数学物理学报  2017, Vol. 37 Issue (5): 834-845   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
张亮
杨国朋
一类非线性Keller-Segel方程的局部零能控性
张亮, 杨国朋     
武汉理工大学数学系 武汉 430070
摘要:该文研究一类由抛物方程和椭圆方程耦合的非线性Keller-Segel方程的局部零能控性.该方程不仅具有非线性的drift-diffuion项,而且具有非线性的人口增长项.作者利用抛物-椭圆结构的非局部特性将方程组化为单个非线性抛物型方程并利用Kakutani不动点定理证明了局部零能控性的存在性.
关键词局部零能控性    Keller-Segel方程    能观性估计    
Null Controllability of a Nonlinear Keller-Segel Equation
Zhang Liang, Yang Guopeng     
Department of Mathematics, Wuhan University of Technology, Wuhan 430070
Abstract: In this paper, we study the local controllability for a nonlinear Keller-Segel equation coupled by an elliptic partial differential equation and a parabolic one, in which nonlinearity lies both on its drift-diffusion term and population growth. We prove the local null controllability by Kakutani's fixed point theorem. The method is established on the nonlocal structure of the elliptic-parabolic equation so that it can be treated as a single nonlinear parabolic equation.
Key words: Null controllability     Keller-Segel equation     Observability estimate    
1 引言

$\Omega \subset \mathbb{R}^{N}(N\geq1)$为带光滑边界的有界区域, $\omega$$\Omega $的开集.记$Q_{T}=\Omega \times(0,T),$ $\Sigma_{T}=\partial \Omega \times(0,T);$并以$|\cdot |_{p}$$\left\Vert \cdot\right\Vert _{p}$记函数空间$L^{p}(\Omega )$$L^{p}(Q_{T})$ ($1\leq p\leq\infty$)的范数; $H^{m}(\Omega )$ ($m$为正整数)为Sobolev函数空间.令

$W(0,T)=\{u\in L^{2}(0,T;H^{1}(\Omega ));\partial _{t}u\in L^{2}(0,T;H^{1} (\Omega )^{\ast})\},$

其中$H^{1}(\Omega )^{\ast}$$H^{1}(\Omega )$的对偶空间, 并记$\left\langle \cdot,\cdot\right\rangle $$H^{1}(\Omega )$$H^{1}(\Omega )^{\ast}$之间的对偶关系. $W(0,T)$的范数定义为

$ \left\Vert u\right\Vert _{W(0,T)}=\left( \left\Vert u\right\Vert _{L^{2}(0,T;H^{1}(\Omega))}^{2}+\left\Vert \partial_{t}u\right\Vert _{L^{2}(0,T;H^{1}(\Omega)^{\ast})}^{2}\right) ^{1/2}. $

考虑如下受控的Keller-Segel方程

$\left\{\begin{array}{ll}\partial_{t}u=\nabla\cdot\left( \nabla u-u\nabla\Phi(ⅳ)\right) +h(u)+{\bf 1}_{\omega}f,~~& \mbox{在$Q_{T}$中},\\ 0=\Delta v-\gamma v+\delta u,& \mbox{在$Q_{T}$中},\\ {\partial }_{\nu}u=0,\partial _{\nu}v=0,& \mbox{在$\Sigma_{T}$上},\\ u(x,0)=u_{0}(x),& x\in\Omega . \end{array}\right. $ (1.1)

这里, $u=u(x,t),v=v(x,t),\partial _{t}=\partial /\partial t$; $\partial _{\nu}=\partial /\partial \nu$表示边界$\partial \Omega $的外法向量; $\gamma,\delta $均为正常数; ${\bf 1}_{\omega}$表示集合$\omega$的特征函数; $f=f(x,t)$表示控制函数; $u_{0}=u_{0}(x)$表示初始值.函数$\Phi:\mathbb{R} \rightarrow\mathbb{R} $$h:\mathbb{R} \rightarrow\mathbb{R} $分别满足

(H$_{1}$) $\Phi\in C^{1}(\mathbb{R} ),\Phi(0)=0;$

(H$_{2}$) $h$局部Lipschitz连续, $h(0)=0.$

函数$(u,v)\in W(0,T)\times L^{2}(0,T;H^{1}(\Omega ))$称为方程(1.1)的弱解, 如果$\forall\varphi ,\phi\in L^{2}(0,T;H^{1}(\Omega )),$下列等式成立

$\begin{eqnarray*} &&\int_{0}^{T}\left\langle \partial_{t}u,\varphi\right\rangle {\rm d}t+ {\displaystyle\iint_{Q_{T}}} \left[ (\nabla u-u\nabla\Phi(v))\cdot\nabla\varphi-h(u)\varphi-{{\bf 1} }_{\omega}f\varphi\right] {\rm d}x{\rm d}t =0,\\ &&{\displaystyle\iint_{Q_{T}}} \left[ \nabla v\cdot\nabla\phi+(\gamma v-\delta u)\phi\right] {\rm d}x{\rm d}t =0. \end{eqnarray*}$

函数$u(x,t)$$v(x,t)$分别表示细菌和药物在时空坐标下的浓度值.当$f=0$时, 模型方程(1.1)可用以描述粘液菌形态形成的药物趋化过程, 该模型由Bonami[1]等给出. (1.1)式的第二个方程不含关于时间的导数项, 表示药物瞬时溶解.这是一类典型的非线性Keller-Segel方程.自1970年Keller和Segel[2]给出Keller-Segel模型以来, Keller-Segel方程广泛应用于生物医学和环境科学等相关领域中, 成为这些领域的核心模型之一.在偏微分方程领域中, Keller-Segel方程因其特殊的数学结构带来挑战而成为经典的偏微分方程模型.近40年内, Keller-Segel方程无论在理论还是应用领域一直是研究的热点问题之一, 目前已经积累了丰富的成果, 比如Keller-Segel方程解的存在性、爆破性、坍塌、聚集、渐近分析等[3-5].

但是关于Keller-Segel方程在控制领域的研究, 目前相关的成果非常少.在文献[6]中, 作者探讨了Keller-Segel方程的最优控制问题.在文献[7]中, 作者证明了当$\Phi(ⅳ)=v,h(u)=0$时系统的局部零能控性.在文献[8]中, 作者给出如下描述药物浓度不瞬时溶解情况下Keller-Segel方程的局部精确能控性

$\left\{\begin{array}{ll}\partial_{t}u=\nabla\cdot\left( \nabla u-u\nabla v\right) +{{\bf 1} }_{\omega}f,~&\mbox{在$Q_{T}$中},\\\partial_{t}v=\Delta v-\gamma v+\delta u,& \mbox{在$Q_{T}$中} . \end{array}\right. $

更多的关于抛物型方程及其方程组的能控性参见文献[9].

