\documentclass[reqno]{amsart} \usepackage{hyperref} \AtBeginDocument{{\noindent\small \emph{Electronic Journal of Differential Equations}, Vol. 2010(2010), No. 148, pp. 1--8.\newline ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu \newline ftp ejde.math.txstate.edu} \thanks{\copyright 2010 Texas State University - San Marcos.} \vspace{9mm}} \begin{document} \title[\hfilneg EJDE-2010/148\hfil An iteration method] {An iteration method for controllability of semilinear parabolic equations} \author[B. Sun\hfil EJDE-2010/148\hfilneg] {Bo Sun} \address{Bo Sun \newline College of Mathematics and Computers, Changsha University of Science and Technology, Changsha, Hunan, China} \email{sunbo52002@yahoo.com.cn} \thanks{Submitted May 30, 2010. Published October 20, 2010.} \subjclass[2000]{35K61, 93B05} \keywords{Semilinear; parabolic equation; controllability; iteration method; \hfill\break\indent coupled systems} \begin{abstract} We present a method based on Picard's idea to construct a sequence of controls and a sequence of solutions of linearized systems such that their limits form a solution to the control problem. By doing this, we simplified the works in the references, and deduced the controllability for semilinear coupled parabolic systems. \end{abstract} \maketitle \numberwithin{equation}{section} \newtheorem{theorem}{Theorem}[section] \newtheorem{proposition}[theorem]{Proposition} \newtheorem{remark}[theorem]{Remark} \section{Introduction} The controllability for semilinear parabolic equations has been intensively studied in previous decades; see for example \cite{Fabre,Zuazua,Fernandez,Khapalov, Khapalov2,Khapalov3}. Some new systems were studied recently, such as nonlinear heat equation with memory effects, Lavanya \cite{Lavanya}, and some semilinear parabolic equations arising in finance, Sakthivel \cite{Sakthivel}. Tang and Zhang \cite{Tang} studied null controllability for forward and backward linear stochastic parabolic equations. It may not be easy to generalize their results to nonlinear cases. Generally speaking, all these works are based on the same idea: the methods of approximating controllability for linear systems are based on the unique continuation of solutions from an open set, while the exact null-controllability depends on the inverse estimates of dual observed systems. Furthermore, the controllability are generalized to semilinear systems by Schauder fixed-point arguments. All the arguments are based on the compact embedding of $H_0^1$ into $L^2$. The Schauder fixed-point arguments are good for semilinear parabolic systems with globally Lipschitz nonlinearity, but difficult for those with superlinear terms. In this paper, we try to do it in other ways, and give a method based on iteration idea, by which we simplified some works concerning the controllability of semilinear parabolic systems in literatures. Moreover, we deduced the controllability of coupled parabolic systems, which seems impossible to be obtained by previous arguments. This paper is organized as follows. In Section~2 we recall two lemmas which are necessary for our analysis. In Section~3 we prove the exact null-controllability of semilinear parabolic systems with mobile distributed control, to illustrate our method. We discuss the controllability of coupled parabolic systems in Section~4. \section{Preliminaries} For convenience, we recall the following two propositions, from \cite{Robinson}. \begin{proposition}\label{Robison 1} Let $X, H, Y$ be Banach spaces with $X\subset\subset H\subseteq Y$ and $X$ reflexive. Suppose