |\nu_2| \mathrm{d} s + \frac{3}{2} \int_{D_z}\omega^2 |\omega_{x_2}| \mathrm{d}A ). \end{eqnarray*} $

再代回(2.15)式得到

$ \begin{eqnarray} \int_\Omega \omega^{\frac{9}{2}}\mathrm{d}x & \leq& \int_{x_{3m}}^{x_{3M}} ( \int_{D_z}\omega^3 \mathrm{d}A )^{\frac{1}{2}} \times (\frac{1}{2} \int_{\partial D_z} \omega^3 |\nu_1| \mathrm{d} s + \frac{3}{2} \int_{D_z}\omega^2 |\omega_{x_1}| \mathrm{d}A )^{\frac{1}{2}} \\ &&\times ( \frac{1}{2} \int_{\partial D_z} \omega^3 |\nu_2| \mathrm{d} s + \frac{3}{2} \int_{D_z}\omega^2 |\omega_{x_2}| \mathrm{d}A )^{\frac{1}{2}} \mathrm{d} x_3 \\ &\leq& \max\limits_{z} ( \int_{D_z}\omega^3 \mathrm{d}A )^{\frac{1}{2}} \int_{x_{3m}}^{x_{3M}} (\frac{1}{2} \int_{\partial D_z} \omega^3 |\nu_1| \mathrm{d} s + \frac{3}{2} \int_{D_z}\omega^2 |\omega_{x_1}| \mathrm{d}A )^{\frac{1}{2}} \\ && \times ( \frac{1}{2} \int_{\partial D_z} \omega^3 |\nu_2| \mathrm{d} s + \frac{3}{2} \int_{D_z}\omega^2 |\omega_{x_2}| \mathrm{d}A )^{\frac{1}{2}} \mathrm{d} x_3 \\ &\leq& \max\limits_{z} ( \int_{D_z}\omega^3 \mathrm{d}A )^{\frac{1}{2}} F_1^{\frac{1}{2}} F_2^{\frac{1}{2}}, \end{eqnarray} $ (2.18)

其中

$ \begin{equation} F_i = \frac{1}{2} \int_{\partial \Omega} \omega^3 |\nu_i| \mathrm{d} S + \frac{3}{2} \int_{\Omega}\omega^2 |\omega_{x_i}| \mathrm{d}x, \quad i = 1, 2. \end{equation} $ (2.19)

最后, 为了计算$ \omega^3 $$ D_z $上的积分, 用$ \Omega^+ $表示$ \Omega $$ D_z $上的部分, 而相应的边界是$ \partial \Omega^+ $, 而下面部分与边界则分别用$ \Omega^- $$ \partial \Omega^- $表示.这样根据散度定理,

$ \begin{aligned} & \int_{D_z} \omega^3 \mathrm{d}A = \int_{\partial \Omega^+} \omega^3 \nu_3 \mathrm{d} S - 3 \int_{\Omega^+} \omega^2 \omega_{x_3} \mathrm{d}x, \\ & \int_{D_z} \omega^3 \mathrm{d}A = -\int_{\partial \Omega^-} \omega^3 \nu_3 \mathrm{d} S + 3 \int_{\Omega^-} \omega^2 \omega_{x_3} \mathrm{d}x, \end{aligned} $

由此得到

$ \int_{D_z} \omega^3 \mathrm{d}A \leq \frac{1}{2} \int_{\partial \Omega} \omega^3 |\nu_3| \mathrm{d} S + \frac{3}{2} \int_{\Omega}\omega^2 |\omega_{x_3}| \mathrm{d}x. $

即可以扩展(2.19)式到$ i = 3 $的情况.又由于

$ F_1^{\frac{1}{3}} F_2^{\frac{1}{3}} F_3^{\frac{1}{3}} \leq \frac{1}{3}(F_1 + F_2 + F_3), \; \; \; \; \; \; \; (a+b+c)^2 \leq 3(a^2 + b^2 + c^2), \quad a, b, c>0. $

因此从(2.18)式得

$ \begin{align} && \int_\Omega \omega^{\frac{9}{2}} \mathrm{d}x & \leq F_1^{\frac{1}{2}} F_2^{\frac{1}{2}} F_3^{\frac{1}{2}} \leq \frac{1}{3^{3/2}} (F_1 + F_2 + F_3)^{\frac{3}{2}} \\ && & = \frac{1}{3^{3/2}} [ (A_1 + A_2 + A_3] + (B_1 + B_2 + B_3) ]^{\frac{3}{2}} \\ && & \leq \frac{1}{3^{3/4}} ( ( A_1^2 + A_2^2 + A_3^2 )^{1/2} + ( B_1^2 + B_2^2 + B_3^2 )^{1/2} )^{\frac{3}{2}}, \end{align} $ (2.20)