本文考虑方程(1.1)的局部零能控性问题.记

${\cal R}(u_{0},T)=\{u(x,T);(u,v)\ \mbox{是方程 (1.1) 相应于$f\in L^{\infty}(\omega\times(0,T))$的解\}.} $

如果存在常数$M>0,$使得对于任意的$u_{0}\in\{u\in L^{\infty}(\Omega );|u_{0}|_{\infty}\leq M\},$都有$0\in {\cal R}(u_{0},T),$则称系统(1.1)在时刻$T$局部零能控.本文考虑系统(1.1)的能控性, 原因如下:其一, 相比其他的带diffusion算子的抛物型方程或方程组, 由于drift-diffusion项$\nabla(u\nabla\Phi(ⅳ))$的存在, 建立能控性所需的一些重要的基本结论更为困难, 比如方程(1.1)的正则性以及线性方程的能观性估计等.其二, 相比于文献[7]的工作, 非线性项$\Phi (ⅳ)$$h(u)$的存在使得方程(1.1)的难度更大, 但是结果更为丰富.本文的结果是文献[7]中工作的进一步深化.其三, 本文考察的是方程组在一个控制下的能控性问题, 类似的问题目前还没有完全解决[10].本文中提出的研究方案不同于现有文献主流的研究方法.具体的, 本文根据该方程的数学结构将之视为单一的非局部抛物型偏微分方程, 从而该方程组的能控性可以视为单一的非线性抛物型偏微分方程的能控性.这种方法可以推广到更广的可视为非局部抛物型偏微分方程的能控性问题的研究, 比如半导体方程的能控性.其四, 方程(1.1)为耦合的抛物方程和椭圆方程组合, 目前相关的研究工作尚没有完全展开, 其中还有大量的控制问题值得探讨, 比如更为一般的抛物-椭圆型方程组的能控性, 能控性的渐近行为等问题.

本文采用近年来发展的处理非线性发展方程能控性问题的经典方法[11-13], 即将非线性方程线性化, 运用线性发展方程的能控性结果结合不动点定理得到非线性发展方程的能控性.本文首先得到如下一般的线性热方程的零能控性

$\left\{\begin{array}{ll}\partial_{t}u-\Delta u+\nabla\cdot\left( B_{0}u\right) +B_{1}\cdot\nabla u+au={\bf 1}_{\omega}f,~~&\mbox{在$ Q_{T}$中,}\\ {\partial }_{\nu}u=0,& \mbox{在$\Sigma_{T}$上,}\\ u(x,0)=u_{0}(x),& x\in\Omega ,\end{array}\right. $ (1.2)

其中$a\in L^{\infty}(Q_{T})$, $B_{i}\in L^{\infty} (Q_{T})^{N}$, 且在边界$\Sigma_{T}$上满足$B_{i}\cdot\nu=0$, $i=0,1$.

定理1.1 对任意的$T>0$和初始值$u_{0} \in L^{2}(\Omega )$, 都存在控制$f\in L^{\infty }(\omega\times(0,T))$, 使得方程(1.2)对应的唯一解$u\in W(0,T)$满足对几乎处处的$x\in\Omega $$u(x,T)=0$.此外, 控制函数$f$满足

$\begin{equation} \left\Vert f\right\Vert _{L^{\infty}(\omega\times(0,T))}\leq {\rm e}^{C\varrho (T)}\left\vert u_{0}\right\vert _{2}, \label{fcontrol} \end{equation}$ (1.3)

其中常数$C=C(\Omega )$,

$\varrho(T)=(1+\left\Vert B_{0}\right\Vert _{\infty}^{2}+\left\Vert B_{1}\right\Vert _{\infty}^{2}+\Vert a\Vert_{\infty})(1+T)+\left\Vert a\right\Vert _{\infty}^{2/3}+\frac{1}{T}. $ (1.4)

定理1.2 设(H$_{1}$)和(H$_{2} $)成立, $T>0.$则存在常数$M(T)$使对任意的初始值$u_{0}$, 只要它满足

$|u_{0}|_{\infty}\leq M(T),$

则系统(1.1)在时刻$T$处是零能控的, 即存在控制函数$f\in L^{\infty }(\omega\times(0,T))$使方程(1.1)的解

$(u,v)\in(C([0,T];L^{2}(\Omega ))\times L^{2}(0,T;H^{1}(\Omega )))\cap L^{\infty }(Q_{T})^{2} $

满足对几乎处处的$x\in\Omega $$u(x,T)=0.$

注1.1 当$f=0$时, 无控制Keller-Segel方程(1.1)的解的性态比较复杂.当$\Phi(u)=u,h(u)=0$时, Keller-Segel方程的解和维数有关.如果维数大于3, 则方程的解在有限时间内爆破.如果维数为2, 当初始值小于一个确定的阈值时, 方程有整体解; 当初始值大于这个阈值时, 解在有限时间内爆破[4].

注1.2 方程(1.1)局部零能控时, 非线性项$h(u)$仅仅要求局部Lipschitz连续, 于是$h(u)$可以超线性增长, 譬如$h(u)=u(1-u)(u-a)$, 其中$0<a<1$[1].注意, 如下受控的非线性抛物型偏微分方程

$\partial _{t}u-\Delta u+h(u)={\bf 1}_{\omega}f,~~\mbox{ 在 $\Omega $ 中.} $ (1.5)

如果非线性项$h(u)$超线性增长, 比如当$p>2$时, 非线性函数$h(s)$满足

$\frac{|h(s)|}{|s|^{p}\log(1+|s|)}\rightarrow0,~~\mbox{ 当$s\rightarrow \infty$ 时,} $

则方程(1.5)不存在全局零能控性[11].因此, 无论是从偏微分方程理论的角度还是控制论的角度, 本文考虑方程(1.1)的局部零能控性是合理的.

注1.3 如果方程(1.1)在时刻$T$零能控, 则方程(1.1)的解$v=v(x,t)$满足

$\limsup\limits_{t\rightarrow T^{-}}\Vert v(\cdot,t)\Vert_{2}=0. $
2 线性方程的正则性与能控性

设椭圆型偏微分方程

$\left\{\begin{array}{ll} -\Delta v+\gamma v=\eta,& \mbox{在$\Omega $中},\\ {\partial }_{\nu}v=0,& \mbox{在$\partial \Omega $上}. \end{array}\right. $ (2.1)

定理2.1 [14] 设$\eta\in L^{p}(\Omega )$, $p>1$, $\gamma >0$, 则方程(2.1)有唯一解$v\in W^{2,p}(\Omega )$, 且有仅依赖于$\Omega $$p$的常数$C$, 使

$\left\Vert v\right\Vert _{W^{2,p}(\Omega )}\leq C\left\vert \eta\right\vert _{p}. $

设抛物型偏微分方程

$\left\{\begin{array}{ll}\partial_{t}u-\Delta u+\nabla\cdot\left( B_{0}u\right) +B_{1}\cdot\nabla u+au=F,~&\mbox{在$Q_{T}$上},\\ {\partial }_{\nu}u=0,& \mbox{在$\Sigma_{T}$上},\\ u(x,0)=u_{0}(x),& x\in\Omega . \end{array}\right. $ (2.2)

为证明方程(2.2)的正则性, 我们需要如下的引理.

