Exact Boundary Controllability of Maxwell's Equations in Heterogeneous Media and an Application to an Inverse Source Problem

From MaRDI portal





The author deals with the controllability of Maxwell's equations in a heterogeneous medium with a nonsmooth boundary by means of control currents on the boundary of that medium. He considers Maxwell's equations: \[ \begin{cases} \varepsilon{\partial E\over\partial t}- \text{curl }H= 0\quad &\text{in }Q_T,\\ \mu{\partial H\over\partial t}+ \text{curl }E= 0\quad &\text{in }Q_T,\\ \text{div}(\varepsilon E)= \text{div}(\mu H)= 0\quad &\text{in }Q_T,\\ H\times\nu= J\quad &\text{on }\Sigma_{0T},\\ H\times\nu= 0\quad &\text{on }\Sigma_1\setminus\Sigma_{0T},\;E\cdot\nu= 0\text{ on }\Sigma_T,\\ E(0)= E_0,\;H(0)= H_0\quad &\text{in }\Omega,\end{cases}\tag{1} \] where \(T> 0\), \(Q_T:= \Omega\times ]0,T[\), \(\Sigma_T= \Gamma\times ]0,T[\), \(\Gamma= \partial\Omega\), \(\Gamma_0\subset\Gamma\), \(\Sigma_{0T}= \Gamma_0\times ]0,T[\), and \(\nu\) denotes the unit outer normal vector on \(\Gamma\). The interest is to solve the following exact controllability problem: given a time \(T>0\) and initial data \(\{E_0, H_0\}\), find a surface density of current \(J\) in appropriate function spaces such that the solution of (1) satisfies \(E(T)= H(T)= 0\) in \(\Omega\). Lions' HUM-method is used here. The exact controllability problem is equivalent to the unique solvability of the following adjoint problem of (1): \[ \begin{cases} \mu{\partial\varphi\over\partial t}- \text{curl }\psi= 0\quad &\text{in }Q,\\ \varepsilon{\partial\psi\over\partial t}+ \text{curl }\varphi=0\quad &\text{in }Q,\\ \text{div}(\mu \varphi)= \text{div}(\varepsilon\psi)= 0\quad &\text{in }Q,\\ \varphi\times \nu= 0,\;\psi\cdot\nu= 0\quad &\text{on }\Sigma,\\ \varphi(0)= \varphi_0,\;\psi(0)= \psi_0\quad &\text{in }\Omega,\end{cases}\tag{2} \] which is obtained with the help of observation estimates. One introduces the following spaces: \[ \begin{aligned} J(\Omega,\varepsilon) & =\{\chi\in L^2(\Omega)^3\mid \text{div}(\varepsilon\chi)= 0\},\\ \widehat J(\Omega,\varepsilon) &= \{\chi\in J(\Omega, \varepsilon)\mid \chi\cdot\nu= 0\text{ on }\Gamma\},\\ J^1_\nu(\Omega, \mu) &= \{\chi\in\widehat J(\Omega, \mu)\mid \text{curl }\chi\in L^2(\Omega)^3\},\\ J^1_\tau(\Omega, \varepsilon) &= \{\chi\in J(\Omega, \varepsilon)\mid \text{curl }\chi\in L^2(\Omega)^3\text{ and }\chi\times \nu=0\text{ on }\Gamma\}.\end{aligned} \] One defines the operator \(A\) as \[ D(A)= J^1_\tau(\Omega, \mu)\times J^1_\nu(\Omega,\varepsilon), \] \[ A\begin{pmatrix} \varphi\\ \psi\end{pmatrix}= \begin{pmatrix} \mu^{-1}\text{curl }\psi\\ -\varepsilon^{-1} \text{curl } \varphi.\end{pmatrix} \] Also, if \((\varphi, \psi)\) is a strong solution of (2) with initial datum in \(D(A)\), then one defines the energy at time \(t\) by \[ \widetilde E(t)= {1\over 2} \int_\Omega \{\mu^{- 1}|\text{curl }\psi(x, t)|^2+ \varepsilon^{-1}|\text{curl } \varphi(x,t)|^2\} dx. \] If \(\Gamma_0\) is an open part of \(\Gamma\), then \(\Sigma_0= \Gamma_0\times (0,T)\) is said to satisfy the \((\varepsilon, \mu)\)-strong observation estimate if there exists \(C>0\) such that: \[ \widetilde E(0)\leq C\int_{\Sigma_0} \varepsilon^{- 1}|\text{curl }\varphi|^2 d\sigma dt \] for all solutions \((\varphi, \psi)\) of (2). Analogously, \(\Sigma_0\) satisfies the \((\varepsilon, \mu)\)-weak obsrvation estimate if \[ \widetilde E(0)\leq C \int_{\Sigma_0} (\varepsilon^{- 1}|\text{curl } \varphi|^2+ \mu^{- 1}|\text{curl }\psi|^2) d\sigma dt. \] It is proved that if \(\Sigma_0\) satisfies the \((\varepsilon, \mu)\)-strong observation estimate, then for all initial data \(\{H_0, -E_0\}\), the control \(J= -\psi\in L^2(\Sigma_0)^3\) derives the system (1) to rest at time \(T\). Analogously, if \(\Sigma_0\) satisfies the \((\varepsilon, \mu)\)-weak observation estimate, then for all \(\{H_0, -E_0\}\in D(A)'\), the control \(J= \psi- \mu\sigma\times \varphi'\) drives our system (1) to rest at time \(T\), where \(\mu\sigma\times \varphi'\) is defined as: \[ (\mu\sigma\times \varphi', \psi)= \int^t_0 \int_{\Gamma_0} \mu(\varphi'\cdot \nu)(\sigma\times \psi)\cdot \nu d\sigma dt. \] On the other hand, an application of the above results to an inverse problem is given, namely the determination and reconstruction of antennas from boundary measurements. Finally, for \(\varepsilon\) and \(\mu\) piecewise constant with Lipschitz polyhedral subdomains, there are established different observation estimates for \(\Sigma\) adapting to our setting the results of other works, and it is given an estimate of the minimal time of control.




Cited in
(31)








This page was built for publication: Exact Boundary Controllability of Maxwell's Equations in Heterogeneous Media and an Application to an Inverse Source Problem

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4507401)