其中$ A_i = \frac{1}{2} \int_{\partial\Omega} \omega^3 |\nu_i| \mathrm{d}S, B_i = \frac{3}{2}\int_\Omega \omega^2 |\omega_{x_i}| \mathrm{d}x, i = 1, 2, 3. $

另外根据Schwarz不等式, 有

$ \begin{aligned} A_1^2 + A_2^2 + A_3^2 = \frac{1}{4} \sum\limits_{i = 1}^3 (\int_{\partial\Omega}\omega^3 |\nu_i|\mathrm{d}S )^2 \leq \frac{1}{4} \sum\limits_{i = 1}^3 \int_{\partial\Omega}\omega^3 \mathrm{d}S \int_{\partial\Omega}\omega^3 \nu_i^2 \mathrm{d}S = \frac{1}{4} (\int_{\partial\Omega}\omega^3 \mathrm{d}S )^2. \end{aligned} $

类似地,

$ \begin{aligned} && B_1^2 + B_2^2 + B_3^2 & = \frac{9}{4} \sum\limits_{i = 1}^3 (\int_{\Omega}\omega^2 |\omega_{x_i}|\mathrm{d}x )^2 \\ && & \leq \frac{9}{4} \sum\limits_{i = 1}^3 \int_{\Omega}\omega^2 \mathrm{d}x \int_{\Omega}\omega^2 |\omega_{x_i}|^2 \mathrm{d}x = \frac{9}{16} \int_{\Omega}\omega^2 \mathrm{d}x \int_\Omega |\nabla \omega^2 |^2 \mathrm{d}x. \end{aligned} $

因此(2.20)式可以变成

$ \begin{equation} \int_\Omega \omega^{\frac{9}{2}}\mathrm{d}x \leq \frac{1}{3^{3/4}} ( \frac{1}{2} \int_{\partial\Omega} \omega^3 \mathrm{d}S + \frac{3}{4} (\int_\Omega \omega^2 \mathrm{d}x )^{1/2} (\int_\Omega |\nabla\omega^2|^2 \mathrm{d}x )^{1/2} )^{\frac{3}{2}}. \end{equation} $ (2.21)

类似于(2.7)与(2.8)式, 可以得到

$ \begin{equation} \int_{\partial\Omega} \omega^3 \mathrm{d}S \leq \frac{3}{l_0} \int_{\Omega} \omega^3 \mathrm{d}x + \frac{3d}{2 l_0} (\int_\Omega \omega^2 \mathrm{d}x )^{1/2} (\int_\Omega |\nabla\omega^2|^2 \mathrm{d}x )^{1/2}. \end{equation} $ (2.22)

把(2.22)式代入(2.21)式,

$ \begin{equation} \int_\Omega \omega^{\frac{9}{2}}\mathrm{d}x \leq \frac{1}{3^{3/4}} ( \frac{3}{2l_0} \int_{\Omega} \omega^3 \mathrm{d}x + \frac{3}{4} (\frac{d}{l_0}+1 ) (\int_\Omega \omega^2 \mathrm{d}x )^{1/2} (\int_\Omega |\nabla\omega^2|^2 \mathrm{d}x )^{1/2} )^{\frac{3}{2}}. \end{equation} $ (2.23)

再使用不等式$ (a+b)^{\frac{3}{2}} \leq 2^{\frac{1}{2}}(a^{\frac{3}{2}} + b^{\frac{3}{2}}), a, b > 0, $ (2.23)式就可以变成

$ \begin{equation} \int_\Omega \omega^{\frac{9}{2}}\mathrm{d}x \leq \frac{2^{1/2}}{3^{3/4}} ( ( \frac{3}{2l_0} \int_{\Omega} \omega^3 \mathrm{d}x )^{\frac{3}{2}}+ (\frac{3}{4} )^{\frac{3}{2}} (\frac{d}{l_0}+1 )^{\frac{3}{2}} (\int_\Omega \omega^2 \mathrm{d}x )^{3/4} (\int_\Omega |\nabla\omega^2|^2 \mathrm{d}x )^{3/4} ). \end{equation} $ (2.24)