引理2.1  [15]设正常数$c,\varepsilon $$b\geq1$, 使非负序列$\{Y_{s}\}_{s=0}^{\infty} $满足

$Y_{s+1}\leq cb^{s}Y_{s}^{1+\varepsilon },\mbox{ }s=0,1,2,\cdots,$

如果$Y_{0}\leq c^{-\frac{1}{\varepsilon }}b^{-\frac{1} {\varepsilon ^{2}}},$则当$s\rightarrow\infty$时, 有$Y_{s}\rightarrow0$.

定理2.2 设$B_{0},B_{1}\in L^{\infty}(Q_{T})^{N}$且在$\Sigma_{T} $上有$B_{0}\cdot\nu=0,$ $a,F\in L^{\infty}(Q_{T})$, $u_{0}\in L^{\infty}(\Omega )$, 则方程(2.2)有唯一解$u\in L^{\infty}(Q_{T})\cap W(0,T)$,

$\left\Vert u\right\Vert _{W(0,T)}+\Vert u\Vert_{\infty}\leq {\rm e}^{C\kappa (T)}\left( |u_{0}|_{\infty}+\Vert F\Vert_{\infty}\right) . $ (2.3)

其中常数$C=C(\Omega )$,

$\kappa(T)=(1+\left\Vert B_{0}\right\Vert _{\infty}^{2}+\left\Vert B_{1}\right\Vert _{\infty}^{2}+\Vert a\Vert_{\infty})(1+T). $

 弱解的存在性可以参见文献[15].方程的解$u$$W(0,T)$中的估计可利用经典的能量估计得到.因此这里只证明解的有界性.设数$l$满足$l\geq L_{0}=\left\Vert F\right\Vert _{\infty}+\left\vert u_{0}\right\vert _{\infty}$, $t\in\lbrack0,T]$.记$(u-l)_{+}=\max\{u-l,0\}$$A_{l}(t)=\{x\in\Omega ,u(x,t)>l\}.$$(u-l)_{+}$乘以(2.2)式两端, 并积分得

$\begin{eqnarray*} && \frac{\rm d}{{\rm d}t}\int_{\Omega}\left\vert (u-l)_{+}\right\vert ^{2} {\rm d}x+\int_{\Omega}\left\vert \nabla(u-l)_{+}\right\vert ^{2}{\rm d}x\\ & \leq&(2\left\Vert B_{0}\right\Vert _{\infty}^{2}+2\left\Vert B_{1} \right\Vert _{\infty}^{2}+\Vert a\Vert_{\infty}+1)\left( \int_{\Omega }\left\vert (u-l)_{+}\right\vert ^{2}{\rm d}x+\int_{A_{l}(t)}l^{2}{\rm d}x\right) . \end{eqnarray*}$

根据Gronwall引理, 有

$\Vert(u-l)_{+}\Vert_{V_{2}}^{2}\leq {\rm e}^{C\kappa(T)}\int_{0}^{T}\int_{A_{l} (t)}l^{2}{\rm d}x.$ (2.4)

这里Sobolev空间$V_{2}=L^{\infty}(0,T;L^{2} (\Omega ))\cap L^{2}(0,T;H^{1}(\Omega ))$赋以范数

$\left\Vert u\right\Vert _{V_{2}}=\left( \left\Vert u\right\Vert _{L^{\infty }(0,T;L^{2}(\Omega ))}^{2}+\left\Vert u\right\Vert _{L^{2}(0,T;H^{1}(\Omega ))}^{2}\right) ^{1/2},$

且有嵌入$V_{2}\hookrightarrow L^{\frac{2(N+2)}{N} }(Q_{T})$和不等式(参见文献[16, 第8页])

$\left\Vert (u-l)_{+}\right\Vert _{\frac{2(N+2)}{N}}\leq C(1+T)^{\frac {N}{2(N+2)}}\Vert(u-l)_{+}\Vert_{V_{2}}.$ (2.5)

$\varphi (l)$为可测集$\{(x,t)\in Q_{T};u(x,t)>l\}$的Lebesgue测度, 并设$h>l.$根据$A_{h}(t)\subset A_{l}(t),$以及(2.4)和(2.5)式可知

$\varphi (h)\leq {\rm e}^{C\kappa(T)}\left( \frac{l}{h-l}\right) ^{\frac{2(N+2)}{N} }\varphi (l)^{1+{\frac{2}{N}}}.$ (2.6)

$Y_{s}=\varphi (k_{s}),k_{s}=M(2-\frac{1}{2^{s}})$.在(2.6)式中取$h=k_{s+1}$$l=k_{s}$

$Y_{s+1}\leq {\rm e}^{C\kappa(T)}4^{\frac{2(N+2)}{N}}\left[2^{\frac{2(N+2)}{N} }\right] ^{s}Y_{s}^{1+\frac{2}{N}},s=0,1,2,\cdots. $

根据引理2.1, 要使$\varphi (2M)=\lim\limits _{s\rightarrow\infty}Y_{s}=0$, 只须

$Y_{0}=\varphi (k_{0})=\varphi (M)\leq\left[{\rm e}^{C\kappa(T)}4^{\frac{2(N+2)}{N} }\right] ^{-\frac{N}{2}}\left[2^{\frac{2(N+2)}{N}}\right] ^{-\frac{N^{2} }{4}}.$ (2.7)

为确定常数$M$, 可取$m>1$为整数, 并待定$M=mL_{0}$.在(2.6)式中取$h=mL_{0}$$l=L_{0}$, 并注意到$\varphi (L_{0})\leq$meas$(Q_{T})=T$meas$(\Omega ),$则有

$\varphi (M)\leq {\rm e}^{C\kappa(T)}4^{\frac{2(N+2)}{N}}\left( \frac{1}{m-1}\right) ^{\frac{2(N+2)}{N}}T^{1+\frac{2}{N}}\mbox{meas}(\Omega )^{1+\frac{2}{N} }.$ (2.8)

根据(2.7)和(2.8)式, 可以选取整数$m$使得

${\rm e}^{C\kappa(T)}4^{\frac{2(N+2)}{N}}\left( \frac{1}{m-1}\right) ^{\frac {2(N+2)}{N}}T^{1+\frac{2}{N}}\mbox{meas}(\Omega )^{1+\frac{2}{N}}\leq\left[{\rm e}^{C\kappa(T)}4^{\frac{2(N+2)}{N}}\right] ^{-\frac{N}{2}}\left[2^{\frac{2(N+2)}{N}}\right] ^{-\frac{N^{2}}{4}},$

$m\geq {\rm e}^{C\kappa(T)\frac{(N+2)^{2}}{N}}2^{2+N+\frac{N^{2}}{4}}T^{\frac{1}{2} }\mbox{meas}(\Omega )^{\frac{1}{2}}+1. $

即有$\varphi (2M)=\varphi (2mL_{0})=0$, 这说明

$u\leq2mL_{0}\leq {\rm e}^{C\kappa(T)}(\left\Vert F\right\Vert _{\infty} +|u_{0}|_{\infty}). $

$-u$替代$u$并用类似的分析可最终得到不等式(2.3).

注2.1 定理2.2的证明可以参见文献[15].在这里我们运用文献[15]的方法得到弱解的关于时间$T$的显式估计(2.3), 这样的估计对于证明非线性抛物型方程的能控性等非常有用[11, 17].

