数学物理学报, 2021, 41(3): 740-761 doi:

论文

一类耦合Korteweg-de Vries方程组输运系数反演问题的Lipschitz稳定性

吴斌,, 陈群

Lipschitz Stability for a Transport Coefficient Inverse Problem of a Linearly Coupled Korteweg-de Vries System

Wu Bin,, Chen Qun

通讯作者: 吴斌, E-mail: binwu@nuist.edu.cn

收稿日期: 2020-03-23  

基金资助: 国家自然科学基金.  11601240

Received: 2020-03-23  

Fund supported: the NSFC.  11601240

Abstract

This paper concerns an inverse problem of determining two spatially varying transport coefficients simultaneously in a linearly coupled Korteweg-de Vries (KdV) system with the first order terms. To obtain the stability result for the inverse problem with only one internal measurement data, we first prove a Carleman estimate including only one local integral for this coupled KdV system. Based on this Carleman estimate, we then obtain Lipschitz stability for the inverse problem under some priori information.

Keywords: Carleman estimate ; Coupled KdV system ; Lipschitz stability ; Coefficient inverse problem

PDF (425KB) 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

吴斌, 陈群. 一类耦合Korteweg-de Vries方程组输运系数反演问题的Lipschitz稳定性. 数学物理学报[J], 2021, 41(3): 740-761 doi:

Wu Bin, Chen Qun. Lipschitz Stability for a Transport Coefficient Inverse Problem of a Linearly Coupled Korteweg-de Vries System. Acta Mathematica Scientia[J], 2021, 41(3): 740-761 doi:

1 引言

$ T>0 $, $ L>0 $, $ \Omega: = (0, L) $$ \Omega_T: = \Omega\times (0, T) $. 本文考虑一类带有一阶导数项耦合的线性Korteweg-de Vires (KdV) 方程组

$ \begin{eqnarray} \left\{\begin{array}{ll} u_t+ u_{xxx}+ \left(M_1(x)u\right)_{x}+k_1v_x = 0, &(x, t)\in \Omega_T, \\ v_t+ v_{xxx}+\left(M_2(x)v\right)_{x}+k_2u_x = 0, &(x, t)\in \Omega_T, \\ u(0, t) = g_1(t), \ u(L, t) = g_2(t), \ u_x(L, t) = g_3(t), &t\in (0, T), \\ v(0, t) = h_1(t), \ v(L, t) = h_2(t), \ v_x(L, t) = h_3(t), &t\in (0, T), \\ u(x, 0) = u_0(x), \quad v(x, 0) = v_0(x), &x\in\Omega, \end{array}\right. \end{eqnarray} $

其中常数$ k_1 $, $ k_2\in {{\Bbb R}} $表示色散系数, 仅依赖于$ x\in\Omega $的输运系数$ M_1(x) $, $ M_2(x) $表示相应线性波的速度.

方程组(1.1) 可以看作是包含两个耦合长波的KdV方程组[13]

$ \begin{eqnarray} \left\{\begin{array}{ll} u_t+ \lambda_1u_{xxx}+ \alpha_1 u u_x+\delta u_x+k_1v_x = 0, &(x, t)\in \Omega_T, \\ v_t+\lambda_2 v_{xxx}+\alpha_2 v v_x-\delta v_x+k_2u_x = 0, &(x, t)\in \Omega_T \end{array}\right. \end{eqnarray} $

的线性化形式. 这个方程组是用以描述分层流体中两个长重力内波的强相互作用. 对于更多的物理细节, 参见文献[12, 16].

本文的目的是研究方程组(1.1) 中输运系数的反演问题. 更确切地说, 令$ \omega\subset \Omega $$ \Omega $的一个适当子区域, $ t_0 $是某给定时刻, 本文希望通过下列测量数据

$ \begin{eqnarray} v|_{\omega_T}\quad {\rm{及}}\quad (u, v)(x, t_0), \quad x\in\Omega, \ \ t_0 = \frac{T}{2} \end{eqnarray} $

来确定$ M_1(x) $$ M_2(x) $.

注 1.1   实际上, 只要对势函数$ \varphi $作微小修改, 我们就可用满足$ 0<t^*_0<T $的任意时刻$ t^*_0 $替换$ t_0 = \frac{T}{2} $. 比如, 若选取势函数为$ \varphi(x, t) = \frac{e^{\lambda d(x)}-e^{2\lambda\|d\|_{C(\overline \Omega)}}}{(t-T_1)(T_2-t)} $, 其中$ 0\leq T_1<T_2\leq T $, 那么$ t_0 $就可以选为$ \frac{T_1+T_2}{2} $.

卡勒曼估计是证明反问题稳定性的重要工具, 其是一类与微分算子相关的带权能量估计. 卡勒曼估计有着许多重要应用, 如唯一延拓性问题[29-30], 微分系统的控制理论及稳定化[17, 20, 31], 系数反演问题的稳定性[3-5, 21-23, 27, 32, 34-35]等. 最近, 卡勒曼估计被应用于构造系数反演问题的全局收敛数值方法, 这在实际中有着非常重要的应用, 参见文献[24-25].

关于经典KdV方程的卡勒曼估计, 文献[9, 11, 15, 28] 中证明了不同边界条件下的卡勒曼估计. 文献[8] 中证明了一类更为精巧的依赖于色散系数的卡勒曼估计. 然后, 该卡勒曼估计被应用于讨论线性KdV方程的一致零控问题. 这些文章大部分都是研究单个KdV方程的边界控制. 因此所用的卡勒曼估计都与边界积分项相关. 文献[7] 证明了单个KdV方程在齐次边界$ u(0, t) = u(L, t) = u_x(L, t) = 0 $下带有局部积分项的卡勒曼估计. 另外, 文献[14] 中证明了一类随机KdV方程的卡勒曼估计.

涉及KdV方程的反问题的文章很少. 文献[1] 中得到了一类由边界数据确定主部项系数的反演问题的Lipschitz稳定性. 在更复杂的边界条件下, 文献[10]中证明了一个类似反问题的稳定性, 不同的是文中使用的是部分边界测量数据及某时刻的全空间测量数据. 最近, 文献[26] 中构造了一个Tikhonov正则化泛函的极小化问题来处理KdV方程的输运系数反问题. 作者证明了该Tikhonov泛函的Fréchet可导性以及Fréchet导数的Lipschitz连续性. 然而, 还没有文章研究过耦合KdV方程组的系数反演问题.

本文的主要目的是证明仅利用$ v $的内部测量数据同时反演两个输运系数$ M_1(x) $, $ M_2(x) $的Lipschitz稳定性. 为此, 我们不得不将$ u $$ v $的卡勒曼估计结合起来, 然后应用方程之间的耦合消除$ u $的局部积分项. 因为方程组相互耦合的是一阶项, 为消去卡勒曼估计中$ u $的局部项, 我们需要得到一个包含局部项$ \int_{\varpi_T}|u_x|^2e^{2s\varphi}{\rm d}x{\rm d}t $的卡勒曼估计, 其中$ \varpi\subset\omega $, $ \varphi $是卡勒曼估计中的势函数. 然而, 由于缺乏$ u $在边界$ \partial\varpi $上的齐次Dirichlet边界条件, 我们无法得到下面的估计

其中$ \sigma_1 $$ \sigma_2 $是适当的函数. 基于此, 我们转而研究$ (u_x, v_x) $满足的方程组的卡勒曼估计. 此方程组附加的是非齐次的Neumann边界条件. 更确切地说, 我们考虑下面$ (\xi, \eta) $满足的耦合KdV方程组

$ \begin{eqnarray} \left\{\begin{array}{ll} \xi_t+ \xi_{xxx}+(M_1(x)\xi)_x+k_1\eta_x = f_1, &(x, t)\in \Omega_T, \\ \eta_t+ \eta_{xxx}+(M_2(x)\eta)_{x}+k_2\xi_x = f_2, &(x, t)\in \Omega_T, \end{array}\right. \end{eqnarray} $

带有非齐次的Neumann边界条件

$ \begin{eqnarray} \left\{\begin{array}{ll}\xi(L, t) = 0, \ \xi_{xx}(0, t) = \theta_1(t), \ \xi_{xx}(L, t) = 0, &t\in (0, T), \\ \eta(L, t) = 0, \ \eta_{xx}(0, t) = \theta_2(t), \ \eta_{xx}(L, t) = 0, &t\in (0, T).\end{array}\right. \end{eqnarray} $

为处理我们的反问题, 我们需要证明问题(1.4)–(1.5) 的卡勒曼估计. 值得注意的是, 涉及带有Neumann边界条件的KdV方程组的零控问题及反问题的文章较少. 据我们所知, 文献[6] 是讨论带有Neumann边界条件$ u_x(L, t) = u_{xx}(0, t) = u_{xx}(L, t) = 0 $的KdV方程的边界控制的唯一一篇文章.

现在我们给出定理1.1中所需的条件.

(A1)   $ \omega\subset\Omega $是含在$ \Omega $内端点$ x = L $的某个领域, 即$ \omega = (L_0, L) $, 其中$ 0<L_0<L $;

(A2)   $ (u, v), (\tilde u, \tilde v)\in C^{2}([0, T];W^{2, \infty}(\Omega))\cap C([0, T];H^4(\Omega)) $;

(A3)   存在充分小的正常数$ \epsilon_0 $, 使得$ M_2(x) = \tilde M_2(x) $, $ x\in(L-\epsilon_0, L) $.

注 1.2   在方程组(1.4)–(1.5) 的卡勒曼估计的证明中, 我们需要消除边界项$ \xi_x|_{x = L} $, $ \eta_x|_{x = L} $. 条件(A1) 是用于克服$ \xi_x, \eta_x $$ x = L $处边界条件的缺失. 更确切地说, 我们将应用$ \omega $上的局部积分项及其他全局积分项来控制这些边界项.

注 1.3   条件(A2) 是关于$ u $$ v $的先验界, 这可以通过提升$ g_1, g_2, g_3, h_1, h_2, h_3 $及初始数据$ u_0, v_0 $的光滑性而得到. 类似的条件在讨论微分方程系数反问题中是经常引入的.

注 1.4   条件(A3) 意味着我们事先已知$ M_2 $在边界$ x = L $处的先验信息, 其将被应用于消除$ u $$ \omega $上的局部测量.

另外, 对于正常数$ a_i, b_i, c_i, d_i\ (i = 1, 2) $及充分小的常数$ \epsilon_1>0 $, 我们定义$ M_1 $, $ M_2 $的容许集$ {\cal M} $

其中$ (u, v)[M_1, M_2] $是方程组(1.1) 对应于$ (M_1, M_2) $的解.

注 1.5   类似于文献[2, 33], 引入$ \epsilon_1>0 $是为了将一阶微分算子的卡勒曼估计应用于$ M_1 $$ M_2 $, 详见引理4.1. 然而, 由于缺乏KdV方程的极值原理, 我们不知道如何附加适当的条件能够保证$ \epsilon_1 $的存在性.

现在我们给出本文的主要结果, 其为利用观测数据(1.3) 反演输运系数$ M_1, M_2 $的Lipschitz稳定性.

定理 1.1   设$ (M_1, M_2), (\tilde M_1, {\tilde M}_2)\in {\cal M} $, $ k_2\neq 0 $及条件(A1)–(A3) 成立. 则存在正常数$ C = C(x_0, \Omega, T, \epsilon_1, \|M_1\|_{C^2(\overline\Omega)}, \|M_2\|_{C^2(\overline\Omega)}, k_1, k_2) $, 使得

$ \begin{eqnarray} & &\|{ M}_1-{\tilde M}_1\|_{H^2(\Omega)}+\|{ M}_2-{\tilde M}_2\|_{H^2(\Omega)}\\ &\leq & C\|v-\tilde v\|_{H^2(0, T;L^2(\omega))}+\|(u-\tilde u)(\cdot, t_0)\|_{H^4(\Omega)}+\|(v-\tilde v)(\cdot, t_0)\|_{H^4(\Omega)}, \end{eqnarray} $

其中$ (u, v) $$ (\tilde u, \tilde v) $是方程组(1.1) 分别对应于$ (M_1, M_2) $$ ({\tilde M}_1, \tilde M_2) $的两个解.