再依次使用Young和Hölder不等式, 还有

$ \begin{eqnarray} (\int_\Omega \omega^2 \mathrm{d}x )^{3/4} (\int_\Omega |\nabla\omega^2|^2 \mathrm{d}x )^{3/4} &\leq& \frac{1}{4}\varepsilon_1^{-3} (\int_\Omega \omega^2 \mathrm{d}x )^{3} + \frac{3}{4}\varepsilon_1 \int_\Omega |\nabla\omega^2|^2 \mathrm{d}x, \end{eqnarray} $ (2.25)
$ \begin{eqnarray} \int_\Omega \omega^3 \mathrm{d}x &\leq& ( \int_\Omega \omega^{\frac{4(8p+1)}{8p+q}} \mathrm{d}x )^{\frac{3(8p+q)}{4(8p+1)}} |\Omega|^{1-\frac{3(8p+q)}{4(8p+1)}}, \end{eqnarray} $ (2.26)
$ \begin{eqnarray} \int_\Omega \omega^2 \mathrm{d}x &\leq& ( \int_\Omega \omega^{\frac{4(8p+1)}{8p+q}} \mathrm{d}x )^{\frac{8p+q}{2(8p+1)}} |\Omega|^{1-\frac{8p+q}{2(8p+1)}}, \end{eqnarray} $ (2.27)

这里$ \varepsilon_1 $是任意正的常数, 并且还假设$ p > \frac{3q-4}{8} $.

结合(2.13), (2.14), (2.24)–(2.27)式, 并选择$ \varepsilon_1 $满足

$ \frac{\sigma_1}{2(8p+1)}\frac{2^{1/2}}{3^{3/4}} (\frac{3}{4} )^{\frac{5}{2}} (\frac{d}{l_0}+1 )^{\frac{3}{2}} \varepsilon_1 - (\frac{2}{8p+q} )^{q} (\frac{C_1}{C_0} )^{\frac{q-2}{2}} = 0, $

由此就可以获得

$ \begin{eqnarray} J_1 & \leq &\frac{8p+1-\sigma_1}{2(8p+1)} \varphi(t) + \frac{\sigma_1}{2(8p+1)}\frac{2^{1/2}}{3^{3/4}} (\frac{3}{2l_0} )^{\frac{3}{2}}|\Omega|^{\frac{3}{2}( 1- \frac{3(8p+q)}{4(8p+1)})} \varphi(t)^{\frac{9(8p+q)}{8(8p+1)}} \\ &&+ \frac{\sigma_1}{2(8p+1)}\frac{2^{1/2}}{3^{3/4}} (\frac{3}{4} )^{\frac{3}{2}} ( \frac{d}{l_0} + 1 )^{\frac{3}{2}} \frac{1}{4} \varepsilon_1^{-3} |\Omega|^{3( 1- \frac{8p+q}{2(8p+1)})} \varphi(t)^{\frac{3(8p+q)}{2(8p+1)}}. \end{eqnarray} $ (2.28)

下面估算$ J_2 $.类似于(2.14)式, 首先有

$ \int_\Omega u^{9p} \mathrm{d}x \leq \frac{9}{8p+1}\varphi(t) + \frac{8(p-1)}{8p+1}\int_\Omega u^{\frac{9(8p+1)}{8}}\mathrm{d}x. $

然后从$ J_2 $的定义得到

$ \begin{equation} J_2 \leq \frac{9}{2(8p+1)}\varphi(t) + \frac{4(p-1)}{8p+1}\int_\Omega u^{\frac{9(8p+1)}{8}}\mathrm{d}x - \frac{32p}{(8p+1)^2}\int_\Omega |\nabla u^{\frac{8p+1}{2}}|^2 \mathrm{d}x. \end{equation} $ (2.29)

类似上面的讨论, 也可以估算$ u^{\frac{9(8p+1)}{8}} $的积分,

$ \begin{eqnarray} \int_\Omega u^{\frac{9(8p+1)}{8}}\mathrm{d}x & \leq& \frac{2^{1/2}}{3^{3/4}} (\frac{3}{2l_0} )^{\frac{3}{2}}|\Omega|^{\frac{3}{8}} \varphi(t)^{\frac{9}{8}} + \frac{2^{1/2}}{3^{3/4}} (\frac{3}{4} )^{\frac{3}{2}} ( \frac{d}{l_0} + 1 )^{\frac{3}{2}} \frac{1}{4} \varepsilon_2^{-3} |\Omega|^{\frac{3}{2}} \varphi(t)^{\frac{3}{2}} \\ &&+ \frac{2^{1/2}}{3^{3/4}} (\frac{3}{4} )^{\frac{5}{2}} ( \frac{d}{l_0} + 1 )^{\frac{3}{2}} \varepsilon_2 \int_\Omega |\nabla u^{\frac{8p+1}{2}}|^2 \mathrm{d}x, \end{eqnarray} $ (2.30)