下面证明线性抛物型方程的零能控性, 为此考虑方程(1.2)的共轭方程

$\left\{\begin{array}{ll} -\partial _{t}p-\Delta p-B_{0}\cdot\nabla p-\nabla\cdot(B_{1}p)+ap=0,~~& \mbox{在$Q_{T}$中},\\ {\partial }_{\nu}p=0,& \mbox{在$\Sigma_{T}$上},\\ p(x,T)=p_{T}(x),& x\in\Omega . \end{array}\right. $ (2.9)

有如下的Carleman不等式.

引理2.2 [8] 设$f_{i}\in L^{2}(Q_{T}),i=0,1,\cdots,N,$函数$z$是下列方程的弱解

$\left\{\begin{array}{ll} -\partial _{t}z-\Delta z=f_{0}+\sum\limits_{i=1}^{N}\frac{\partial f_{i}}{\partial x_{i}},~&\mbox{在$Q_{T}$中},\\[2mm] {\partial }_{\nu}z=0,& \mbox{在$\Sigma_{T}$上},\\ z(x,T)=z_{T}(x),& x\in\Omega . \end{array}\right. $

则存在正常数$C=C(\Omega ,\omega),\lambda_{0}=\lambda_{0}(\Omega ,\omega)>1,$使得任意的$\lambda\geq\lambda_{0},s\geq \gamma(\lambda)(T+T^{2}),$

$\begin{eqnarray*} &&{\displaystyle\iint_{Q_{T}}} [\lambda^{2}s\varphi|\nabla z|^{2}+\lambda^{4}(s\varphi)^{3}z^{2}]{\rm e}^{2s\alpha }{\rm d}x{\rm d}t \\ & \leq& C {\displaystyle\iint_{Q_{T}}} {\rm e}^{2s\alpha}f_{0}^{2}{\rm d}x{\rm d}t+C\sum_{i=1}^{N} {\displaystyle\iint_{Q_{T}}} \lambda^{2}(s\varphi)^{2}{\rm e}^{2s\alpha}f_{i}^{2}{\rm d}x{\rm d}t\\ && +C {\displaystyle\iint_{\omega\times(0,T)}} \lambda^{4}(s\varphi)^{3}{\rm e}^{2s\alpha}z^{2}{\rm d}x{\rm d}t, \end{eqnarray*}$

这里

$\varphi =\frac{{\rm e}^{\lambda\beta}}{t(T-t)},\alpha=\frac{{\rm e}^{\lambda\beta }-{\rm e}^{2\lambda\left\Vert \beta\right\Vert _{C(\bar{\Omega })}}}{t(T-t)},\gamma(\lambda)={\rm e}^{2\lambda\left\Vert \beta\right\Vert _{C\left( \bar {\Omega }\right) }},$

其中函数$\beta\in C^{2}(\bar{\Omega })$满足[9]

$\beta(x)>0,\forall x\in\Omega ;\mbox{ }\beta(x)=0,\forall x\in\partial \Omega ;\mbox{ }|\nabla\beta(x)|>0,\forall x\in\overline{\Omega \setminus\omega }. $

有如下的能观性估计.

命题2.1 设$\tau_{0}\in(1,2).$则存在正常数$\lambda$$s$使方程(2.9)的弱解满足

$\begin{equation} |p(\cdot,0)|_{2}^{2}\leq {\rm e}^{C\varrho(T)} {\displaystyle\iint_{{\omega}\times(0,T)}} {\rm e}^{\tau_{0}s\alpha}p^{2}{\rm d}x{\rm d}t, \label{car0} \end{equation}$ (2.10)

其中常数$C=C(\Omega )$, $\varrho(T)$形如(1.4)式.

 一方面, 根据引理2.2, 存在正常数$\lambda_{1}=\lambda_{1}(\Omega ,\omega)$使对任意的$\lambda\geq\lambda_{1},$ $s\geq\gamma(\lambda)(T+T^{2}),$方程(2.9)的弱解满足

$\begin{eqnarray*} && {\displaystyle\iint_{Q_{T}}} [\lambda^{2}s\varphi|\nabla p|^{2}+\lambda^{4}(s\varphi)^{3}p^{2}]{\rm e}^{2s\alpha }{\rm d}x{\rm d}t\\ & \leq& C\left\Vert B_{0}\right\Vert _{\infty}^{2} {\displaystyle\iint_{Q_{T}}} {\rm e}^{2s\alpha}|\nabla p|^{2}{\rm d}x{\rm d}t+C\left\Vert a\right\Vert _{\infty}^{2} {\displaystyle\iint_{Q_{T}}} {\rm e}^{2s\alpha}|p|^{2}{\rm d}x{\rm d}t\\ && +C\left\Vert B_{1}\right\Vert _{\infty}^{2} {\displaystyle\iint_{Q_{T}}} \lambda^{2}(s\varphi)^{2}{\rm e}^{2s\alpha}|p|^{2}{\rm d}x{\rm d}t+C {\displaystyle\iint_{\omega\times(0,T)}} \lambda^{4}(s\varphi)^{3}{\rm e}^{2s\alpha}p^{2}{\rm d}x{\rm d}t. \end{eqnarray*}$

$\lambda_{2}=C(1+\left\Vert B_{0}\right\Vert _{\infty}^{2}+\left\Vert B_{1}\right\Vert _{\infty}^{2}+\left\Vert a\right\Vert _{\infty}^{2/3} )$使对任意的$\lambda>\lambda_{2}>\lambda _{1}>1,s>\gamma(\lambda)(T+T^{2})$

$s\varphi >2C\max\{\left\Vert B_{0}\right\Vert _{\infty}^{2},\left\Vert B_{1}\right\Vert _{\infty}^{2},\left\Vert a\right\Vert _{\infty}^{2/3}\},$

$\begin{eqnarray} {\displaystyle\iint_{Q_{T}}} [s\varphi|\nabla p|^{2}+(s\varphi)^{3}p^{2}]{\rm e}^{2s\alpha}{\rm d}x{\rm d}t & \leq& C {\displaystyle\iint_{\omega\times(0,T)}} \lambda^{4}(s\varphi)^{3}{\rm e}^{2s\alpha}p^{2}{\rm d}x{\rm d}t\nonumber\\ & \leq& C {\displaystyle\iint_{\omega\times(0,T)}} {\rm e}^{\tau_{0}s\alpha}p^{2}{\rm d}x{\rm d}t. \end{eqnarray}$ (2.11)

另一方面, 根据方程(2.9), 有如下不等式

$\frac{\rm d}{{\rm d}t}\left( {\rm e}^{2(\left\Vert B_{0}\right\Vert _{\infty}^{2}+\left\Vert B_{1}\right\Vert _{\infty}^{2}+\left\Vert a\right\Vert _{\infty})t}|p( \cdot,t)|_2^2\right) \geq0,$

由此, 可得

$|p(\cdot,0)|_{2}^{2}\leq {\rm e}^{2(\left\Vert B_{0}\right\Vert _{\infty} ^{2}+\left\Vert B_{1}\right\Vert _{\infty}^{2}+\left\Vert a\right\Vert _{\infty})T}|p(\cdot,t)|_{2}^{2},\forall t\in\lbrack0,T]. $

对上式在$[T/4,3T/4]$上积分, 有

$|p(\cdot,0)|_{2}^{2}\leq\frac{2}{T}{\rm e}^{2(\left\Vert B_{0}\right\Vert _{\infty }^{2}+\left\Vert B_{1}\right\Vert _{\infty}^{2}+\left\Vert a\right\Vert _{\infty})T}\int_{T/4}^{3T/4}\int_{\Omega }p^{2}{\rm d}x{\rm d}t. $ (2.12)

由于在$\Omega \times\lbrack T/4,3T/4]$上, $(s\varphi )^{-3}{\rm e}^{-2s\alpha}\leq {\rm e}^{Cs/T^{2}},$根据(2.11)和(2.12)式, 可以取足够大的$\lambda$$s$使(2.10)式成立.