注 1.6   值得注意的是, 定理1.1中的稳定性没有用到$ u $$ \omega_T $上的测量数据. 实际上, 我们将应用方程组的耦合关系消除$ u $$ \omega_T $上的测量.

注 1.7   为了将KdV方程的解延拓至$ (-T, 0) $, 在文献[1] 中作者不得不引入对初始数据及输运系数的对称性假设, 即$ M_1(x) = M_1(L-x), M_2(x) = M_2(L-x) $$ u_{0, xxx}(x) = u_{0, xxx}(L-x) $, $ v_{0, xxx}(x) = v_{0, xxx}(L-x) $, $ x\in\Omega $. 这意味着文献[1] 中仅仅确定了一半区间$ [0, \frac{L}{2}] $上的未知函数. 本文在附加$ t_0 $时刻的测量下消除了这种对称性假设.

本文的其余部分安排如下: 在第2节, 我们给出了耦合KdV方程组(1.1) 的解的适定性. 第3节证明这类耦合KdV方程组的仅包含一个局部积分项的卡勒曼估计. 最后, 在第4节中我们证明输运系数反演问题的Lipschitz稳定性, 即定理1.1.

2 适定性

本节我们将证明下面带有一阶项耦合的线性KdV方程组的解的适定性

$ \begin{eqnarray} \left\{\begin{array}{ll} u_t+ u_{xxx}+ \left(M_1(x)u\right)_{x}+k_1v_x = 0, &(x, t)\in \Omega_T, \\ v_t+ v_{xxx}+\left(M_2(x)v\right)_{x}+k_2u_x = 0, &(x, t)\in \Omega_T, \\ u(0, t) = g_1(t), \ u(L, t) = g_2(t), \ u_{x}(L, t) = g_3(t), &t\in (0, T), \\ v(0, t) = h_1(t), \ v(L, t) = h_2(t), \ v_{x}(L, t) = h_3(t), &t\in (0, T), \\ u(x, 0) = u_0(x), \ v(x, 0) = v_0(x), &x\in\Omega. \end{array}\right. \end{eqnarray} $

首先, 我们引入函数变换将边界齐次化. 令

那么方程组(2.1) 可变形为

$ \begin{eqnarray} \left\{\begin{array}{ll} \varrho_t+ \varrho_{xxx}+ \left(M_1(x)\varrho\right)_{x}+k_1\vartheta_x = F, &(x, t)\in \Omega_T, \\ \vartheta_t+\vartheta_{xxx}+\left(M_2(x)\vartheta\right)_{x}+k_2\varrho_x = G, &(x, t)\in \Omega_T, \\ \varrho(0, t) = \varrho(L, t) = \varrho_{x}(L, t) = 0, &t\in (0, T), \\ \vartheta(0, t) = \vartheta(L, t) = \vartheta_{x}(L, t) = 0, &t\in (0, T), \\ \varrho(x, 0) = \varrho_0(x), \quad \vartheta(x, 0) = \vartheta_0(x), &x\in\Omega, \end{array}\right. \end{eqnarray} $

其中

下面我们先证明方程组(2.2) 在适当的Sobolev空间中解的存在性和唯一性. 由此, 我们立刻可以推知方程组(2.1) 的适定性.

为简便计, 令

其范数定义为

引理 2.1   设$ M_1, M_2\in C^{1}(\overline\Omega) $, $ F, G\in L^2(\Omega_T) $$ \varrho_0, \vartheta_0\in L^2(\Omega) $. 则方程组(2.2) 存在唯一解$ (\varrho, \vartheta)\in {{\cal X}}_T $, 且成立以下估计

$ \begin{eqnarray} \|(\varrho, \vartheta)\|_{{{\cal X}}_T}\leq C\left(\|F\|_{L^2(\Omega_T)}+\|G\|_{L^2(\Omega_T)}+\|\varrho_0\|_{L^2(\Omega)}+\|\vartheta_0\|_{L^2(\Omega)}\right), \end{eqnarray} $

其中$ C $是依赖于$ \Omega, T, \|M_1\|_{C^{1}(\overline\Omega)}, \|M_2\|_{C^{1}(\overline\Omega)} $$ k_1, k_2 $的常数.

  我们应用不动点方法证明方程组(2.2) 的解的存在性和唯一性. 为此, 定义函数集

$ (\hat\varrho, \hat\vartheta)\in {{\cal B}}_R $. 考虑

$ \begin{eqnarray} \left\{\begin{array}{ll} \varrho_t+ \varrho_{xxx} = -\left(M_1(x)\hat\varrho\right)_{x}-k_1\hat\vartheta_x+F, &(x, t)\in \Omega_T, \\ \vartheta_t+\vartheta_{xxx} = -(M_2(x)\hat\vartheta)_{x}-k_2\hat\varrho_x+G, &(x, t)\in \Omega_T, \\ \varrho(0, t) = \varrho(L, t) = \varrho_{x}(L, t) = 0, &t\in (0, T), \\ \vartheta(0, t) = \vartheta(L, t) = \vartheta_{x}(L, t) = 0, &t\in (0, T), \\ \varrho(x, 0) = \varrho_0(x), \ \ \vartheta(x, 0) = \vartheta_0(x), &x\in\Omega. \end{array}\right. \end{eqnarray} $

注意到算子$ {\cal A}[z]: = -z_{xxx} $及其伴随算子在$ D({\cal A}) = \{ z\in H^3(\Omega): \ z(0) = z(L) = z_x(L) = 0\} $上都是耗散的. 对于$ -\left(M_1(x)\hat\varrho\right)_{x}-k_1\hat\vartheta_x+F, -(M_2(x)\hat\vartheta)_{x}-k_2\hat\varrho_x+G\in L^2(\Omega_T) $$ \varrho_0, \vartheta_0\in L^2(\Omega) $, 由标准的算子半群理论, 我们可得方程组(2.4) 存在唯一解$ (\varrho, \vartheta)\in \left(L^\infty(0, T;L^2(\Omega))\right)^2 $. 另外, 由标准的能量估计可知$ (\varrho, \vartheta)\in \left(L^2(0, T;H^1(\Omega))\right)^2 $, 于是$ (\varrho, \vartheta)\in {\cal X}_T $. 从而, 由(2.4) 式定义的解映射

$ \begin{equation} S: {\cal B}_R\rightarrow {\cal X}_T, \quad (\hat \varrho, \hat \vartheta)\mapsto (\varrho, \vartheta) \end{equation} $

是有意义的.

现在我们证明当$ T $充分小时, $ S $$ {\cal B}_R $$ {\cal B}_R $的映到, 即

$ \begin{eqnarray} \|(\varrho, \vartheta)\|_{{\cal X}_T}\leq R. \end{eqnarray} $

$ (x+L)\varrho, (x+L)\vartheta $分别乘以$ \varrho $, $ \vartheta $的方程, 在$ \Omega_t: = \Omega\times(0, t) $上积分, 并利用Young不等式, 我们得到

$ \begin{eqnarray} &&\frac{1}{2}\int_\Omega (x+L)\left(|\varrho(x, t)|^2+|\vartheta(x, t)|^2\right){\rm d}x+\frac{3}{2}\int_{\Omega_t}(|\varrho_x(x, s)|^2+|\vartheta_x(x, s)|^2){\rm d}x{\rm d}s\\ & = &\frac{1}{2}\int_\Omega (x+L)\left(|\varrho_0(x)|^2+|\vartheta_0(x)|^2\right){\rm d}x-\frac{L}{2}\int_0^t (|\varrho_x(0, s)|^2+|\vartheta_x(0, s)|^2){\rm d}x\\ &&+\int_{\Omega_t}\left(M_1(x) \hat\varrho(x, s)+k_1 \hat\vartheta(x, s)\right)\left((x+L)\varrho_x(x, s)+\varrho(x, s)\right){\rm d}x{\rm d}s\\ &&+\int_{\Omega_t}\left(M_2(x)\hat\vartheta(x, s)+k_2 \hat\varrho(x, s)\right)\left((x+L)\vartheta_x(x, s)+\vartheta(x, s)\right){\rm d}x{\rm d}s\\ &&+\int_{\Omega_t}(x+L)(F(x, s)\varrho(x, s)+G(x, s)\vartheta(x, s)){\rm d}x{\rm d}s\\ &\leq & C\int_\Omega\left(|\varrho_0(x)|^2+|\vartheta_0(x)|^2\right){\rm d}x+\epsilon\int_{\Omega_t}(|\varrho_x(x, s)|^2+|\vartheta_x(x, s)|^2){\rm d}x{\rm d}s\\ &&+C(\epsilon)\int_{\Omega_t}(|\hat \varrho(x, s)|^2+|\hat \vartheta(x, s)|^2){\rm d}x{\rm d}s+C\int_{\Omega_t}\left(|F(x, s)|^2+|G(x, s)|^2\right){\rm d}x{\rm d}s\\ &&+C\int_{\Omega_t} (x+L)\left(|\varrho(x, s)|^2+|\vartheta(x, s)|^2\right){\rm d}x{\rm d}s, \end{eqnarray} $

其蕴含着当$ \epsilon $充分小时, 成立

$ \begin{eqnarray} &&\int_\Omega \left(|\varrho(x, t)|^2+|\vartheta(x, t)|^2\right){\rm d}x+\int_{\Omega_t}(|\varrho_x(x, s)|^2+|\vartheta_x(x, s)|^2){\rm d}x{\rm d}s\\ &\leq& C\int_{\Omega_t} \left(|\varrho(x, s)|^2+|\vartheta(x, s)|^2\right){\rm d}x{\rm d}s+C\int_\Omega\left(|\varrho_0(x)|^2+|\vartheta_0(x)|^2\right){\rm d}x\\ &&+C\int_{\Omega_t}(|\hat \varrho(x, s)|^2+|\hat \vartheta(x, s)|^2){\rm d}x{\rm d}s+C\int_{\Omega_t}\left(|F(x, s)|^2+|G(x, s)|^2\right){\rm d}x{\rm d}s. \end{eqnarray} $

进一步, 可得

$ \begin{eqnarray} &&\int_\Omega\left(|\varrho(x, t)|^2+|\vartheta(x, t)|^2\right){\rm d}x+\int_{\Omega_t}(|\varrho_x(x, s)|^2+|\vartheta_x(x, s)|^2){\rm d}x{\rm d}s\\ &\leq & C\int_{\Omega_t}\left(|\varrho(x, s)|^2+|\vartheta(x, s)|^2\right){\rm d}x{\rm d}s+CT\left(\|\hat \varrho\|_{L^\infty(0, T;L^2(\Omega))}^2+\|\hat \vartheta\|_{L^\infty(0, T;L^2(\Omega))}^2\right)\\ &&+C\left(\|\varrho_0\|^2_{L^2(\Omega)}+\|\vartheta_0\|^2_{L^2(\Omega)}+\|F\|^2_{L^2(\Omega_T)}+\|G\|^2_{L^2(\Omega_T)}\right), \quad t\in (0, T). \end{eqnarray} $

然后, 应用Gronwall不等式可得

$ \begin{eqnarray} \|(\varrho, \vartheta)\|_{{\cal X}_T}\leq C\left(\|\varrho_0\|_{L^2(\Omega)}+\|\vartheta_0\|_{L^2(\Omega)}+\|F\|_{L^2(\Omega_T)}+\|G\|_{L^2(\Omega_T)}+T^{\frac{1}{2}}R\right)e^{CT}. \end{eqnarray} $

现在我们取

$ \begin{eqnarray} R = 2C\left(\|\varrho_0\|_{L^2(\Omega)}+\|\vartheta_0\|_{L^2(\Omega)}+\|F\|_{L^2(\Omega_T)}+\|G\|_{L^2(\Omega_T)}\right), \end{eqnarray} $

则存在充分小的$ T_0 $, 使得(2.6) 式对于所有满足$ 0<T\leq T_0 $$ T $成立.