这里$ \varepsilon_2 $也是一个任意正的常数.把(2.30)式代进(2.29)式, 并且选择$ \varepsilon_2 $满足

$ 4(p-1) \frac{2^{1/2}}{3^{3/4}} (\frac{3}{4} )^{\frac{5}{2}} ( \frac{d}{l_0} + 1 )^{\frac{3}{2}} \varepsilon_2 - \frac{32p}{8p+1} = 0, $

(2.29)式就变成

$ \begin{align} && J_2 \leq & \frac{9}{2(8p+1)}\varphi(t) + \frac{4(p-1)}{8p+1} \frac{2^{1/2}}{3^{3/4}} (\frac{3}{2l_0} )^{\frac{3}{2}}|\Omega|^{\frac{3}{8}} \varphi(t)^{\frac{9}{8}} \\ && & \qquad \qquad \qquad + \frac{4(p-1)}{8p+1} \frac{2^{1/2}}{3^{3/4}} (\frac{3}{4} )^{\frac{3}{2}} ( \frac{d}{l_0} + 1 )^{\frac{3}{2}} \frac{1}{4} \varepsilon_2^{-3} |\Omega|^{\frac{3}{2}} \varphi(t)^{\frac{3}{2}}. \end{align} $ (2.31)

把(2.28)和(2.31)式代入(2.4)式, 就可以获得关于函数$ \varphi $的一阶微分不等式

$ \begin{equation} \varphi^\prime(t) \leq K_1 \varphi(t) + K_2 \varphi(t)^{\frac{9(8p+q)}{8(8p+1)}} + K_3 \varphi(t)^{\frac{3(8p+q)}{2(8p+1)}} + K_4 \varphi(t)^{\frac{9}{8}} + K_5 \varphi(t)^{\frac{3}{2}}, \end{equation} $ (2.32)

其中

$ \begin{equation} \left\{ \begin{aligned} & K_1 = \frac{9p(8p+1)}{8(p-1)+9p}+\frac{9}{2}, \\ & K_2 = \frac{\sigma_1}{2} \frac{2^{1/2}}{3^{3/4}} (\frac{3}{2 l_0} )^{\frac{3}{2}}|\Omega|^{\frac{3}{2}( 1- \frac{3(8p+q)}{4(8p+1)})}, \\ & K_3 = \frac{\sigma_1}{8} \frac{2^{1/2}}{3^{3/4}} (\frac{3}{4} )^{\frac{3}{2}} ( \frac{d}{l_0} + 1 )^{\frac{3}{2}}\varepsilon_1^{-3} |\Omega|^{3( 1- \frac{8p+q}{2(8p+1)})}, \\ & K_4 = 4(p-1) \frac{2^{1/2}}{3^{3/4}} ( \frac{3}{2 l_0} )^{\frac{3}{2}}|\Omega|^{\frac{3}{8}}, \\ & K_5 = (p-1) \frac{2^{1/2}}{3^{3/4}} ( \frac{3}{4} )^{\frac{3}{2}} ( \frac{d}{l_0} + 1 )^{\frac{3}{2}}\varepsilon_2^{-3} |\Omega|^{ \frac{3}{2}}. \end{aligned} \right. \end{equation} $ (2.33)

再积分得到

$ \begin{equation} \int_{\varphi(0)}^{\varphi(t)} \frac{\mathrm{d} \eta}{K_1 \eta + K_2 \eta^{\frac{9(8p+q)}{8(8p+1)}} + K_3 \eta^{\frac{3(8p+q)}{2(8p+1)}} + K_4 \eta^{\frac{9}{8}} + K_5 \eta^{\frac{3}{2}} } \leq t, \end{equation} $ (2.34)

由此得到爆破时间$ T $的一个下界估计

$ \begin{equation} T \geq T_0 = \int_{\varphi(0)}^{\infty} \frac{\mathrm{d} \eta}{K_1 \eta + K_2 \eta^{\frac{9(8p+q)}{8(8p+1)}} + K_3 \eta^{\frac{3(8p+q)}{2(8p+1)}} + K_4 \eta^{\frac{9}{8}} + K_5 \eta^{\frac{3}{2}} }, \end{equation} $ (2.35)

其中$ \varphi(0) = \int_\Omega [u_0(x)]^{8p+1} \mathrm{d}x. $

把上面的分析总结成如下定理.