定理1.1的证明 在命题2.1中, 我们可以取足够大的$\lambda $$s$使能观性不等式(2.10)和下列不等式成立

$\eta(\lambda)={\rm e}^{-\lambda\left\Vert \beta\right\Vert _{C(\bar{\Omega })}} <\tau_{0}-1. $ (2.13)

$\varepsilon >0,$考虑最优化控制问题

$\inf\limits_{(u,f)\in{\cal U}}\frac{1}{2} {\displaystyle\iint_{{\omega}\times(0,T)}} f^{2}{\rm e}^{-\tau_{0}s\alpha}{\rm d}x{\rm d}t+\frac{1}{2\varepsilon } {\displaystyle\int_{\Omega }} |u(x,T)|^{2}{\rm d}x. $

这里${\cal U}$是所有的满足方程(1.2)的函数对$(u,f)\in W(0,T)\times L^{2}(\omega \times(0,T)).$易知最优对$(u_{\varepsilon },f_{\varepsilon } )$存在[18], 且满足如下最优系统

$\left\{\begin{array}{ll}\partial_{t}u_{\varepsilon }-\Delta u_{\varepsilon }+\nabla\cdot\left( B_{0}u_{\varepsilon }\right) +B_{1}\cdot\nabla u_{\varepsilon }+au_{\varepsilon }={\bf 1}_{\omega}f_{\varepsilon },~&\mbox{在$Q_{T}$中},\\ {\partial }_{\nu}u_{\varepsilon }=0,& \mbox{在$\Sigma_{T}$上},\\ u_{\varepsilon }(x,0)=u_{0}(x),& x\in\Omega ; \end{array}\right. $ (2.14)
$\left\{\begin{array}{ll} -\partial _{t}p_{\varepsilon }-\Delta p_{\varepsilon }-B_{0}\cdot\nabla p_{\varepsilon }-\nabla\cdot(B_{1}p_{\varepsilon })+ap_{\varepsilon }=0,~~& \mbox{在$Q_{T}$中},\\ {\partial }_{\nu}p_{\varepsilon }=0,& \mbox{在$\Sigma_{T}$上},\\[2mm] p_{\varepsilon }(x,T)=-\frac{1}{\varepsilon }u_{\varepsilon }(x,T),& x\in\Omega ; \end{array}\right. $ (2.15)
$f_{\varepsilon }={\bf 1}_{\omega}p_{\varepsilon }{\rm e}^{\tau_{0}s\alpha}. $ (2.16)

根据(2.14)-(2.16)式, 以及能观性估计即命题2.1, 有

$\iint_{\omega\times(0,T)} p_{\varepsilon }^{2}{\rm e}^{\tau_{0}s\alpha}{\rm d}x{\rm d}t+\frac{1}{\varepsilon } {\displaystyle\int_{\Omega }} |u_\varepsilon (x,T)|^{2}{\rm d}x\leq {\rm e}^{C\varrho(T)}|u_{0}|_{2}^{2},$ (2.17)

由此可得控制函数$f_{\varepsilon }$$L^{2}$ -估计

$\left\Vert {{\bf 1}}_{\omega}f_{\varepsilon }\right\Vert _{2}\leq {\rm e}^{C\varrho(T)}|u_{0}|_{2}. $

下面我们证明控制函数$f_{\varepsilon }$可以在$L^{\infty}(\omega\times(0,T))$中选取.为此, 设$\alpha_{0}=\min\limits _{\bar{\Omega }}\alpha,$

$\alpha_{0}\leq\alpha\leq\alpha_{0}/[1+\eta(\lambda)]. $ (2.18)

取序列$\{l_{i}\}_{i=0}^{M}$$\{p_{i}\}_{0=1}^{M} $分别满足$0<l_{i}\leq l_{i+1}<l_{M}=l,i=0,1,\cdots,M-1$$2=p_{0} <p_{i}<p_{i+1}<p_{M}=\infty,i=1,\cdots,M-1.$定义

$\begin{eqnarray*}&&q_{i}(x,t) ={\rm e}^{(s+l_{i})\alpha_{0}}p_{\varepsilon }(x,T-t),\\ &&F_{i}(x,t) =\partial_{t}({\rm e}^{(s+l_{i})\alpha_{0}} p_{\varepsilon }(x,T-t)),\\ &&\tilde{B}_{i}(x,t) =B(x,T-t),i=0,1,\\ &&\tilde{a}(x,t) =a(x,T-t).\end{eqnarray*}$

于是对每个$i=0,1,\cdots,M,$函数$q_{i} $满足方程

$\left\{\begin{array}{ll}\partial_{t}q_{i}-\Delta q_{i}-\tilde{B}_{0}\cdot\nabla q_{i}-\nabla \cdot(\tilde{B}_{1}q_{i})+\tilde{a}q_{i}=F_{i},~~& \mbox{在 $Q_{T}$中},\\ {\partial }_{\nu}q_{i}=0,& \mbox{在$Q_{T}$上},\\ q_{i}(x,0)=0,& x\in\Omega . \end{array}\right. $ (2.19)

设带零Neumann边界条件的Laplace算子$-\Delta _{N}$$L^{p}(\Omega )$ ($p>1$)上生成的半群为$\{S(t)\}_{t\geq0},$则根据半群理论[19]有如下的$L^{p}$-$L^{q}$估计

$|S(t)u|_{q}\leq Cm(t)^{-\frac{N}{2}(\frac{1}{p}-\frac{1}{q})}|u|_{p},\forall u\in L^{p}(\Omega ),t>0,1<p\leq q\leq\infty,$

且对任意的$\varepsilon >0,$存在正常数$C_{\varepsilon },$

$|S(t)\nabla u|_{p}\leq C_{\varepsilon }m(t)^{-\frac{1}{2}-\varepsilon-\frac {N}{2}(\frac{1}{p}-\frac{1}{q})}|u|_{p},\forall u\in L^{p}(\Omega ),t>0,1<p\leq q<\infty,$

其中$m(t)=\min\{t,1\}.$注意方程(2.19)的解可表示为

$q_{i}(\cdot,t)=\int_{0}^{t}S(t-s)\left( \tilde{B}_{0}\cdot\nabla q_{i} +\nabla\cdot(\tilde{B}_{1}q_{i})-\tilde{a}q_{i}+F_{i}\right) (\cdot,s){\rm d}s. $