另一方面, 设$ (\hat \varrho_1, \hat \vartheta_1), (\hat \varrho_2, \hat \vartheta_2)\in {\cal B}_R $, 定义$ (\varrho_1, \vartheta_1) = S(\hat \varrho_1, \hat \vartheta_1) $$ (\varrho_2, \vartheta_2) = S(\hat \varrho_2, \hat \vartheta_2) $. 考虑

$ \begin{eqnarray} \left\{\begin{array}{ll} (\varrho_1-\varrho_2)_t+ (\varrho_1-\varrho_2)_{xxx} = -\left(M_1(x)(\hat\varrho_1-\hat \varrho_2)\right)_{x}-k_1(\hat\vartheta_1-\hat\vartheta_2)_x, &(x, t)\in \Omega_T, \\ (\vartheta_1-\vartheta_2)_t+ (\vartheta_1-\vartheta_2)_{xxx} = -(M_2(x)(\hat\vartheta_1-\hat\vartheta_2))_{x}-k_2(\hat\varrho_1-\hat\varrho_2)_x, &(x, t)\in \Omega_T, \\ (\varrho_1-\varrho_2)(0, t) = (\varrho_1-\varrho_2)(L, t) = (\varrho_1-\varrho_2)_{x}(L, t) = 0, &t\in (0, T), \\ (\vartheta_1-\vartheta_2)(0, t) = (\vartheta_1-\vartheta_2)(L, t) = (\vartheta_1-\vartheta_2)_{x}(L, t) = 0, &t\in (0, T), \\ (\varrho_1-\varrho_2)(x, 0) = 0, \ \ (\vartheta_1-\vartheta_2)(x, 0) = 0, &x\in\Omega. \end{array}\right. \end{eqnarray} $

类似于(2.9) 式的计算, 我们有

$ \begin{eqnarray} &&\int_\Omega \left(|(\varrho_1-\varrho_2)(x, t)|^2+|(\vartheta_1-\vartheta_2)(x, t)|^2\right){\rm d}x\\ &&+\int_{\Omega_t}(|(\varrho_1-\varrho_2)_x(x, s)|^2+|(\vartheta_1-\vartheta_2)_x(x, s)|^2){\rm d}x{\rm d}s\\ &\leq & C\int_{\Omega_t}\left(|(\varrho_1-\varrho_2)(x, s)|^2+|(\vartheta_1-\vartheta_2)(x, s)|^2\right){\rm d}x{\rm d}s\\ &&+CT\left(\|\hat \varrho_1-\hat \varrho_2\|_{L^\infty(0, T;L^2(\Omega))}^2+\|\hat \vartheta_1-\hat \vartheta_2\|_{L^\infty(0, T;L^2(\Omega))}^2\right). \end{eqnarray} $

再次应用Gronwall不等式, 可得

$ \begin{equation} \|(\varrho_1-\varrho_2, \vartheta_1-\vartheta_2)\|_{{\cal X}_T} \leq CT^{\frac{1}{2}}e^{CT} \|(\hat \varrho_1-\hat\varrho_2, \hat\vartheta_1-\hat\vartheta_2)\|_{{\cal X}_T} \leq \frac{1}{2}\|(\hat \varrho_1-\hat\varrho_2, \hat\vartheta_1-\hat\vartheta_2)\|_{{\cal X}_T} \end{equation} $

对于所有满足$ T\leq T_0 $$ T $成立, 这里$ T_0 $满足$ CT_0^{\frac{1}{2}}e^{CT_0}\leq \frac{1}{2} $.

从而当$ T\leq T_0 $时, $ S $$ {\cal B}_R $上的压缩映像. 由Banach不动点定理可得方程组(2.2) 存在唯一解$ (\varrho, \vartheta)\in {\cal X}_{T_0} $. 因为$ T_0 $仅仅依赖于$ \Omega, T, M_1, M_2 $$ k_1, k_2 $, 而与$ F, G $$ \varrho_0, \vartheta_0 $无关, 所以重复上述过程有限次, 对于任意的$ T $, 我们可以得到方程组(2.2) 在$ (0, T) $上存在唯一的全局解.

最后, 利用(2.8)式($ \varrho, \vartheta $替换$ \hat \varrho, \hat \vartheta $) 及Gronwall不等式可得(2.3)式. 证毕.

由引理2.1, 结合$ (\varrho, \vartheta) $$ (u, v) $之间的联系, 我们即可得到方程组(2.1) 的解的适定性.

定理 2.2   设$ M_1, M_2\in C^{1}(\overline\Omega) $, $ g_1, g_2, g_3, h_1, h_2, h_3\in H^1(0, T) $$ u_0, v_0\in L^2(\Omega) $. 则方程组(2.1) 存在唯一解$ (u, v)\in {\cal X}_T $, 且成立以下估计

$ \begin{eqnarray} \|(u, v)\|_{{\cal X}_T}\leq &&C\left(\|g_1\|_{H^1(0, T)}+\|g_2\|_{H^1(0, T)}+\|g_3\|_{H^1(0, T)}+\|h_1\|_{H^1(0, T)}\right.\\ &&+\|h_2\|_{H^1(0, T)}+\|h_3\|_{H^1(0, T)}+\left.\|u_0\|_{L^2(\Omega)}+\|v_0\|_{L^2(\Omega)}\right), \end{eqnarray} $

其中$ C $是依赖于$ \Omega, T, \|M_1\|_{C^{1}(\overline\Omega)}, \|M_2\|_{C^{1}(\overline\Omega)} $$ k_1, k_2 $的常数.

3 卡勒曼估计

考虑耦合KdV方程组

$ \begin{eqnarray} \left\{\begin{array}{ll} p_t+ p_{xxx}+(M_1(x)p)_x+k_1q_x = f_1, &(x, t)\in \Omega_T, \\ q_t+ q_{xxx}+(M_2(x)q)_{x}+k_2p_x = f_2, &(x, t)\in \Omega_T, \end{array}\right. \end{eqnarray} $

其满足边界条件

$ \begin{equation} \left\{\begin{array}{ll}p(0, t) = p(L, t) = p_{x}(L, t) = 0, &t\in (0, T), \\ q(0, t) = q(L, t) = q_{x}(L, t) = 0, &t\in (0, T).\end{array}\right. \end{equation} $

本节我们将证明方程组(3.1) 仅带有$ q $局部观测的卡勒曼估计. 这是证明我们反问题稳定性的关键步骤. 为了消除$ p $的局部积分项, 我们不得不转而研究$ (p_x, q_x) $满足的带有非齐次Neumann边界条件耦合KdV方程组的卡勒曼估计. 我们将结合$ p $的卡勒曼估计和$ (p_x, q_x) $的卡勒曼估计来证明我们需要的卡勒曼估计.

为构造我们的卡勒曼估计, 我们引入一些记号. 令$ d(x) = (x-x_0)^2 $, 其中$ x_0\in (-\infty, -\frac{1}{2}) $是某一固定点. 显然, 函数$ d $满足

选择这样的$ d $是为了处理Neumann边界条件带来的困难.

进一步, 我们定义

$ \begin{equation} \varphi(x, t) = \frac{e^{\lambda d(x)}-e^{2\lambda\|d\|_{C(\overline{\Omega})}}}{t(T-t)}, \quad\phi(x, t) = \frac{e^{\lambda d(x)}}{t(T-t)}, \end{equation} $

其中$ \lambda $是一个充分大的正参数.

我们还需引入截断函数$ \rho_i\in C^2(\overline\Omega) $$ (1\leq i\leq 4) $满足$ 0\leq \rho_i(x)\leq 1 $, $ x\in\Omega $

$ \begin{eqnarray} \left\{\begin{array}{ll} \rho_i(x) = 1, &x\in \omega^{(i)}, \\ \rho_i(x) = 0, &x\in \overline\Omega\setminus \omega^{(i+1)}, \end{array} \right. \end{eqnarray} $

其中$ \omega^{(i)}\subset \Omega $满足

本节的主要结果是方程组(3.1)–(3.2) 的仅带有$ q $局部积分项的卡勒曼估计.

定理 3.1   设$ f_1, f_2\in H^1(Q_T) $满足$ f_1(L, t) = f_2(L, t) = 0 $, $ t\in (0, T) $$ f_2(x, t) = 0 $, $ (x, t)\in(L-\epsilon_0, L)\times (0, T) $, 且$ M_1, M_2\in C^{2}(\overline\Omega) $, $ k_2\neq 0 $. 则存在正常数$ \lambda_0 = \lambda_0(x_0, \Omega, T, $$ \|M_1\|_{C^2(\overline\Omega)}, $$ \|M_2\|_{C^2(\overline\Omega)}, k_1, k_2) $, $ s_0 = s_0(\lambda, x_0, \Omega, T, $$ \|M_1\|_{C^2(\overline\Omega)}, \|M_2\|_{C^2(\overline\Omega)}, k_1, k_2) $$ C = C(\lambda, $$ x_0, \Omega, T, \|M_1\|_{C^2(\overline\Omega)}, $$ \|M_2\|_{C^2(\overline\Omega)} $, $ k_1, k_2) $, 使得

$ \begin{eqnarray} &&{{\cal I}}(p_x)+{{\cal I}}(q_x)+s{{\cal I}}(p)+s{{\cal I}}(q)+{\cal B} (p_x)|_{x = 0}+{\cal B} (q_x)|_{x = 0}\\ &\leq& C\int_{\Omega_T}s\phi(|f_1|^2+|f_{1, x}|^2+|f_2|^2+|f_{2, x}|^2)e^{2s\varphi}{\rm d}x{\rm d}t+Cs^{35}e^{Cs}\|q\|^2_{L^2(\omega_T)} \end{eqnarray} $

对于所有的$ \lambda\geq \lambda_0 $, $ s\geq s_0 $和所有满足(3.1)–(3.2) 的$ p, q\in L^2(0, T;H^3(\Omega)) $都成立, 其中

3.1 带非齐次Neumann边界KdV方程的卡勒曼估计

本节研究满足非齐次Neumann边界条件

$ \begin{eqnarray} y(L, t) = 0, \ y_{xx}(0, t) = \theta(t), \ y_{xx}(L, t) = 0, \quad t\in (0, T) \end{eqnarray} $

的KdV算子

$ \begin{equation} {\cal L}[y]: = y_t+ y_{xxx}, \quad (x, t)\in\Omega_T \end{equation} $

的卡勒曼估计.

引理 3.2   设$ \theta\in L^2(0, T) $. 则存在正常数$ \lambda_1 = \lambda_1(x_0, \Omega, T) $, $ s_1 = s_1(\lambda, x_0, \Omega, T) $$ C = C(x_0, \Omega, T) $, 使得

$ \begin{eqnarray} {{\cal I}}(y)+{{\cal B}}(y)|_{x = 0} & \leq& C\int_{\Omega_T}s\lambda\phi|{\cal L}[y]|^2e^{2s\varphi}{\rm d}x{\rm d}t+C\int_0^T(s\lambda\phi|\theta|^2e^{2s\varphi})|_{x = 0}{\rm d}t\\ &&+C\int_{{\omega^{(3)}_T}}s^9\lambda^6\phi^9|y|^2 e^{2s\varphi}{\rm d}x{\rm d}t \end{eqnarray} $

对于所有的$ \lambda\geq \lambda_1 $, $ s\geq s_1 $和所有满足(3.6)–(3.7) 式的$ y\in L^2(0, T;H^2(\Omega)) $都成立.

  令

$ \begin{equation} Y(x, t) = e^{s\varphi(x, t)}y(x, t), \quad (x, t)\in \Omega_T. \end{equation} $

直接计算$ {\cal L}[y] $, 可得

$ \begin{equation} e^{s\varphi}{\cal L}[y]+{\mathcal R}[Y] = {\mathcal L}_1[Y]+{\mathcal L}_2[Y], \end{equation} $

其中

$ \varphi $的定义, 显然可得

$ \begin{equation} \varphi|_{t = 0} = \varphi|_{t = T} = 0. \end{equation} $

另外, 由(3.10) 式得

$ \begin{equation} \|{\mathcal L}_1[Y]\|^2_{L^2(\Omega_T)}+\|{\mathcal L}_2[Y]\|^2_{L^2(\Omega_T)}+2({\cal L}_1[Y], {\cal L}_2[Y]) = \|e^{s\varphi}{\cal L}[y]+{\mathcal R}[Y]\|_{L^2(\Omega_T)}^2. \end{equation} $

$ \begin{equation} ({\cal L}_1[Y], {\cal L}_2[Y]) = \sum\limits_{i = 1}^3\sum\limits_{j = 1}^3I_{ij}, \end{equation} $

其中$ I_{ij} $表示$ {\cal L}_1[Y] $中第$ i $项与$ {\cal L}_2[Y] $中第$ j $项之间的内积. 类似于文献[7], 应用分部积分公式, 可得

我们将上述所有项分为三部分: $ L(Y) $, $ N(Y) $$ B(Y) $, 即

$ \begin{eqnarray} ({\cal L}_1[Y], {\cal L}_2[Y]) = L(Y)+N(Y)+B(Y), \end{eqnarray} $

其中

这里$ I_{ij}^k $$ I_{ij} $表达式中的第$ k $项.