定理1   如果$ p > \max\{ 1, \ \frac{3q-4}{8}\} $, 以及(2.10)式成立, 那么当问题(2.1)的非负古典解$ u $在(2.2)式测度$ \varphi $的意义下有限时刻$ T $发生爆破时, 则$ T $的下界$ T_0 $由(2.35)式给出.

3 非爆破情况

这里考虑$ p < 1 $的情况, 此时问题(2.1)的解不会在有限时刻内发生爆破.令$ \varphi(t) $依然是(2.2)式定义的辅助函数, 从(2.4)式得到

$ \begin{equation} \varphi^\prime (t) \leq (8p+1)(J_1 + J_2). \end{equation} $ (3.1)

类似(2.11)式的讨论, 有

$ \begin{equation} \int_\Omega |\nabla u^{\frac{8p+q}{2}}|^2 \mathrm{d}x \geq C_2 \int_\Omega u^{8p+q} \mathrm{d}x, \end{equation} $ (3.2)

其中

$ C_2 = \frac{2 l_0 \lambda_1 - 3k(16p+q) - dk(16p+q)\theta}{2 l_0 + dk \theta^{-1} (16p+q)} > 0. $

因此从(2.13)式得到

$ \begin{equation} J_1 \leq \frac{1}{2}\int_\Omega u^{9p} \mathrm{d}x - (\frac{2}{8p+q} )^q (\frac{C_1}{C_0} )^{\frac{q-2}{2}}C_2 \int_\Omega u^{8p+q} \mathrm{d}x. \end{equation} $ (3.3)

另外, 根据Hölder不等式, 也有

$ \begin{equation} \int_\Omega u^{9p} \mathrm{d}x \leq (\int_\Omega u^{8p+1}\mathrm{d}x )^{\frac{9p}{8p+1}}|\Omega|^{\frac{1-p}{8p+1}} = \varphi(t)^{\frac{9p}{8p+1}}|\Omega|^{\frac{1-p}{8p+1}} \end{equation} $ (3.4)

$ \begin{equation} \int_\Omega u^{8p+q} \mathrm{d}x \geq ( \int_\Omega u^{8p+1} \mathrm{d}x )^{\frac{8p+q}{8p+1}} |\Omega|^{\frac{1-q}{8p+1}} = \varphi(t)^{\frac{8p+q}{8p+1}} |\Omega|^{\frac{1-q}{8p+1}}. \end{equation} $ (3.5)

把(3.4), (3.5)式代入(3.3)式得

$ \begin{align} && J_1 & \leq \frac{1}{2}|\Omega|^{\frac{1-p}{8p+1}}\varphi(t)^{\frac{9p}{8p+1}} - (\frac{2}{8p+q} )^q (\frac{C_1}{C_0} )^{\frac{q-2}{2}}C_2 |\Omega|^{\frac{1-q}{8p+1}}\varphi(t)^{\frac{8p+q}{8P+1}} \\ && & = \varphi(t)^{\frac{9p}{8p+1}} ( \frac{1}{2}|\Omega|^{\frac{1-p}{8p+1}} - (\frac{2}{8p+q} )^q (\frac{C_1}{C_0} )^{\frac{q-2}{2}}C_2 |\Omega|^{\frac{1-q}{8p+1}}\varphi(t)^{\frac{q-p}{8P+1}} ). \end{align} $ (3.6)

下面估算$ J_2 $.类似(2.11)式的讨论也得到

$ \begin{equation} \int_\Omega |\nabla u^{\frac{8p+1}{2}}|^2 \mathrm{d}x \geq C_3 \int_\Omega u^{8p+1} \mathrm{d}x = C_3 \varphi(t), \end{equation} $ (3.7)

这里

$ C_3 = \frac{2 l_0 \lambda_1 - 3k(16p+1) - dk(16p+1)\theta}{2 l_0 + dk \theta^{-1} (16p+1)} > 0. $

利用(3.4)式与(3.7)式有

$ \begin{eqnarray} J_2 & = & \frac{1}{2} \int_\Omega u^{9p} \mathrm{d}x - \frac{32p}{(8p+1)^2} \int_\Omega |\nabla u^{\frac{8p+1}{2}}|^2 \mathrm{d}x \leq \frac{1}{2}|\Omega|^{\frac{1-p}{8p+1}}\varphi(t)^{\frac{9p}{8p+1}} - \frac{32p}{(8p+1)^2}C_3 \varphi(t) \\ & = & \varphi(t)^{\frac{9p}{8p+1}} ( \frac{1}{2}|\Omega|^{\frac{1-p}{8p+1}} - \frac{32p}{(8p+1)^2}C_3 \varphi(t)^{\frac{1-p}{8p+1}} ). \end{eqnarray} $ (3.8)