从而应用半群的$L^{p_{i-1}} $-$L^{p_{i}}$估计, 以及不等式$\left\Vert F_{i}\right\Vert _{p_{i-1}}\leq CT\left\Vert q_{i-1}\right\Vert _{p_{i-1}},$得迭代式

$\left\Vert q_{i}\right\Vert _{p_{i}}\leq {\rm e}^{C\varrho(T)}\left\Vert q_{i-1}\right\Vert _{p_{i-1}},$

由此对$i=0,1,\cdots,M-1$迭代得

$\left\Vert q_{M}\right\Vert _{p_{M}}\leq {\rm e}^{C\varrho(T)}\left\Vert q_{0}\right\Vert _{2}. $ (2.20)

因为$p_{M}=\infty,$ $l_M=l$, 注意到$p_\varepsilon $满足不等式(2.11), 所以根据$q_{i}$的定义, (2.17)和(2.20)式得

$\left\Vert p_{\varepsilon }{\rm e}^{(s+l)\alpha_{0}}\right\Vert _{\infty}\leq {\rm e}^{C\varrho(T)}|u_{0}|_{2}. $

由此式和(2.18)式有

$\left\Vert {\rm e}^{\left[-s(\tau_{0}-1-\eta(\lambda))+l(1+\eta(\lambda))\right] \alpha}f_{\varepsilon }\right\Vert _{L^{\infty}(\omega\times(0,T))}\leq {\rm e}^{C\varrho(T)}|u_{0}|_{2}. $

$l$足够小使$-s(\tau_{0}-1-\eta(\lambda ))+l(1+\eta(\lambda))<0,$从而根据上面最后一个不等式得

$\left\Vert f_{\varepsilon }\right\Vert _{L^{\infty}(\omega\times(0,T))}\leq {\rm e}^{C\varrho(T)}|u_{0}|_{2}. $ (2.21)

这说明序列$\{f_{\varepsilon }\}$$L^{\infty}(\omega\times(0,T))$中的有界集, 于是存在子列$\{f_{\varepsilon _{k}} \}_{k=1}^{\infty}$满足当$\varepsilon _{k}\rightarrow 0$时, $f_{\varepsilon _{k}}$$L^{\infty}(\omega\times (0,T))$中弱*收敛于$f\in L^{\infty}({\omega \times(0,T)}).$$u_{\varepsilon }$为方程(2.14)相应于$f_{\varepsilon }$的弱解, 根据定理2.1和(2.21)式知$\{u_{\varepsilon }\}$在函数空间$W(0,T)$中有界, 从而可以抽取子列$\{u_{\varepsilon _{k}}\}_{k=1}^{\infty}$使当$\varepsilon _{k}\rightarrow0$时, $u_{\varepsilon _{k}}$$W(0,T)$中弱收敛于$u\in W(0,T).$最后对方程(2.14)取$\varepsilon _{k}\rightarrow0$时的极限, 可知$u$是相应于$f$$u_{0}$的方程(1.2)的弱解, 而且根据(2.17)式知对几乎处处的$x\in \Omega $$u(x,T)=0.$

3 非线性方程的局部零能控性

定理1.2的证明 分两步证明.

步骤1 设$h$$C^{1} $函数.令

$H(s)=\left\{ \begin{array}{ll} \frac{h(s)}{s},~~& s\neq0,\\[2mm] h^{\prime}(0),& s=0. \end{array} \right. $

$X=L^{2}(Q_{T}),$$X$的子集$Z=\{z\in L^{\infty}(Q_{T});\left\Vert z\right\Vert _{\infty}\leq1\}\cap L^{\infty }(0,T;L^{p}(\Omega )),p>\max\{N,2\}.$对于任意的$z\in Z,$考虑如下线性椭圆型偏微分方程

$\left\{\begin{array}{ll} -\Delta v_{z}+\gamma v_{z}=\delta z,~~& \mbox{在$\Omega $上},\\ {\partial }_{\nu}v_{z}=0,& \mbox{在$\partial \Omega $上} \end{array}\right. $ (3.1)

和线性抛物型偏微分方程

$\left\{\begin{array}{ll}\partial_{t}u_{z}-\Delta u_{z}-\nabla\cdot\left( B_{z}u_{z}\right) +a_{z}u_{z}={\bf 1}_{\omega}f_{z},~~& \mbox{在$Q_{T}$上},\\ {\partial }_{\nu}u_{z}=0,& \mbox{在$\Sigma_{T}$上},\\ u_{z}(x,0)=u_{0}(x),& x\in\Omega . \end{array}\right. $ (3.2)

其中

$B_{z}=\nabla \Phi(v_{z}),a_{z}=H(z).$

根据定理2.1, 对于$z\in Z$和几乎处处$t\in\lbrack0,T],$方程(3.1)有唯一的对应于$z(\cdot,t)$的弱解$v_{z}\in L^{\infty}(0,T;W^{2,p}(\Omega )),$且满足

$\left\Vert v_{z}\right\Vert _{L^{\infty}(0,T;W^{2,p}(\Omega ))}\leq c_{0}\left\Vert z\right\Vert _{L^{\infty}(0,T;L^{p}(\Omega ))}\leq c_{0}. $

由此可以定义从$Z$$L^{\infty }(0,T,W^{2,p}(\Omega ))\times L^{2}(Q_{T})$的有界线性算子

${\cal L}(z)=(v_{z},z). $

根据(H$_{1} $)和(H$_{2}$), 非线性映射$B_{z}=\nabla \Phi(v_{z})=\Phi^{\prime}(v_{z})\nabla v_{z}$$a_{z}=H(z)$满足

$\left\Vert B_{z}\right\Vert _{\infty} \leq\left\Vert \Phi\right\Vert _{C^{1}([-c_{0},c_{0}])}\left\Vert \nabla v_{z}\right\Vert _{\infty}\leq c_{0}c_{1}\left\Vert \Phi\right\Vert _{C^{1}([-c_{0},c_{0}])},$ (3.3)
$\left\Vert a_{z}\right\Vert _{\infty} \leq\left\Vert H\right\Vert _{L^{\infty}(-1,1)},$ (3.4)

其中$c_{1}$为嵌入不等式$\left\Vert u\right\Vert _{W^{1,\infty}(\Omega )}\leq c_{1}\left\Vert u\right\Vert _{W^{2,p}(\Omega )} $中的常数.根据定理1.1, 对于每对$(v,z)=(v,z)\in L^{\infty}(0,T;W^{1,\infty }(\Omega ))\times Z,$方程(3.2)在时刻$T$零能控, 亦即存在控制函数$f_{z}\in L^{\infty}(\omega\times(0,T))\,,$使方程(3.2)对应$f_{z}$与初始值$u_{0}$的弱解$u$对几乎处处的$x\in\Omega $$u_{z}(x,T)=0$, 且$f_{z}$满足

$\left\Vert f_{z}\right\Vert _{L^{\infty}(\omega\times(0,T))}\leq C{\rm e}^{m_{0} (T)}|u_{0}|_{2},$ (3.5)