下面我们通过估计$ L(Y) $, $ N(Y) $$ B(Y) $中的所有项来寻求$ ({\cal L}_1[Y], {\cal L}_2[Y]) $的正下界. 由$ \varphi $的定义得

$ \begin{eqnarray} \left\{\begin{array}{l} \varphi_x = \lambda d_x \phi, \quad \varphi_{xx} = \left(\lambda^2d_x^2+\lambda d_{xx}\right)\phi, \quad |\varphi_t|\leq Ce^{C\lambda}\phi^2, \\ |\varphi_{xxx}|+|\varphi_{xxxx}|\leq C\lambda^4\phi, \quad |\varphi_{xt}|\leq C\lambda\phi^2.\end{array}\right. \end{eqnarray} $

然后, 利用(3.15) 式和$ d_x>1 $, $ d_{xx}>0 $, 我们得到下面关于$ L(Y) $$ N(Y) $的估计

$ \begin{eqnarray} L(Y)& = &\frac{9}{2}\int_{\Omega_T}s\varphi_{xx}|Y_{xx}|^2{\rm d}x{\rm d}t+\frac{15}{2}\int_{\Omega_T}s^5\varphi_x^4\varphi_{xx}|Y|^2{\rm d}x{\rm d}t\\ &\geq &\frac{9}{2}\int_{{\Omega_T}}s\lambda^2\phi|Y_{xx}|^2{\rm d}x{\rm d}t+\frac{15}{2}\int_{\Omega_T}s^5\lambda^6\phi^5|Y|^2{\rm d}x{\rm d}t, \end{eqnarray} $

$ \begin{equation} N(Y)\geq -C\int_{{\Omega_T}}\left(s\lambda\phi^2|Y_x|^2+s^3\lambda^6\phi^4|Y|^2\right){\rm d}x{\rm d}t. \end{equation} $

为吸收(3.17) 式右端的第一项, 我们利用分部积分得到

这蕴含着

$ \begin{eqnarray} &&\int_{\Omega_T}s^3\lambda^4\phi^3|Y_x|^2{\rm d}x{\rm d}t\\ &\leq&\int_{{\Omega_T}}s\lambda^2\phi|Y_{xx}|^2{\rm d}x{\rm d}t+2\int_{\Omega_T}s^5\lambda^6\phi^5|Y|^2{\rm d}x{\rm d}t+2\int_0^T[s^3\lambda^4\phi^3YY_x]_{x = 0}^{x = L}{\rm d}t. \end{eqnarray} $

于是, 结合(3.16)–(3.18) 式及$ y = Ye^{-s\varphi} $, 我们得到对于充分大的$ s $$ \lambda $, 成立

$ \begin{eqnarray} L(Y)+N(Y) &\geq &\int_{{\Omega_T}}s\lambda^2\phi|y_{xx}|^2e^{2s\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T}s^3\lambda^4\phi^3|y_x|^2e^{2s\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T}s^5\lambda^6\phi^5|y|^2e^{2s\varphi}{\rm d}x{\rm d}t\\ &&-C\sum\limits_{x = \hat x}\int_0^T(s^2\lambda^4\phi^3|y_x|^2e^{2s\varphi}+s^4\lambda^6\phi^5|y|^2e^{2s\varphi})|_{\hat x = 0, L}{\rm d}t. \end{eqnarray} $

现在我们估计边界项$ B(Y) $. 首先, 因为$ d_x(0)>0 $$ d_x(L)>0 $, 所以我们有

$ \begin{eqnarray} I_{12}^2+I_{13}^2+I_{32}^4+I_{33}^2 & = & \int_0^T\left[-\frac{3}{2}s\varphi_x|Y_{xx}|^2-4s^3\varphi_x^3|Y_x|^2-\frac{3}{2}s^5\varphi_x^5|Y|^2\right]_{x = 0}^{x = L}{\rm d}t\\ &\geq & \int_0^T\left.\left(\frac{3}{2}s|\varphi_x||Y_{xx}|^2+4s^3|\varphi_x|^3|Y_x|^2+\frac{3}{2}s^5|\varphi_x|^5|Y|^2\right)\right|_{x = 0}{\rm d}t\\ &&-C\int_0^T\left(s|\varphi_x||Y_{xx}|^2+s^3|\varphi_x|^3|Y_x|^2+s^5|\varphi_x|^5|Y|^2\right)|_{x = L}{\rm d}t. \end{eqnarray} $

利用Hölder不等式, 我们得到

$ \begin{eqnarray} I_{22}^3\geq -C\sum\limits_{\hat x = 0, L}\int_0^T \left(|\varphi_x||Y_{xx}|^2{\rm d}t+s^2|\varphi_x|^3|Y_{x}|^2\right)|_{x = \hat x}{\rm d}t, \end{eqnarray} $

$ \begin{eqnarray} I_{32}^3 \geq-\frac{1}{2}\sum\limits_{\hat x = 0, L}\int_{0}^T(s|\varphi_{x}||Y_{xx}|^2+s^5|\varphi_x|^5|Y|^2)|_{x = \hat x}{\rm d}t \end{eqnarray} $

$ \begin{equation} I_{22}^4+I_{32}^5+I_{32}^6\geq -C\sum\limits_{\hat x = 0, L}\int_{0}^T (s^2|\varphi_x|^3|Y_x|^2+s^4|\varphi_x|^5|Y|^2)|_{x = \hat x}{\rm d}t. \end{equation} $

这里我们应用了$ |\varphi_{xx}|\leq C|\varphi_x|^2 $, $ |\varphi_{xxx}|\leq C|\varphi_x|^3 $$ |\left(\varphi_x^2\varphi_{xx}\right)_x|\leq C|\varphi_x|^5 $.

直接计算, 可得

$ \begin{eqnarray} \left\{\begin{array}{l} Y|_{x = 0, L} = (ye^{s\varphi})|_{x = 0, L}, \quad Y_x|_{x = 0, L} = (y_xe^{s\varphi}+s\varphi_xye^{s\varphi})|_{ x = 0, L}, \\ Y_{xx}|_{x = 0, L} = (y_{xx}e^{s\varphi}+2s\varphi_xy_xe^{s\varphi}+(s\varphi_{xx}+s^2\varphi_x^2)ye^{s\varphi})|_{ x = 0, L}. \end{array}\right. \end{eqnarray} $

利用(3.20)–(3.24) 式和Neumann边界条件(3.6), 有

$ \begin{eqnarray} B(Y)-I_{11}^3&\geq &\int_0^T\left(s|\varphi_x||Y_{xx}|^2+s^3|\varphi_x|^3|Y_x|^2+s^5|\varphi_x|^5|Y|^2\right)|_{x = 0}{\rm d}t\\ &&-C\int_0^T\left(s|\varphi_x||Y_{xx}|^2+s^3|\varphi_x|^3|Y_x|^2+s^5|\varphi_x|^5|Y|^2\right)|_{x = L}{\rm d}t\\ &\geq &\frac{1}{4}\int_0^T\left(s^3\lambda^3\phi^3|y_x|^2e^{2s\varphi}+s^5\lambda^5\phi^5|y|^2e^{2s\varphi}\right)|_{x = 0}{\rm d}t\\ &&-C\int_0^T \left(s^3\lambda^3\phi^3|y_x|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t. \end{eqnarray} $

对于$ I_{11}^3 $, 由(3.6) 和(3.24) 式得

$ \begin{eqnarray} I_{11}^3& = &-3\int_0^T\left[s\varphi_x(y_x+s\varphi_x y)(y_t+s\varphi_t y)e^{2s\varphi}\right]_{x = 0}^{x = L}{\rm d}t\\ & = &\int_0^T\left(-\frac{3}{2}s^2\left(\varphi_x^2e^{2s\varphi}\right)_t|y|^2+3s^2\varphi_x\varphi_tyy_xe^{2s\varphi}+3s^3\varphi_x^2\varphi_t |y|^2e^{2s\varphi}\right)|_{x = 0}{\rm d}t\\ &&+\int_0^T(3s\varphi_xy_xy_t e^{2s\varphi})|_{x = 0}{\rm d}t\\ &\geq &-Ce^{C\lambda}\int_0^t \left(s^3\lambda^2\phi^4|y|^2e^{2s\varphi}\right)|_{ x = 0}{\rm d}t-Ce^{C\lambda}\int_0^t \left(s\phi^2|y_x|^2e^{2s\varphi}\right)|_{ x = 0}{\rm d}t\\ &&+3\int_0^T\left(s\varphi_xy_xy_t e^{2s\varphi}\right)|_{x = 0}{\rm d}t. \end{eqnarray} $

另一方面, 用$ (s\varphi_x e^{2s\varphi})|_{x = 0}y_{xx} $乘以方程(3.7) 并应用分部积分, 我们有

$ \begin{eqnarray} &&-\int_0^T\left[(s\varphi_x e^{2s\varphi})|_{x = 0}y_xy_t\right]_{x = 0}^{x = L}{\rm d}t\\ & = &-\int_{\Omega_T}(s\varphi_xe^{2s\varphi})|_{x = 0}y_{xt}y_x{\rm d}x{\rm d}t-\int_{\Omega_T}(s\varphi_xe^{2s\varphi})|_{x = 0}y_{xx}(-y_{xxx}+{{\cal L}}[y]){\rm d}x{\rm d}t\\ & = &\frac{1}{2}\int_{\Omega_T}\left((s\varphi_xe^{2s\varphi})|_{x = 0}\right)_t|y_x|^2{\rm d}x{\rm d}t+\frac{1}{2}\int_{0}^T\left[(s\varphi_xe^{2s\varphi})|_{x = 0}|y_{xx}|^2\right]_{x = 0}^{x = L}{\rm d}t\\ &&-\int_{\Omega_T}(s\varphi_xe^{2s\varphi})|_{x = 0}y_{xx}{{\cal L}}[y]{\rm d}x{\rm d}t. \end{eqnarray} $

因为在$ \Omega $上有$ d(0)\leq d(x) $, 那么在$ \Omega_T $上我们有$ \phi|_{x = 0}\leq \phi $, $ \varphi|_{x = 0}\leq \varphi $. 结合(3.6) 式, 我们进一步有

$ \begin{eqnarray} \int_0^T\left(s\varphi_xy_xy_t e^{2s\varphi}\right)|_{x = 0}{\rm d}t &\geq &-Ce^{C\lambda}\int_{\Omega_T}s^2\lambda\phi^3|y_x|^2e^{2s\varphi}{\rm d}x{\rm d}t -C\int_0^T\left(s\lambda\phi|\theta|^2e^{2s\varphi}\right)|_{x = 0}{\rm d}t \\ &&-C\int_{\Omega_T}s\lambda\phi|y_{xx}|^2e^{2s\varphi}{\rm d}x{\rm d}t-C\int_{\Omega_T}s\lambda\phi |{\mathcal L}[y]|^2e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

因此, 结合(3.25), (3.26) 和(3.28)式, 可得到当$ s\geq e^{C\lambda} $时, 成立以下关于$ B(Y) $的估计

$ \begin{eqnarray} B(Y) &\geq&\frac{1}{8}\int_0^T\left(s^3\lambda^3\phi^3|y_x|^2e^{2s\varphi}+s^5\lambda^5\phi^5|y|^2e^{2s\varphi}\right)|_{x = 0}{\rm d}t-C\int_0^T \left(s^3\lambda^3\phi^3|y_x|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t\\ &&-C\int_0^T\left(s\lambda\phi|\theta|^2e^{2s\varphi}\right)|_{x = 0}{\rm d}t-C\int_{\Omega_T}s^3\lambda\phi^3|y_x|^2e^{2s\varphi}{\rm d}x{\rm d}t\\ &&-C\int_{\Omega_T}s\lambda\phi|y_{xx}|^2e^{2s\varphi}{\rm d}x{\rm d}t-C\int_{\Omega_T}s\lambda\phi |{\mathcal L}[y]|^2e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