最后, 结合(3.1), (3.6)与(3.8)式就可以得到下面的一阶微分不等式

$ \begin{eqnarray} \varphi^\prime(t) & \leq& (8p+1)\varphi(t)^{\frac{9p}{8p+1}} ( \frac{1}{2}|\Omega|^{\frac{1-p}{8p+1}} - (\frac{2}{8p+q} )^q (\frac{C_1}{C_0} )^{\frac{q-2}{2}}C_2 |\Omega|^{\frac{1-q}{8p+1}}\varphi(t)^{\frac{q-p}{8P+1}} ) \\ &&+ (8p+1)\varphi(t)^{\frac{9p}{8p+1}} ( \frac{1}{2}|\Omega|^{\frac{1-p}{8p+1}} - \frac{32p}{(8p+1)^2}C_3 \varphi(t)^{\frac{1-p}{8p+1}} ). \end{eqnarray} $ (3.9)

显然, 如果问题(2.1)的解$ u $在测度$ \varphi $的意义下发生爆破, 那么从(3.9)式导出$ \varphi^\prime(t) \leq 0 $, 这是一个矛盾.这样就得到了下面的定理.

定理2   假设$ u $是问题(2.1)的一个非负古典解, 如果$ p < 1 $和第一特征值$ \lambda_1 $满足$ 2 l_0 \lambda_1 - 3k(16p+q) > 0 $, 那么$ u $不能在测度$ \varphi $的意义下发生爆破.

  为了给出定理2的证明, 可以使用另外的辅助函数来代替(2.2)式定义的函数$ \varphi(t) $.例如, 类似文献[6], 可以定义函数$ \Phi(t) = \int_\Omega u^2 \mathrm{d}x $来完成定理的证明.

参考文献
[1] Quittner R, Souplet P. Superlinear parabolic problems: blow-up, global existence and steady states[M]. Basel: Birkhauser, 2007.
[2] Levine H A. The role of critical exponents in blow-up theorems[J]. SIAM Reviews, 1990, 32: 262–288. DOI:10.1137/1032046
[3] Payne L E, Schaefer P W. Lower bounds for blow-up time in parabolic problems under Neumann conditions[J]. Appl. Anal., 2006, 85: 1301–1311.
[4] Li Haixia, Gao Wenjie, Han Yuzhu. Lower bounds for the blowup time of solutions to a nonlinear parabolic problem[J]. Electron J. Differential Equations, 2014, 20: 1–6. DOI:10.1080/10236198.2013.800868
[5] Payne L E, SchaeferP W. Lower bounds for blow-up time in parabolic problems under Dirichlet conditions[J]. J. Math. Anal. Appl., 2007, 328: 1196–1205. DOI:10.1016/j.jmaa.2006.06.015
[6] Payne L E, Philippin G A, Schaefer P W. Bounds for blow-up time in nonlinear parabolic problems[J]. J. Math. Anal. Appl., 2008, 338: 438–447. DOI:10.1016/j.jmaa.2007.05.022
[7] Payne L E, Philippin G A, Schaefer P W. Blow-up phenomena for some nonlinear parabolic problems[J]. Nonlinear Anal., 2008, 69: 3495–3502. DOI:10.1016/j.na.2007.09.035
[8] Liu Dengming, Mu Chunlai, Xin Qiao. Lower bounds estimate for the blow-up time of a nonlinear nonlocal porous medium equation[J]. Acta Math. Sci., 2013, 32B(3): 1206–1212.
[9] Song Xianfa, Lv Xiaoshuang. Bounds of blowup time and blowup rate estimates for a type of parabolic equations with weighted source[J]. Appl. Math. & Computation, 2014, 236: 78–92.
[10] Baghaei K, Hesaaraki M B. Lower bounds for the blow-up time of nonlinear parabolic problems with Robin boundary conditions[J]. Electron J. Differential Equations, 2014, 113: 1–5.
[11] Baghaei K, Ghaemi M B, Hesaaraki M. Lower bounds for the blow-up time in a semilinear parabolic problem involving a variable source[J]. Appl. Math. Letters, 2014, 27: 49–52. DOI:10.1016/j.aml.2013.07.012