其中常数$C=C(\Omega )$,

$m_{0}(T)=C(\Omega ,\left\Vert H\right\Vert _{L^{\infty}(-1,1)},\left\Vert \Phi\right\Vert _{C^{1}([-c_{0},c_{0}])})(1+T+\frac{1}{T}). $ (3.6)

由此可以定义从$L^{\infty }(0,T;W^{2,p}(\Omega ))\times Z$$2^{L^{2}(Q_{T})}$的多值映射

${\cal M}(v,z)=\left\{ u_{z}\in L^{2}(Q_{T})\left\vert \begin{array} [c]{c} \mbox{存在满足(3.5)式的$f_{z}\in L^{\infty}(\omega\times(0,T))$使方程(3.2)}\\ \mbox{对应的解$u_{z}\in W(0,T)$满足 $u_{z}(x,T)=0\ \mbox{ a.e }\ x\in\Omega $} \end{array} \right. \right\} . $

${\cal A}={\cal M}\circ{\cal L,}$${\cal A} $为从$Z\subset L^{2}(Q_{T})$$2^{L^{2}(Q_{T})} $的多值映射, 下面证明它满足Kakutani不动点定理[18].

首先, $Z$$L^{2}(Q_{T})$中的凸集, 且根据定理2.2, 对于每个$z\in Z$, 集合${\cal A} (z)$$W(0,T)$的有界集, 从而根据Aubin-Lions引理[18], ${\cal A}(z)$$L^{2}(Q_{T})$的紧集.

其次, ${\cal A}$是上半连续的.事实上, 取$Z$中依$L^{2}(Q_{T})$范数强收敛于$z$的序列$\{z_{n}\}_{n=1}^{\infty}$.对每个$n,$$v_{n}={\cal L}(z_{n})$为下列椭圆型偏微分方程的解

$\left\{\begin{array}{ll} -\Delta v_{n}+\gamma v_{n}=\delta z_{n},~&\mbox{在$\Omega $上},\\ {\partial }_{\nu}v_{n}=0,& \mbox{在$\partial \Omega $上}. \end{array}\right. $ (3.7)

$B_{n}=\nabla\Phi(v_{n}),a_{n}=H(z_{n}),$$u_{n}\in {\cal M}(v_{n},z_{n})={\cal M}\circ{\cal L}(z_{n})={\cal A} (z_{n})$为下列抛物型偏微分方程

$\left\{\begin{array}{ll}\partial_{t}u_{n}-\Delta u_{n}-\nabla\cdot\left( B_{n}u_{n}\right) +a_{n}u_{n}={\bf 1}_{\omega}f_{n},~&\mbox{在$Q_{T}$上},\\ {\partial }_{\nu}u_{n}=0,& \mbox{在$\Sigma_{T}$上},\\ u_{n}(x,0)=u_{0}(x),& x\in\Omega \end{array}\right. $ (3.8)

的解且满足$u_{n}(x,T)=0;$控制函数$f_{n}$满足不等式$(3.5).$于是, 序列$\{f_{n}\}_{n=1}^{\infty}$可以取子序列$\{f_{n_{k}}\}_{k=1}^{\infty}$使其在$L^{\infty}(\omega\times(0,T))$中弱*收敛于$f.$根据定理2.2可知$\{u_{n} \}_{n=1}^{\infty}$$W(0,T)$中一致有界, 从而存在子序列$\{u_{n_{k}} \}_{k=1}^{\infty}$使其在$W(0,T)$中弱收敛于$u\in W(0,T);$再次根据Aubin-Lions引理知$\{u_{n_{k} }\}_{k=1}^{\infty}$$L^{2}(Q_{T})$中强收敛于$u.$一方面, 根据Sobolev空间嵌入$W(0,T)\subset C([0,T];L^{2}(\Omega ))$$u\in C([0,T];L^{2}(\Omega )).$另一方面, 根据(3.3)和(3.4)式知$\{B_{n}u_{n} \}_{n=1}^{\infty}$$\{a_{n}u_{n}\}_{n=1}^{\infty}$均在$L^{2}(Q_{T})$中有界, 从而分别存在子列$\{B_{n_{k} }u_{n_{k}}\}_{k=1}^{\infty}$$\{a_{n_{k}}u_{n_{k}}\}_{k=1}^{\infty} $$L^{2}(Q_{T})$中弱收敛; 又因为函数$\Phi$$h$$C^{1}$函数, $\{B_{n_{k}}u_{n_{k}}\}_{k=1}^{\infty}$$\{a_{n_{k} }u_{n_{k}}\}_{k=1}^{\infty}$分别几乎处处收敛于$B_{z} u$$a_{z}u$, 所以当$k\rightarrow\infty$时, $B_{n_{k}}u_{n_{k}}$$a_{n_{k}}u_{n_{k}}$分别在$L^{2}(Q_{T})$中弱收敛于$B_{z}u$$a_{z}u.$最后, 在(3.8)式中取$k\rightarrow\infty $的极限可知函数$u\in W(0,T)$为方程(3.1)对应于$f\in L^{\infty}(\omega\times(0,T))$的弱解, 且满足$u(x,T)=0,$ a.e. $x\in\Omega .$由此证明了如果$Z$中有收敛于$z$的序列$\{z_{n} \}_{n=1}^{\infty},$可以找到$\{u_{n}\}_{n=1}^{\infty} $ (其中$u_{n}\in{\cal A}(z_{n})$)的子列$\{u_{n_{n}}\}_{k=1}^{\infty}$$L^{2}(Q_{T})$中强收敛于$u,$从而证明了多值映射${\cal A} $的上半连续.

最后, ${\cal A}(Z)\subset Z.$根据线性抛物型偏微分方程的能量估计, 对任意的$z\in Z,$ ${\cal A}(z)$中的元素, 亦即方程(3.2)的解满足

$\left\Vert u_{z}\right\Vert _{L^{\infty}(0,T;L^{p}(\Omega ))}\leq {\rm e}^{c((\left\Vert B_{o}\right\Vert _{\infty}^{2}+\left\Vert B_{1}\right\Vert _{\infty}^{2}+1)T+1+\left\Vert a\right\Vert _{\infty})}\left( |u_{0} |_{p}+\left\Vert {\bf 1}_{\omega}f\right\Vert _{p}\right),$

由此式, $(3.5)$式和定理2.2, 有

$\left\Vert u_{z}\right\Vert _{\infty}+\left\Vert u_{z}\right\Vert _{L^{\infty }(0,T;L^{p}(\Omega ))}\leq {\rm e}^{Cm_{0}(T)}|u_{0}|_{\infty},$

其中$m_{0}(T)$形如(3.6)式.取$M(T)={\rm e}^{-Cm_{0}(T)},$则只要$|u_{0}|_{\infty}\leq M(T),$就有$u_{z}\in Z.$根据上面的分析知, ${\cal A}$$Z$中存在不动点$u\in {\cal A}(u)$, 且$u$为方程(1.1)对应于一个控制函数$f\in L^{\infty}(\omega\times(0,T))$的弱解, 它对几乎处处$x\in\Omega $满足$u(x,T)=0.$