从(3.19) 和(3.29) 式可得对于满足$ s\geq \max\{\lambda^2, e^{c\lambda}\} $的充分大的$ \lambda $$ s $, 有

$ \begin{eqnarray} 2({\cal L}_1[Y], {\cal L}_2[Y]) &\geq &\int_{{\Omega_T}}s\lambda^2\phi|y_{xx}|^2e^{2s\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T}s^3\lambda^4\phi^3|y_x|^2e^{2s\varphi}{\rm d}x{\rm d}t \\ &&+\int_{\Omega_T}s^5\lambda^6\phi^5|y|^2e^{2s\varphi}{\rm d}x{\rm d}t\!+\!\frac{1}{8}\int_0^T \left(s^3\lambda^3\phi^3|y_x|^2e^{2s\varphi}\!+\!s^5\lambda^5\phi^5|y|^2e^{2s\varphi}\right)|_{x = 0}{\rm d}t \\ &&-C\int_0^T\left(s\lambda\phi|\theta|^2e^{2s\varphi}\right)|_{x = 0}{\rm d}t -C\int_0^T \left(s^3\lambda^3\phi^3|y_x|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t{}\\ &&-C\int_{\Omega_T}s\lambda\phi |{\mathcal L}[y]|^2e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

将(3.30) 式代入(3.12) 式, 并应用

$ \begin{eqnarray} |{\cal R}[Y]|\leq C\left(se^{C\lambda} \phi^2+s\lambda^3\phi+s^2\lambda^3\phi^2\right)|y|e^{s\varphi}, \end{eqnarray} $

我们得到

$ \begin{eqnarray} {{\cal I}}(y) +{\cal B} (y)|_{x = 0} &\leq & C\int_{\Omega_T}s\lambda\phi|{\cal L}[y]|^2e^{2s\varphi}{\rm d}x{\rm d}t+C\int_0^T\left(s\lambda\phi|\theta|^2e^{2s\varphi}\right)|_{x = 0}{\rm d}t\\ &&+C\int_0^T \left(s^3\lambda^3\phi^3|y_x|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t. \end{eqnarray} $

现在估计(3.32) 右端的最后一项. 应用分部积分和Young不等式, 我们有

$ \begin{eqnarray} & &\int_0^T\left(s^3\lambda^3\rho_1\phi^3|y_x|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t\\ & = &2\int_{\Omega_T}s^3\lambda^3\rho_1\phi^3y_xy_{xx}e^{2s\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T}s^3\lambda^3\left(\rho_1\phi^3e^{2s\varphi}\right)_x |y_x|^2{\rm d}x{\rm d}t\\ & \leq & \epsilon\int_{\Omega_T}s\lambda^2\phi |y_{xx}|^2e^{2s\varphi}{\rm d}x{\rm d}t+C(\epsilon)\int_{\omega^{(2)}_T}s^5\lambda^4\phi^5 |y_x|^2e^{2s\varphi}{\rm d}x{\rm d}t, \end{eqnarray} $

其中$ \epsilon $是任意正常数. 另一方面, 对任意正常数$ \epsilon>0 $, 成立

$ \begin{eqnarray} &&\int_{\omega^{(2)}_T}s^5\lambda^4\phi^5 |y_x|^2e^{2s\varphi}{\rm d}x{\rm d}t\leq\int_{\omega^{(3)}_T} s^5\lambda^4\phi^5\rho_2 |y_x|^2e^{2s\varphi}{\rm d}x{\rm d}t\\ & = & -\int_{\omega^{(3)}_T} s^5\lambda^4\phi^5\rho_2yy_{xx}e^{2s\varphi}{\rm d}x{\rm d}t-\int_{\omega^{(3)}_T} s^5\lambda^4\left(\phi^5\rho_2e^{2s\varphi}\right)_xyy_{x}{\rm d}x{\rm d}t\\ &\leq &\epsilon \int_{\Omega_T}s\lambda^2\phi |y_{xx}|^2e^{2s\varphi}{\rm d}x{\rm d}t+\epsilon \int_{\Omega_T}s^3\lambda^4\phi^3 |y_{x}|^2e^{2s\varphi}{\rm d}x{\rm d}t +C(\epsilon)\int_{\omega^{(3)}_T}s^9\lambda^6\phi^9 |y|^2e^{2s\varphi}{\rm d}x{\rm d}t. {}\\ \end{eqnarray} $

由(3.33) 和(3.34) 式得

$ \begin{equation} \int_0^T\left(s^3\lambda^3\phi^3|y_x|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t \leq \epsilon{{\cal I}}(y)+C(\epsilon)\int_{\omega^{(3)}_T}s^9\lambda^6\phi^9 |y|^2e^{2s\varphi}{\rm d}x{\rm d}t. \end{equation} $

最后, 结合(3.32) 和(3.35) 式, 并选择$ \epsilon $充分小以吸收(3.35) 式右端的第一项, 我们得到(3.8)式. 证毕.

注 3.1   文献[7] 中证明了以下方程组

$ \begin{eqnarray} \left\{\begin{array}{ll}-v_t-\xi(x, t)v_x-v_{xxx} = 0, &(x, t)\in \Omega_T, \\ v(0, t) = v(L, t) = u_x(0, t) = 0, & t\in (0, T) \end{array}\right. \end{eqnarray} $

的带有局部项的卡勒曼估计

$ \begin{eqnarray} & &\int_{{\Omega_T}}s\tilde\varphi|v_{xx}|^2e^{2s\tilde\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T}s^3\tilde \varphi^3|v_x|^2e^{2s\tilde\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T}s^5\tilde\varphi^5|v|^2e^{2s\tilde\varphi}{\rm d}x{\rm d}t\\ &\leq&\int_{\omega_T}\left(s\tilde\varphi|v_{xx}|^2+s^3\tilde \varphi^3|v_x|^2+s^5\tilde\varphi^5|v|^2\right)e^{2s\tilde\varphi}{\rm d}x{\rm d}t, \end{eqnarray} $

其中$ \tilde\varphi $是一个适当的势函数. 然而, 我们不能直接应用这个估计来证明引理3.2. 一方面因为我们讨论的方程组的边界条件与(3.36) 式中的边界条件不同. 另一方面, 为研究我们的反问题, 我们不得不证明一个非齐次KdV方程的卡勒曼估计, 这使得我们无法应用类似于文献[7]或[15]的方法来消除$ y_{xx} $的局部项.

3.2 带Neumann边界耦合KdV方程组的卡勒曼估计

我们现在来证明带有Neumann边界的耦合KdV方程组的卡勒曼估计. 考虑

$ \begin{eqnarray} \left\{\begin{array}{ll} \xi_{t}+ \xi_{xxx}+(M_1(x)\xi)_{x}+k_1\eta_{x} = f_1, &(x, t)\in \Omega_T, \\ \eta_{t}+ \eta_{xxx}+(M_2(x)\eta)_{x}+k_2\xi_{x} = f_2, &(x, t)\in \Omega_T, \\ \xi(L, t) = 0, \ \xi_{xx}(0, t) = \theta_1(t), \ \xi_{xx}(L, t) = 0, &t\in (0, T), \\ \eta(L, t) = 0, \ \eta_{xx}(0, t) = \theta_2(t), \ \eta_{xx}(L, t) = 0, &t\in (0, T).\end{array}\right. \end{eqnarray} $

引理 3.3   设$ f_1, f_2\in L^2(Q_T) $, $ \theta_1, \theta_2\in L^2(0, T) $, $ M_1, M_2\in C^{1}(\overline\Omega) $. 则存在正常数$ \lambda_2 = \lambda_2(x_0, \Omega, T, \|M_1\|_{C^1(\overline\Omega)}, \|M_2\|_{C^1(\overline\Omega)}, k_1, k_2) $, $ s_2 = s_2(\lambda, x_0, \Omega, T, \|M_1\|_{C^1(\overline\Omega)} $, $ \|M_2\|_{C^1(\overline\Omega)}, k_1, k_2) $$ C = C(\lambda, x_0, \Omega, T, \|M_1\|_{C^1(\overline\Omega)}, \|M_2\|_{C^1(\overline\Omega)}, $$ k_1, k_2) $, 使得

$ \begin{eqnarray} &&{{\cal I}}(\xi)+{{\cal I}}(\eta)+{\cal B} (\xi)|_{x = 0}+{\cal B} (\eta)|_{x = 0}\\ &\leq &C\int_{\Omega_T}s\lambda\phi(|f_1|^2+|f_2|^2)e^{2s\varphi}{\rm d}x{\rm d}t+C\int_0^T(s\lambda\phi(|\theta_1|^2+|\theta_2|^2)e^{2s\varphi})|_{x = 0}{\rm d}t\\ &&+C\int_{{\omega^{(3)}_T}}s^9\lambda^6\phi^9(|\xi|^2+|\eta|^2)e^{2s\varphi}{\rm d}x{\rm d}t \end{eqnarray} $

对于所有的$ \lambda\geq \lambda_2 $, $ s\geq s_2 $和所有满足(3.38) 式的$ \xi, \eta\in L^2(0, T;H^2(\Omega)) $成立.

  对$ (\xi, \eta) $应用引理3.2, 可得对所有的$ \lambda\geq \lambda_1 $, $ s\geq s_1 $, 有

$ \begin{eqnarray} {{\cal I}}(\xi)+{{\cal B}}(\xi)|_{x = 0} &\leq& C\int_{\Omega_T}s\lambda\phi\left(|\xi_{x}|^2+|\eta_{x}|^2+|f_1|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t\\ &&+C\int_0^T(s\lambda\phi|\theta_1|^2e^{2s\varphi})|_{x = 0}{\rm d}t+C\int_{{\omega^{(3)}_T}}s^9\lambda^6\phi^9|\xi|^2 e^{2s\varphi}{\rm d}x{\rm d}t \end{eqnarray} $

$ \begin{eqnarray} {{\cal I}}(\eta)+{{\cal B}}(\eta)|_{x = 0} &\leq& C\int_{\Omega_T}s\lambda\phi\left(|\eta_{x}|^2+|\xi_{x}|^2+|f_2|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t\\ &&+C\int_0^T(s\lambda\phi|\theta_2|^2e^{2s\varphi})|_{x = 0}{\rm d}t+C\int_{{\omega^{(3)}_T}}s^9\lambda^6\phi^9|\eta|^2 e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

然后将(3.40) 式和(3.41) 式相加, 并选择$ s $充分大来吸收关于$ \xi_{x}, \eta_{x} $的全局积分项, 我们得到(3.39) 式. 证毕.

3.3 定理3.1的证明

为证明定理3.1, 我们需要下面的引理.

引理 3.4   存在正常数$ \lambda_3 = \lambda_3(x_0, \Omega, T) $, $ s_3 = s_3(\lambda, x_0, \Omega, T) $$ C = C(x_0, \Omega, T) $, 使得

$ \begin{eqnarray} {{\cal I}}(z)\leq C\int_{\Omega_T}|{\cal L}[z]|^2e^{2s\varphi}{\rm d}x{\rm d}t+Cs^{-1}{\cal I} (z_x) \end{eqnarray} $

对于所有的$ \lambda\geq \lambda_3 $, $ s\geq s_3 $和所有满足

$ \begin{eqnarray} \left\{ \begin{array}{ll}{\cal L}[z]: = z_t+ z_{xxx}, &(x, t)\in\Omega_T, \\ z(0, t) = z(L, t) = z_x(L) = 0, & t\in (0, T) \end{array} \right. \end{eqnarray} $

$ z\in L^2(0, T;H^3(\Omega)) $都成立.

  为简单记, 这里我们沿用引理3.2中的记号. 令$ Z = ze^{s\varphi} $. 因为估计式(3.19) 不涉及边界条件, 所以(3.19) 式对于$ Z $仍然成立, 即

$ \begin{eqnarray} L(Z)+N(Z) & \geq & \int_{{\Omega_T}}s\lambda^2\phi|z_{xx}|^2e^{2s\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T}s^3\lambda^4\phi^3|z_x|^2e^{2s\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T}s^5\lambda^6\phi^5|z|^2e^{2s\varphi}{\rm d}x{\rm d}t\\ & &-C\sum\limits_{x = \hat x}\int_0^T(s^2\lambda^4\phi^3|z_x|^2e^{2s\varphi}+s^4\lambda^6\phi^5|z|^2e^{2s\varphi})|_{\hat x = 0, L}{\rm d}t. \end{eqnarray} $

对于$ B(Z)-I_{11}^3 $, 应用$ z(0, t) = z(L, t) = z_x(L, t) = 0 $, 我们有

$ \begin{eqnarray} B(Z)-I_{11}^3&\geq &\int_0^T\left(s|\varphi_x||Z_{xx}|^2+s^3|\varphi_x|^3|Z_x|^2+s^5|\varphi_x|^5|Z|^2\right)|_{x = 0}{\rm d}t\\ &&-C\int_0^T\left(s|\varphi_x||Z_{xx}|^2+s^3|\varphi_x|^3|Z_x|^2+s^5|\varphi_x|^5|Z|^2\right)|_{x = L}{\rm d}t\\ &\geq &\int_0^T\left(s^3\lambda^3\phi^3|z_x|^2e^{2s\varphi}\right)|_{x = 0}{\rm d}t-C\int_0^T \left(s\lambda\phi|z_{xx}|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t. \end{eqnarray} $

另外, 由$ Z(0, t) = Z_x(L, t) = 0 $可得$ I_{11}^3 = 0 $. 因此, 我们得到以下关于$ B(Z) $的估计

$ \begin{eqnarray} B(Z)\geq \int_0^T\left(s^3\lambda^3\phi^3|z_x|^2e^{2s\varphi}\right)|_{x = 0}{\rm d}t-C\int_0^T \left(s\lambda\phi|z_{xx}|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t. \end{eqnarray} $

然后, 将(3.44) 和(3.46) 式代入$ ({\cal L}_1[Z], {\cal L}_2[Z]) = L(Z)+N(Z)+B(Z) $, 可得

$ \begin{eqnarray} ({\cal L}_1[Z], {\cal L}_2[Z]) &\geq &\int_{{\Omega_T}}s\lambda^2\phi|z_{xx}|^2e^{2s\varphi}{\rm d}x{\rm d}t +\int_{\Omega_T}s^3\lambda^4\phi^3|z_x|^2e^{2s\varphi}{\rm d}x{\rm d}t \\ &&+\int_{\Omega_T}s^5\lambda^6\phi^5|z|^2e^{2s\varphi}{\rm d}x{\rm d}t-C\int_{0}^T\left(s\lambda\phi|z_{xx}|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t. \end{eqnarray} $

由(3.47) 式及

$ \begin{eqnarray} \|{\mathcal L}_1[Z]\|^2_{L^2(\Omega_T)}+\|{\mathcal L}_2[Z]\|^2_{L^2(\Omega_T)}+2({\cal L}_1[Z], {\cal L}_2[Z]) = \|e^{s\varphi}{\cal L}[z]+{\mathcal R}[Z]\|_{L^2(\Omega_T)}^2, \end{eqnarray} $

我们得到

$ \begin{equation} {{\cal I}}(z)\leq C\int_{\Omega_T}|{\cal L}[z]|^2e^{2s\varphi}{\rm d}x{\rm d}t+C\int_{0}^T\left(s\lambda\phi|z_{xx}|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t. \end{equation} $

另一方面, 应用分部积分和Young不等式, 我们有

$ \begin{eqnarray} \int_{0}^T\left(s\lambda\rho_1\phi|z_{xx}|^2e^{2s\varphi}\right)|_{x = L}{\rm d}t & = &2\int_{\Omega_T}s\lambda\rho_1\phi z_{xx}z_{xxx}e^{2s\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T}s\lambda\left(\rho_1\phi e^{2s\varphi}\right)_x |z_{xx}|^2{\rm d}x{\rm d}t\\ &\leq& Cs^{-1}{\cal I} (z_x). \end{eqnarray} $

最后, 通过(3.49) 和(3.50) 式, 我们立即可到(3.42) 式. 证毕.

现在我们证明定理3.1.

定理 3.1的证明   应用(3.1), (3.2) 和$ f(L, t) = 0 $, 可知$ (p_x, q_x) $满足

$ \begin{eqnarray} \left\{\begin{array}{ll} p_{x, t}+ p_{x, xxx}+(M_1(x)p)_{xx}+k_1q_{xx} = f_{1, x}, &(x, t)\in \Omega_T, \\ q_{x, t}+ q_{x, xxx}+(M_2(x)q)_{xx}+k_2p_{xx} = f_{2, x}, &(x, t)\in \Omega_T, \\ p_{x}(L, t) = 0, \ p_{x, xx}(0, t) = \theta_1(t), \ p_{x, xx}(L, t) = 0, &t\in (0, T), \\ q_{x}(L, t) = 0, \ q_{x, xx}(0, t) = \theta_2(t), \ q_{x, xx}(L, t) = 0, &t\in (0, T), \end{array}\right. \end{eqnarray} $

其中

然后对$ (p_x, q_x) $应用引理3.3, 可得对于所有$ \lambda\geq\lambda_2 $$ s\geq s_2 $, 成立

$ \begin{eqnarray} &&{{\cal I}}(p_x)+{{\cal I}}(q_x)+{\cal B} (p_x)|_{x = 0}+{\cal B} (q_x)|_{x = 0}\\ &\leq &C\int_{\Omega_T}s\lambda\phi(|f_{1, x}|^2+|f_{2, x}|^2)e^{2s\varphi}{\rm d}x{\rm d}t+C\int_0^T(s\lambda\phi(|p_x|^2+|q_x|^2+|f_1|^2+|f_2|^2)e^{2s\varphi})|_{x = 0}{\rm d}t\\ &&+C\int_{{\omega^{(3)}_T}}s^9\lambda^6\phi^9(|p_x|^2+|q_x|^2)e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

另一方面, 由引理3.4, 可得对于所有$ \lambda\geq \lambda_3 $$ s\geq s_3 $, 成立

$ \begin{equation} {{\cal I}}(p)+{{\cal I}}(q) \leq C\int_{\Omega_T}(|p|^2+|p_x|^2+|q|^2+|q_x|^2+|f_1|^2+|f_2|^2)e^{2s\varphi}{\rm d}x{\rm d}t+Cs^{-1}({\cal I} (p_x)+{\cal I} (q_x)). \end{equation} $

此外, 注意到$ \phi|_{x = 0}\leq \phi $, $ \varphi|_{x = 0}\leq \varphi $, 我们有

$ \begin{eqnarray} \int_0^T(s\lambda\phi(|f_1|^2+|f_2|^2)e^{2s\varphi})|_{x = 0}{\rm d}t & = &\int_0^T(s\lambda\phi e^{2s\varphi})|_{x = 0}{\rm d}t\left(\left|\int_\Omega f_{1, x} dx\right|^2+\left|\int_\Omega f_{2, x} dx\right|^2\right)\\ &\leq& \int_0^T(s\lambda\phi e^{2s\varphi})|_{x = 0}\int_{\Omega}(|f_{1, x}|^2+|f_{2, x}|^2){\rm d}x{\rm d}t\\ &\leq &C\int_{\Omega_T}s\lambda\phi(|f_{1, x}|^2+|f_{2, x}|^2)e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

因此, 从(3.52)–(3.54)式可得

$ \begin{eqnarray} &&{{\cal I}}(p_x)+{{\cal I}}(q_x)+s{\cal I} (p)+s{\cal I} (q)+{\cal B} (p_x)|_{x = 0}+{\cal B} (q_x)|_{x = 0}\\ &\leq &C\int_{\Omega_T}s\lambda\phi(|f_1|^2+|f_{1, x}|^2+|f_2|^2+|f_{2, x}|^2)e^{2s\varphi}{\rm d}x{\rm d}t\\ &&+C\int_{{\omega^{(3)}_T}}s^9\lambda^6\phi^9(|p_x|^2+|q_x|^2)e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

现在我们消除(3.55) 式右端关于$ p_x $的局部积分项. 显然, 在$ (L-\epsilon_0, L)\times (0, T) $$ f_2 = 0 $.$ \omega^{(4)}_T\subset (L-\epsilon_0, L)\times (0, T) $可得$ \rho_3 f_2 = 0 $. 然后, 用$ s^9\lambda^6\rho_3\phi^9 p_xe^{2s\varphi} $乘以$ q $并应用分部积分, 我们得到

$ \begin{eqnarray} &&\int_{{\Omega}_T}k_2s^9\lambda^6\rho_3\phi^9|p_x|^2e^{2s\varphi}{\rm d}x{\rm d}t\\ & = &\int_{\Omega_T} s^9\lambda^6\rho_3\phi^9 p_x\left(-q_t-q_{xxx}-(M_2(x)q)_x+f_2\right)e^{2s\varphi}{\rm d}x{\rm d}t{}\\ & = &\int_{\Omega_T} s^9\lambda^6\rho_3\phi^9 p q_{xt}e^{2s\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T} s^9\lambda^6 p q_{t}\left(\rho_3\phi^9e^{2s\varphi}\right)_x{\rm d}x{\rm d}t\\ &&+\int_{\Omega_T}s^9\lambda^6\rho_3\phi^9p_{xx}q_{xx}e^{2s\varphi}{\rm d}x{\rm d}t+\int_{\Omega_T}s^9\lambda^6 p_{x}q_{xx}\left(\rho_3\phi^9e^{2s\varphi}\right)_x{\rm d}x{\rm d}t\\ & &-\int_{\Omega_T}s^9\lambda^6\rho_3\phi^9 p_x(M_2(x)q)_x e^{2s\varphi}{\rm d}x{\rm d}t{}\\ & = &-\int_{\Omega_T}s^9\lambda^6\rho_3\phi^9(p_t+p_{xxx})q_xe^{2s\varphi}{\rm d}x{\rm d}t-\int_{\Omega_T}s^9\lambda^6 pq_x\left(\rho_3\phi^9e^{2s\varphi}\right)_t{\rm d}x{\rm d}t\\ &&-2\int_{\Omega_T}s^9\lambda^6 p_{xx}q_x\left(\rho_3\phi^9e^{2s\varphi}\right)_x{\rm d}x{\rm d}t-\int_{\Omega_T}s^9\lambda^6 p_xq_x\left(\rho_3\phi^9e^{2s\varphi}\right)_{xx}{\rm d}x{\rm d}t {}\\ &&-\int_{\Omega_T} s^9\lambda^6 p_t q\left(\rho_3\phi^9e^{2s\varphi}\right)_x{\rm d}x{\rm d}t-\int_{\Omega_T}s^9\lambda^6 pq\left(\rho_3\phi^9e^{2s\varphi}\right)_{xt}{\rm d}x{\rm d}t\\ &&-\int_{\Omega_T}s^9\lambda^6 p_x(M_2(x)q)_x \rho_3\phi^9 e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

结合

和Young不等式, 我们进一步得到对于所有的$ \epsilon>0 $, 有

$ \begin{eqnarray} \int_{\omega^{(3)}_T}s^9\lambda^6\phi^9|p_x|^2e^{2s\varphi}{\rm d}x{\rm d}t&\leq &\int_{{\Omega}_T}s^9\lambda^6\rho_3\phi^9|p_x|^2e^{2s\varphi}{\rm d}x{\rm d}t\\ & \leq & \epsilon{{\cal I}}(p_x)+{\cal I} (p)+\epsilon\int_{\Omega_T} s\phi|p_t|^2e^{2s\varphi}{\rm d}x{\rm d}t\\ & &+ C(\epsilon, \lambda)\int_{\omega^{(4)}_T}s^{19}\phi^{21}\left(|q_x|^2+|q|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

另外, 由$ p $的方程, 我们有

$ \begin{equation} \int_{\Omega_T}s\phi|p_t|^2e^{2s\varphi}{\rm d}x{\rm d}t\leq {\cal I} (p_x)+{\cal I}(q_x)+C\int_{\Omega_T} s\phi |f|^2e^{2s\varphi}{\rm d}x{\rm d}t. \end{equation} $