步骤2 设(H$_{2}$)成立.引入函数$\rho\in C_{0}^{\infty}(\mathbb{R} ),$它满足$\rho\geq0,$ supp$\rho\subset\lbrack-1,1],$ $\int_{\mathbb{R} }\rho(x){\rm d}x=1.$$\rho _{n}(s)=n\rho(ns),s\in\mathbb{R},$并记$H$的光滑化函数为$H_{n}=\rho_{n}\ast H,n\geq1.$$h_{n}(s)=H_{n}(s)s,$ $s\in\mathbb{R},$$h_{n} $满足步骤1中的条件.对每个$n,$考虑如下非线性方程

$\left\{\begin{array}{ll}\partial_{t}u_{n}=\nabla\cdot\left( \nabla u_{n}-u\nabla\Phi(v_{n})\right) +h_{n}(u_{n})+{\bf 1}_{\omega}f_{n},~&\mbox{在$Q_{T}$中},\\ 0=\Delta v_{n}-\gamma v_{n}+\delta u_{n},& \mbox{在$Q_{T}$中},\\ {\partial }_{\nu}u_{n}=0,\partial _{\nu}v_{n}=0,~~& \mbox{在$\Sigma _{T}$上},\\ u_{n}(x,0)=u_{0}(x),& x\in\Omega . \end{array}\right. $ (3.9)

根据步骤1的论述, 为证明方程(3.9)的局部零能控性, 我们先将其线性化如下

$\left\{\begin{array}{ll} -\Delta v_{n}^{z}+\gamma v_{n}^{z}=\delta z,~~& \mbox{在$\Omega $上},\\ {\partial }_{\nu}v_{n}^{z}=0,& \mbox{在$\partial \Omega $上} \end{array}\right. $ (3.10)

$\left\{\begin{array}{ll}\partial_{t}u_{n}^{z}-\Delta u_{n}^{z}-\nabla\cdot\left( B_{n}^{z}u_{n} ^{z}\right) +a_{n}^{z}u_{n}^{z}={\bf 1}_{\omega}f_{n}^{z},~~& \mbox{在$Q_{T}$上},\\ {\partial }_{\nu}u_{n}^{z}=0,& \mbox{在$\Sigma_{T}$上},\\ u_{n}^{z}(x,0)=u_{0}(x),& x\in\Omega ,\end{array}\right. $ (3.11)

其中$B_{n}^{z}=\nabla\Phi(v_{n}^{z}),a_{n}^{z}=H_{n}(z).$如果取$z\in Z,$则根据(3.3)-(3.6)式知$\{B_{n} ^{z}\}_{n=1}^{\infty}$$\{a_{n}^{z}\}_{n=1}^{\infty}$关于$n$$z\in Z$都在$L^{\infty}(Q_{T})$中一致有界.根据步骤1的分析, 非线性映射在$Z$中的不动点$u_{n}$为非线性方程(3.9)的解, 且

$\left\Vert f_{n}\right\Vert _{L^{\infty}(\omega\times(0,T))}\mbox{和} \left\Vert u_{n}\right\Vert _{\infty}+\left\Vert u_{n}\right\Vert _{W(0,T)} $

一致有界.从而存在在$L^{\infty}(\omega\times (0,T))$中弱*收敛于$f\in L^{\infty}(\omega \times(0,T))$的子列$\{f_{n_{k}}\}_{k=1}^{\infty} $和在$W(0,T)$中弱收敛于$u$的子列$\{u_{n_{k}}\}_{k=1}^{\infty}.$于是对方程(3.9)取$k\rightarrow \infty$的极限可知$u$为方程(1.1)对应于控制函数$f$的弱解, 且对几乎处处的$x\in\Omega $$u(x,T)=0$.

参考文献
[1] Bonami A, Hilhorst D, Logak E, et al. Singular limit of a chemotaxis-growth model. Adv Differential Equations, 2001, 6: 1173–1218.
[2] Keller E F, Segel L A. Initiation of slime mold aggregation viewed as an instability. J Theor Biol, 1970, 26: 399–415. DOI:10.1016/0022-5193(70)90092-5
[3] Hillen T, Painter K J. A user's guide to PDE models for chemotaxis. J Math Biol, 2009, 58: 183–217. DOI:10.1007/s00285-008-0201-3
[4] Horstmann D. From 1970 until present:the Keller-Segel model in chemotaxis and its consequences. Jahresber DMV, 2003, 105: 103–165.
[5] Zhong X, Jiang S. Globally bounded in-time solutions to a parabolic-elliptic system modeling chemotaxis. Acta Mathematica Scientia, 2007, 27B: 421–429.
[6] Ryu S U, Yagi A. Optimal control of Keller-Segel equations. J Math Anal Appl, 2001, 256: 45–66. DOI:10.1006/jmaa.2000.7254
[7] Guo B Z, Zhang L. Local null controllability for a chemotaxis system of parabolic-elliptic type. Systems & Control Letters, 2014, 65: 106–111.
[8] Guo B Z, Zhang L. Local exact controllability to positive trajectory for parabolic system of chemotaxis. Math Control Relat Fields, 2016, 6: 143–165. DOI:10.3934/mcrf
[9] Barbu V. Controllability of parabolic and Navier-Stokes equations. Sci Math Jpn, 2002, 56: 143–211.
[10] Ammar-Khodja F, Benabdallah A, Gonzalez-Burgos M, et al. Recent results on the controllability of linear coupled parabolic problems:a survey. Math Control Relat Fields, 2011, 1: 267–306. DOI:10.3934/mcrf
[11] Fernández-Cara E, Zuazua E. Null and approximate controllability for weakly blowing up semilinear heat equations. Ann Inst Henri Poincaré, Analyse non linéaire, 2000, 17: 583–616. DOI:10.1016/S0294-1449(00)00117-7
[12] Wang L, Wang G. The Bang-Bang principle of time optimal controls for the heat equation with internal controls. Systems & Control Letters, 2007, 56: 709–713.
[13] Wang G. L-Null Controllability for the heat equation and its consequences for the time optimal control problem. SIAM J Control & Optimization, 2008, 47: 1701–1720.
[14] Agmon S, Douglis A, Nirenberg L. Estimates near the boundary for solutions of elliptic partial differential equations satisfying general boundary conditions I. Comm Pure Appl Math, 1959, 12: 623–727. DOI:10.1002/(ISSN)1097-0312
[15] Ladyzhenskaja O A, Solonnikov V A, Ural'ceva N N. Linear and Quasi-linear Equations of Parabolic Type. Providence:AMS, 1968
[16] DiBenedetto E. Degenerate Parabolic Equations. New York: Springer-Verlag, 1993.
[17] Wang L, Wang G. The optimal time control of a phase-field system. SIAM J Control & Optimization, 2006, 42: 1483–1508.
[18] Barbu V. Analysis and Control of Nonlinear Infinite-Dimensional Systems. Boston: Academic Press, 1993.
[19] Rothe F. Global Solutions of Reaction-Diffusion systems. Berlin: Springer-Verlag, 1984.