将(3.58) 式代入(3.57) 式得到

$ \begin{eqnarray} \int_{\omega^{(3)}_T}s^9\lambda^6\phi^9|p_x|^2e^{2s\varphi}{\rm d}x{\rm d}t &\leq & 2\epsilon{{\cal I}}(p_x)+\epsilon{\cal I} (q_x)+{\cal I} (p)+C\int_{\Omega_T} s\phi |f|^2e^{2s\varphi}{\rm d}x{\rm d}t\\ &&+ C(\epsilon, \lambda)\int_{\omega^{(4)}_T}s^{19}\phi^{21}\left(|q_x|^2+|q|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

因此, 从(3.55) 和(3.59) 式我们推知

$ \begin{eqnarray} &&{{\cal I}}(p_x)+{{\cal I}}(q_x)+s{\cal I} (p)+s{\cal I} (q)+{\cal B} (p_x)|_{x = 0}+{\cal B} (q_x)|_{x = 0}\\ &\leq &C\int_{\Omega_T}s\lambda\phi(|f_1|^2+|f_{1, x}|^2+|f_2|^2+|f_{2, x}|^2)e^{2s\varphi}{\rm d}x{\rm d}t\\ &&+C(\lambda)\int_{{\omega^{(4)}_T}}s^{19}\phi^{21}(|q|^2+|q_x|^2)e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

类似于(3.34)式, 对于$ q_x $的局部积分项, 我们有下面的估计

$ \begin{equation} \int_{\omega^{(4)}_T}s^{19}\phi^{21} |q_x|^2e^{2s\varphi}{\rm d}x{\rm d}t \leq \epsilon {\cal I} (q_x)+C(\epsilon, \lambda) \int_{\omega_T}s^{35}\phi^{39} |q|^2e^{2s\varphi}{\rm d}x{\rm d}t. \end{equation} $

最后, 将(3.61) 式代入(3.60) 式并取$ \epsilon $充分小, 我们得到(3.5) 式. 定理3.1证毕.

4 定理1.1的证明

本节将证明我们反问题的Lipschitz稳定性, 即定理1.1. 证明的过程基于文献[19] 的方法, 这种方法最初由Bukhgeim和Klibanov在文献[5]中提出, 也可参见文献[1, 32]. 以下我们固定$ \lambda $是满足定理3.1中的约束的正常数, 并用$ C $表示依赖于$ x_0, \Omega, T, \|M_1\|_{C^2(\overline\Omega)}, \|M_2\|_{C^2(\overline\Omega)} $, $ \epsilon_1 $$ \lambda $, 但不依赖于$ s $的正常数.

我们引入一个引理, 其是关于一阶偏微分算子

$ \begin{eqnarray} {\cal N} [\zeta] = A(x)\zeta_x(x)+B(x)\zeta(x), \quad x\in\Omega \end{eqnarray} $

的卡勒曼估计, 其中$ A, B\in W^{1, \infty}(\Omega) $.

引理 4.1   设

$ \begin{eqnarray} |A(x)(x-x_0)|>0, \quad x\in\Omega, \end{eqnarray} $

则存在正常数$ \lambda_4 = \lambda_4(x_0, \Omega, \|A\|_{W^{1, \infty}(\Omega)}, $$ \|B\|_{W^{1, \infty}(\Omega)}) $, $ s_4 = s_4(\lambda, x_0 $, $ \Omega $, $ \|A\|_{W^{1, \infty}(\Omega)} $, $ \|B\|_{W^{1, \infty}(\Omega)}) $$ C = C(\lambda, x_0, \Omega, \|A\|_{W^{1, \infty}(\Omega)}, $$ \|B\|_{W^{1, \infty}(\Omega)}) $, 使得

$ \begin{eqnarray} \int_\Omega s^2 |\zeta|^2e^{2s\varphi(x, t_0)}{\rm d}x\leq C \int_\Omega |{\cal N} [\zeta](x)|^2e^{2s\varphi(x, t_0)}{\rm d}x \end{eqnarray} $

$ \begin{eqnarray} \int_\Omega s^2 |\zeta_x|^2e^{2s\varphi(x, t_0)}{\rm d}x\leq C \int_\Omega\left( |{\cal N} [\zeta]|^2+|({\cal N} [\zeta])_x|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x \end{eqnarray} $

对于所有的$ \lambda\geq \lambda_4 $, $ s\geq s_4 $和所有满足$ \zeta(0) = \zeta(L) = 0, \zeta_x(0) = \zeta_x(L) = 0 $$ \zeta\in H^2(\Omega) $都成立.

此引理的证明可以参见文献[2, 18].

定理 1.1的证明   为简单记, 令

那么$ (\overline u, \overline v) $满足

$ \begin{eqnarray} \left\{\begin{array}{ll} \overline u_{t}+ \overline u_{xxx}+ \left(M_1(x)\overline u\right)_{x}+k_1\overline v_{x} = -({\overline M}_1(x)\tilde u)_x, &(x, t)\in \Omega_T, \\ \overline v_{t}+ \overline v_{xxx}+ \left(M_2(x)\overline v\right)_{x}+k_2\overline u_{x} = -({\overline M}_2(x)\tilde v)_x, &(x, t)\in \Omega_T, \\ \overline u (0, t) = \overline u(L, t) = \overline u_{x}(L, t) = 0, &t\in (0, T), \\ \overline v(0, t) = \overline v(L, t) = \overline v_{x}(L, t) = 0, &t\in (0, T). \end{array}\right. \end{eqnarray} $

现在我们考虑$ (p_i, q_i): = (\partial_t^i\overline u, \partial_t^i\overline v) $$ (i = 0, 1, 2) $满足的方程组

$ \begin{eqnarray} \left\{\begin{array}{ll} p_{i, t}+ p_{i, xxx}+ \left(M_1(x)p_i\right)_{x}+k_1q_{i, x} = -({\overline M}_1(x)\partial_t^i\tilde u)_x, &(x, t)\in \Omega_T, \\ q_{i, t}+ q_{i, xxx}+ \left(M_2(x)q_i\right)_{x}+k_2p_{i, x} = -({\overline M}_2(x)\partial_t^i\tilde v)_x, &(x, t)\in \Omega_T, \\ p_i(0, t) = p_{i}(L, t) = p_{i, x}(L, t) = 0, &t\in (0, T), \\ {q_i}(0, t) = q_{i}(L, t) = q_{i, x}(L, t) = 0, &t\in (0, T). \end{array}\right. \end{eqnarray} $

注意到$ \overline M_i(L) = \overline M_{i, x}(L) = 0 $$ (i = 1, 2) $$ \overline M_{2}(x) = 0 $, $ x\in(L-\epsilon_0, L) $. 然后应用定理3.1, 我们有

$ \begin{eqnarray} & &{\cal I} (\overline u_{xt})+{\cal I} ( {\overline u}_{xtt})+s{\cal I} ( {\overline u}_{t})+s{\cal I} ( {\overline u}_{tt})+{\cal I} (\overline v_x)+{\cal I} ( {\overline v}_{xt})\\ &\leq & C\sum\limits_{i = 1}^2\int_{\Omega_T}s\left(|{\overline M}_{i, xx}(x)|^2+ |{\overline M}_{i, x}(x)|^2+|{\overline M}_i(x)|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t {}\\ &&+Cs^{35}e^{Cs}\|\overline v\|_{H^2(0, T;L^2(\omega_T))}^2. \end{eqnarray} $

另一方面, 由(4.5) 式的第一个方程得

$ \begin{eqnarray} -{\overline M}_{1, x}(x){\tilde u}(x, t_0)-{\overline M}_{1}(x){\tilde u}_x(x, t_0) = \zeta(x), \quad x\in\Omega, \end{eqnarray} $

这里

利用$ {\overline M}_{1}(x) = 0 $$ x = 0, L $$ {\overline M}_{1, x}(x) = 0 $

我们可以应用引理4.1得到

$ \begin{eqnarray} &&\int_{\Omega}s^2\left(|{\overline M}_{1, x}(x)|^2+|{\overline M}_{1}(x)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x\\ &\leq& C\int_{\Omega}\left(|{\overline u}_{xt}(x, t_0)|^2+|{\overline u}_{t}(x, t_0)|^2+|\overline v_{xx}(x_0, t)|^2+|\overline v_x(x_0, t)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x\\ &&+Ce^{Cs}\|{\overline u}(\cdot, t_0)\|^2_{H^4(\Omega)}. \end{eqnarray} $

另外, 由

$ \begin{eqnarray} -{\overline M}_{1, xx}(x){\tilde u}(x, t_0) = \zeta_x(x)+2{\overline M}_{1, x}(x){\tilde u}_x(x, t_0)+{\overline M}_1(x){\tilde u}_{xx}(x, t_0), \quad x\in\Omega \end{eqnarray} $

可得

$ \begin{eqnarray} &&\int_{\Omega}|{\overline M}_{1, xx}(x)|^2e^{2s\varphi(x, t_0)}{\rm d}x\\ &\leq& C\int_{\Omega}\left(|{\overline M}_{1, x}(x)|^2+|{\overline M}_{1}(x)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x+C\int_{\Omega}\left(|{\overline u}_{xt}(x, t_0)|^2+|{\overline v}_{xx}(x, t_0)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x\\ &&+Ce^{Cs}\|{\overline u}(\cdot, t_0)\|^2_{H^4(\Omega)}. \end{eqnarray} $

从而, 由(4.9) 和(4.11) 式, 我们有

$ \begin{eqnarray} &&\int_{\Omega}\left(s|{\overline M}_{1, xx}(x)|^2+s^2|{\overline M}_{1, x}(x)|^2+s^2|{\overline M}_{1}(x)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x\\ &\leq& C\int_{\Omega}\left(s\left(|{\overline u}_{xt}(x, t_0)|^2+|\overline v_{xx}(x_0, t)|^2\right)+\left(|{\overline u}_{t}(x, t_0)|^2+|\overline v_{x}(x_0, t)|^2\right)\right)e^{2s\varphi(x, t_0)}{\rm d}x\\ &&+Cse^{Cs}\|{\overline u}(\cdot, t_0)\|^2_{H^4(\Omega)}. \end{eqnarray} $

另外

$ \begin{eqnarray} \int_{\Omega}|{\overline u}_{xt}(x, t_0)|^2e^{2s\varphi(x, t_0)}{\rm d}x& = &\int_0^{t_0}\frac{\partial}{\partial t}\left(\int_{\Omega}|{\overline u}_{xt}(x, t)|^2e^{2s\varphi(x, t)}{\rm d}x\right){\rm d}t\\ &\leq & \int_{\Omega_T}2\left(|{\overline u}_{xt}||{\overline u}_{xtt}|+s|\varphi_t|| {\overline u}_{xt}|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t\\ &\leq & C\int_{\Omega_T}\left(s|{\overline u}_{xt}|^2+|{\overline u}_{xtt}|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t. \end{eqnarray} $

类似于(4.13) 式的计算, 我们可以得到关于$ \overline v_{xx}, \overline u_t $$ \overline v_x $的同样的估计. 然后由(4.7), (4.12) 和(4.13) 式, 我们得到

$ \begin{eqnarray} & &\int_{\Omega}s\left(|{\overline M}_{1, xx}(x)|^2+|{\overline M}_{1, x}(x)|^2+|{\overline M}_{1}(x)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x\\ &\leq&{\cal I} (\overline u_{xt})+{\cal I} ( {\overline u}_{xtt})+{\cal I} ( {\overline u}_{t})+{\cal I} ( {\overline u}_{tt})+{\cal I} (\overline v_x)+{\cal I} ( {\overline v}_{xt})+Cse^{Cs}\|{\overline u}(\cdot, t_0)\|^2_{H^4(\Omega)}\\ & \leq &C\sum\limits_{i = 1}^2\int_{\Omega_T}s\left(|{\overline M}_{i, xx}(x)|^2+|{\overline M}_{i, x}(x)|^2+|{\overline M}_i(x)|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t\\ & &+Cs^{35}e^{Cs}\|\overline v\|_{H^2(0, T;L^2(\omega_T))}^2+Cse^{Cs}\|{\overline u}(\cdot, t_0)\|^2_{H^4(\Omega)}. \end{eqnarray} $

类似地, 我们有

$ \begin{eqnarray} &&\int_{\Omega}s\left(|{\overline M}_{2, xx}(x)|^2+|{\overline M}_{2, x}(x)|^2+|{\overline M}_{2}(x)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x\\ &\leq &C\sum\limits_{i = 1}^2\int_{\Omega_T}s\left(|{\overline M}_{i, xx}(x)|^2+|{\overline M}_{i, x}(x)|^2+|{\overline M}_i(x)|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t\\ &&+Cs^{35}e^{Cs}\|\overline v\|_{H^2(0, T;L^2(\omega_T))}^2+Cse^{Cs}\|{\overline v}(\cdot, t_0)\|^2_{H^4(\Omega)}. \end{eqnarray} $

然后将(4.14) 和(4.15) 式相加得到

$ \begin{eqnarray} &&\sum\limits_{i = 1}^2\int_{\Omega}s\left(|{\overline M}_{i, xx}(x)|^2+|{\overline M}_{i, x}(x)|^2+|{\overline M}_{i}(x)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x\\ &\leq &C\sum\limits_{i = 1}^2\int_{\Omega_T}s\left(|{\overline M}_{i, xx}(x)|^2+|{\overline M}_{i, x}(x)|^2+|{\overline M}_i(x)|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t\\ & &+Cs^{35}e^{Cs}\|\overline v\|_{H^2(0, T;L^2(\omega_T))}^2+Cse^{Cs}\|{\overline u}(\cdot, t_0)\|^2_{H^4(\Omega)}+Cse^{Cs}\|{\overline v}(\cdot, t_0)\|^2_{H^4(\Omega)}. \end{eqnarray} $

$ D: = \frac{4}{T^2} \min\limits_{(x, t)\in\Omega_T}\frac{e^{2\lambda\|d\|_{C(\overline \Omega)}}-e^{\lambda d(x)}}{t(T-t)}>0 $.$ s\rightarrow \infty $时, 我们有

$ \begin{eqnarray} \sup\limits_{x\in\Omega}\left|\int_0^T e^{2s(\varphi(x, t)-\varphi(x, t_0))}{\rm d}t\right|\leq \int_0^T e^{-2sD\left(t-\frac{T}{2}\right)^2}{\rm d}t\leq \int_{-\infty}^{+\infty} e^{-2sD t^2}{\rm d}t = o(1). \end{eqnarray} $

然后对于充分小的$ \epsilon>0 $, 如果我们选择$ s $充分大, 可得

$ \begin{eqnarray} &&\sum\limits_{i = 1}^2\int_{\Omega_T}\left(|{\overline M}_{i, xx}(x)|^2+|{\overline M}_{i, x}(x)|^2+|{\overline M}_i(x)|^2\right)e^{2s\varphi}{\rm d}x{\rm d}t\\ & = &\sum\limits_{i = 1}^2\int_{\Omega}\left(|{\overline M}_{i, xx}(x)|^2+|{\overline M}_{i, x}(x)|^2+|{\overline M}_i(x)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x\int_0^T e^{2s(\varphi(x, t)-\varphi(x, t_0))}{\rm d}t\\ &\leq &\epsilon \sum\limits_{i = 1}^2\int_{\Omega}\left(|{\overline M}_{i, xx}(x)|^2+|{\overline M}_{i, x}(x)|^2+|{\overline M}_i(x)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x. \end{eqnarray} $

由(4.16) 和(4.18) 式, 我们得到

$ \begin{eqnarray} &&(s- \epsilon C s)\sum\limits_{i = 1}^2\int_{\Omega_T}\left(|{\overline M}_{i, xx}(x)|^2+|{\overline M}_{i, x}(x)|^2+|{\overline M}_i(x)|^2\right)e^{2s\varphi(x, t_0)}{\rm d}x{\rm d}t\\ &\leq& Cs^{35}e^{Cs}\|\overline v\|_{H^2(0, T;L^2(\omega_T))}^2+Cse^{Cs}\|{\overline u}(\cdot, t_0)\|^2_{H^4(\Omega)}+Cse^{Cs}\|{\overline v}(\cdot, t_0)\|^2_{H^4(\Omega)}. \end{eqnarray} $

如果我们选择$ \epsilon $充分小, 由(4.19) 式立刻可得(1.6) 式. 定理1.1证毕.

参考文献

Baudouin L , Cerpa E , Crépeau E , Mercado A .

On the determination of the principal coefficient from boundary measurements in a KdV equation

J Inverse Ill-Posed Prob, 2014, 22, 819- 845

URL     [本文引用: 4]

Bellassoued M , Yamamoto M .

Lipschitz stability in determining density and two Lamé coefficients

J Math Anal Appl, 2017, 329, 1240- 1259

URL     [本文引用: 2]

Bellassoued M , Yamamoto M .

Carleman estimates and an inverse heat source problem for the thermoelasticity system

Inverse Probl, 2011, 27, 015006

DOI:10.1088/0266-5611/27/1/015006      [本文引用: 1]

Bellassoued M , Yamamoto M .

Carleman estimate and inverse source problem for Biot's equations describing wave propagation in porous media

Inverse Probl, 2013, 29, 115002

DOI:10.1088/0266-5611/29/11/115002     

Bukhgeim A L , Klibanov M V .

Uniqueness in the large of a class of multidimensional inverse problems

Soviet Math Dokl, 1981, 17, 244- 247

URL     [本文引用: 2]

Caicedo M A , Capistrano-Filho R , Zhang B Y .

Neumann boundary controllability of the Korteweg-de Vries equation on a bounded domain

SIAM J Control Optim, 2017, 55, 3503- 3532

DOI:10.1137/15M103755X      [本文引用: 1]

Capistrano-Filho R , Pazoto A F , Rosier L .

Internal controllability of the Korteweg-de Vries equation on a bounded domain

ESAIM Control Optim Calc Var, 2015, 21, 1076- 1107

DOI:10.1051/cocv/2014059      [本文引用: 4]

Carréno N , Guerrero S .

Uniform null controllability of a linear KdV equation using two controls

J Math Anal Appl, 2018, 457, 922- 943

DOI:10.1016/j.jmaa.2017.08.039      [本文引用: 1]

Cerpa E , Rivas I , Zhang B Y .

Boundary controllability of the Korteweg-de Vries equation on a bounded domain

SIAM J Control Optim, 2013, 51, 2976- 3010

DOI:10.1137/120891721      [本文引用: 1]

Chen M .

Lipschitz stability in an inverse problem for the Korteweg-de Vries equation on a finite domain

Boundary Value Problems, 2007, 2017, 48

DOI:10.1186/s13661-017-0779-8      [本文引用: 1]

Coron J M , Crépeau E .

Exact boundary controllability of a nonlinear KdV equation with a critical length

J Eur Math Soc, 2004, 31, 367- 398

[本文引用: 1]

Dias F , Il'ichev A .

Interfacial waves with free-surface boundary conditions: an approach via a model equation

Physica D, 2001, 150, 278- 300

DOI:10.1016/S0167-2789(01)00149-X      [本文引用: 1]

Fochesato C , Dias F , Grimshaw R .

Generalized solitary waves and fronts in coupled Korteweg-de Vries systems

Physica D, 2005, 210, 96- 117

DOI:10.1016/j.physd.2005.07.010      [本文引用: 1]

Gao P .

Carleman estimate and unique continuation property for the linear stochastic Korteweg-de Vries equation

Bull Aust Math Soc, 2014, 90, 283- 294

DOI:10.1017/S0004972714000276      [本文引用: 1]

Glass O , Guerrero S .

Some exact controllability results for the linear KdV equation and uniform controllability in the zero-dispersion limit

Asymptot Anal, 2008, 60, 61- 100

DOI:10.3233/ASY-2008-0900      [本文引用: 2]

Grimshaw R , Pelinovsky E , Talipova T .

The modified Korteweg-de Vries equation in the theory of the large-amplitude internal waves

Nonlinear Process Geophys, 1997, 4, 237- 250

DOI:10.5194/npg-4-237-1997      [本文引用: 1]

Imanuvilov O Y .

Controllability of parabolic equations

Sbornik Math, 1995, 186, 879- 900

DOI:10.1070/SM1995v186n06ABEH000047      [本文引用: 1]

Imanuvilov O Y , Yamamoto M .

Carleman estimates for the non-stationary Lemé system and application to an inverse problem

ESAIM Control Optim Calc Var, 2005, 11, 1- 56

DOI:10.1051/cocv:2004030      [本文引用: 1]

Imanuvilov O Y , Yamamoto M .

Global Lipschitz stability in an inverse hyperbolic problem by interior observations

Inverse Probl, 2001, 17, 717- 728

DOI:10.1088/0266-5611/17/4/310      [本文引用: 1]

Isakov V . Inverse Problems for Partial Differential Equations. Berlin: Springer-Verlag, 1998

[本文引用: 1]

Isakov V , Kim N .

Carleman estimates with second large parameter and applications to elasticity with residual stress

Applicationes Mathematicae, 2008, 35, 447- 465

DOI:10.4064/am35-4-4      [本文引用: 1]

Klibanov M V , Timonov A . Carleman Estimates for Coefficient Inverse Problems and Numerical Applications. Utrecht: VSP, 2004

Klibanov M V .

Carleman estimates for global uniqueness, stability and numerical methods for coefficient inverse problems

J Inverse Ill-Posed Prob, 2013, 21, 477- 560

URL     [本文引用: 1]

Klibanov M V , Kolesov A E .

Convexification of a 3-D coefficient inverse scattering problem

Computers and Mathematics with Applications, 2019, 77, 1681- 1702

DOI:10.1016/j.camwa.2018.03.016      [本文引用: 1]

Klibanov M V , Kolesov A E , Nguyen L , Sullivan A .

Globally strictly convex cost functional for a 1D inverse medium scattering problem with experimental data

SIAM J Applied Mathematics, 2017, 77, 1733- 1755

DOI:10.1137/17M1122487      [本文引用: 1]

Kumarasamy S , Hasanov A .

An inverse problem for the KdV equation with Neumann boundary measured data

J Inverse Ill-Posed Prob, 2018, 26, 133- 151

DOI:10.1515/jiip-2017-0038      [本文引用: 1]

Romanov V G , Yamamoto M .

Recovering a Lamé kernel in viscoelastic equation by a single boundary measurement

Appl Anal, 2010, 89, 377- 390

DOI:10.1080/00036810903518975      [本文引用: 1]

Rosier L .

Exact boundary controllability for the linear Korteweg-de Vries equation on the half-line

SIAM J Control Optim, 2000, 39, 331- 351

DOI:10.1137/S0363012999353229      [本文引用: 1]

Rousseau J , Lebeau G .

On Carleman estimates for elliptic and parabolic operators Applications to unique continuation and control of parabolic equations

ESAIM Control Optim Calc Var, 2012, 18, 712- 747

DOI:10.1051/cocv/2011168      [本文引用: 1]

Saut J , Scheurer B .

Unique continuation for some evolution equations

J Differential Equations, 1987, 66, 118- 139

DOI:10.1016/0022-0396(87)90043-X      [本文引用: 1]

Tang S , Zhang X .

Null controllability for forward and backward stochastic parabolic equations

SIAM J Control Optim, 2009, 48, 2191- 2216

DOI:10.1137/050641508      [本文引用: 1]

Wu B , Gao Y , Wang Z , Chen Q .

Unique continuation for a reaction-diffusion system with cross diffusion

J Inverse Ill-Posed Prob, 2019, 27, 511- 525

DOI:10.1515/jiip-2017-0094      [本文引用: 2]

Wu B , Liu J .

Conditional stability and uniqueness for determining two coefficients in a hyperbolic-parabolic system

Inverse Probl, 2011, 27, 075013

DOI:10.1088/0266-5611/27/7/075013      [本文引用: 1]

Wu B , Yu J , Wang Z .

A coefficient identification problem for a mathematical model related to ductal carcinoma in situ

Stud Appl Math, 2019, 143, 356- 372

DOI:10.1111/sapm.12281      [本文引用: 1]

Yamamoto M .

Carleman estimates for parabolic equations and applications

Inverse Probl, 2009, 25, 123013

DOI:10.1088/0266-5611/25/12/123013      [本文引用: 1]

/