that $\{u_n\}_{n=1}^{\infty}$ is a sequence that is bounded in $L^2(0,T;X)$, and that $\{\frac{{\rm d}u_n}{{\rm d}t}\}_{n=1}^{\infty}$ is bounded in $L^p(0,T;Y)$, for some $p>1$. Then there is a subsequence of $\{u_n\}_{n=1}^{\infty}$ that converges strongly in $L^2(0;T;H)$. \end{proposition} \begin{proposition}\label{Robison 2} Let $\mathcal{O}$ be a bounded open set in $\mathbb{R}^m$, and let $\{g_j\}_{j=1}^{\infty}$ be a sequence of functions in $L^p(\mathcal{O})$ with $$\|g_j\|_{L^p(\mathcal{O})}\le C \quad \text{for all } j\in \mathbb{Z}^+.$$ If $g_j\to g\in L^p(\mathcal{O})$ almost everywhere, then $g_j\rightharpoonup g$ in $L^p(\mathcal{O})$. \end{proposition} \section{Exact null-controllability for semilinear systems with mobile distributed control}\label{sec2} Consider the controllability for the Dirichlet problem \begin{gather} \frac{\partial u}{\partial t}=\Delta u+g(u)+f(x,t)\chi_{\omega(t)}(x) \quad \text{in } Q_T, \label{orignal1} \\ u=0 \quad\text{in } \Sigma_T, \quad u|_{t=0}=u_0,\quad u_0\in L^2(\Omega), \label{orignal2} \end{gather} where $\Omega$ is an open bounded domain in $\mathbb{R}^n$ with boundary $\partial \Omega$, $Q_T=\Omega\times (0,T)$, $\Sigma_T=\partial \Omega \times (0,T)$, $f$ is a distributed control in $L^2(Q_T)$, $\omega(\cdot)$ is a mobile support, and $\chi_{\omega(\cdot)}$denotes its indication function. We assume growth conditions as follows: \begin{gather} |g(u)| \leq c_1+c_2|u|^p, \quad p=1+4/n,\quad \forall u\in R,\label{cond1}\\ g(u)u \leq c_3u^2, \quad \forall u\in R, \label{cond2} \\ g(0)=0,\quad \lim_{u\to 0}\frac{g(u)}{u}=g'(0).\label{cond3} \end{gather} Conditions \eqref{cond1}-\eqref{cond3} ensure the existence of at least one generalized solution of \eqref{orignal1}-\eqref{orignal2} from $C([0,T];L^2(\Omega))\bigcap H_0^{1,0}(Q_T)$ \cite[pp. 466-467]{Ladyzhenskaja}. Khapalov proved that \eqref{orignal1}-\eqref{cond3} is exactly null-controllable \cite{Khapalov2}, but his generalization to nonlinear systems is somewhat complicated. So in this paper we put forward some different methods. The corresponding linear system is as follows: \begin{gather} \frac{\partial w}{\partial t}=\Delta w+a(x,t)w +f(x,t)\chi_{\omega(t)}(x) \quad\text{in } Q_T, \label{linear1}\\ w=0 \quad\text{in} \quad \Sigma_T,\quad w|_{t=0}=w_0\in L^2(\Omega). \label{linear2} \end{gather} The counterparts to \eqref{cond1}-\eqref{cond2} are \begin{gather} a(x,t)\leq c_3, \quad a\in L^{r_2}(0,T;L^{r_1}(\Omega)),\\ \frac{1}{r_2}+\frac{n}{2r_1}=1,\\ r_1\in (\frac{n}{2}, \infty],\quad r_2\in [1, \infty), \quad \text{for } n\geq 2,\\ r_1\in [1, \infty], \quad r_2\in [1, 2], \quad \text{for } n=1. \label{linear conditions} \end{gather} Khapalov proved that \eqref{linear1}-\eqref{linear conditions} is exactly null-controllable by a mobile internal control, while the support may be chosen to be arbitrarily small. Moreover, he gives the following estimates: \begin{gather} \|w\|_{L^q(Q_T)}, \|w\|_{\mathcal{B}} \leq c(T)(\|w_0\|_{L^2(\Omega)}+\|v\|_{L^2(\omega(\cdot))}), \label{estimate1}\\ \|f\|_{L^2(\omega(\cdot))} \leq M \|w_0\|_{L^2(\Omega)}, \label{estimate2} \end{gather} where $q=2+4/n$, $$\|w\|_\mathcal{B}=\|w\|_{C([0,T];L^2(\Omega))} +\Big(\int_0^T\int_\Omega \|\nabla w\|^2_{R^n}\text{d}x\text{d}t\Big)^{1/2},$$ $c(T)$ and $M$ do not depend on the choice of $a(x,t)$. Then he deduced the exact null-controllability for \eqref{orignal1}-\eqref{orignal2} by Schauder fixed-point argument, with $p\le 1+4/n$. Now we do in a different way. As usual, we introduce the non-linearity $$h(s)=\begin{cases} g(s)/s, & s\neq 0,\\ g'(0), & s=0, \end{cases}$$ and construct a linearized'' system $$\frac{\partial u}{\partial t} =\Delta u+h(z)u +f(x,t)\chi_{\omega(t)}(x)\quad \text{in } Q_T \label{linearized}$$ with the same initial and boundary conditions as \eqref{linear2}. We construct a sequence in $L^2(Q_T)$ as follows: Take any $z_1\in L^2(Q_T)$, substitute it for $z$ in \eqref{linearized}, then there exists a control $f_1\in L^2(Q_T)$ such that the corresponding solution $u_1$ satisfies that $u_1(T)=0$; Take $z_2=u_1$, substitute it for $z$ in \eqref{linearized}, then there exists a control $f_2\in L^2(Q_T)$ such that the corresponding solution $u_2$ satisfies that $u_2(T)=0$; Repeating this process yields two sequences $\{z_n\}_{n=1}^{\infty}$(or $\{u_n\}_{n=1}^{\infty}$) and $\{f_n\}_{n=1}^{\infty}$ in $L^2(Q_T)$. This process can be illustrated as follows: \begin{gather} z_1\Rightarrow f_1\Rightarrow u_1=z_2\Rightarrow f_2\Rightarrow u_2= \dots , \notag\\ \frac{\partial u_n}{\partial t}=\Delta u_n+h(z_n)u_n+f_n\chi_{\omega(t)}. \label{iterationeq} \end{gather} Now we prove that they converge in some topology, their limits solve \eqref{orignal1}-\eqref{orignal2}, and the limit satisfies that $u(T)=0$. To simplify notation, we denote $\int_{Q_T}\cdot {\rm d}x{\rm d}t$ by $\int_{Q_T}\cdot$. Condition \eqref{cond3} implies that there is $r>0$ such that $$|h(z)|\le c_4=|g'(0)|+1$$ when $|z|0$ and finite dimensional subspaces $E_1$, $E_2$ of $L^2(\Omega)$, there exist controls $f$ and $g$ in $L^2(Q_T)$ such that \begin{gather} \|u(T)-u_T\|_{L^2(\Omega)}\le \varepsilon, \quad \Pi_1u(T)=\Pi_1u_T, \label{coupledcontro1} \\ |v(T)-v_T|_{L^2(\Omega)}\le \varepsilon, \quad \Pi_2v(T)=\Pi_2v_T, \label{coupledcontro2} \end{gather} where $\Pi_1$ and $\Pi_2$ are the orthogonal projections from $L^2(\Omega)$ into $E_1$ and $E_2$ respectively. It seems difficult or impossible to construct a linearized systems for \eqref{couple1}-\eqref{couple4}, as \eqref{linearized} for \eqref{orignal1}-\eqref{orignal2}. We will deduce the controllability without linearized system. At first, we construct sequences of solutions and controls as follows: Take any $v_1\in L^2(Q_T)$, substitute it for $v$ in \eqref{couple1}, then it follows from the controllability results on single parabolic systems (see \cite{Zuazua}) that there exists control $f_1\in L^2(Q_T)$ such that the corresponding solution $u_1$ of \eqref{couple1} satisfies \eqref{coupledcontro1}. Substitute $u_1$ for $u$ in \eqref{couple2}, there exists control $g_1\in L^2(Q_T)$ such that the corresponding solution $v_2$ satisfies \eqref{coupledcontro2}. Then substitute $v_2$ for $v$ in \eqref{couple1}, and get $f_2$ and $u_2$. Repeating this process yields sequences $\{u_n\}_{n=1}^{\infty}$, $\{v_n\}_{n=1}^{\infty}$, $\{f_n\}_{n=1}^{\infty}$ and $\{g_n\}_{n=1}^{\infty}$, which satisfy \eqref{coupledcontro1} and \eqref{coupledcontro2}. This process can be illustrated as follows: \begin{gather} v_1\Rightarrow f_1\Rightarrow u_1\Rightarrow g_1\Rightarrow v_2\Rightarrow f_2\Rightarrow u_2\Rightarrow \dots , \notag \\ \frac{\partial u_n}{\partial t}-\Delta u_n=\Phi (u_n,v_n)+f_n\chi_{\omega_1}, \label{coupleiteration1} \\ \frac{v_{n+1}}{\partial t}-\Delta v_{n+1}=\Psi (u_n,v_{n+1})+g_n\chi_{\omega_2}. \label{coupleiteration2} \end{gather} We will prove that these sequences converge in some topology, and that their limits solve \eqref{couple1}-\eqref{couple4}, which satisfy \eqref{coupledcontro1} and \eqref{coupledcontro2}. Due to the global Lipschitz continuity of $\Phi$ and $\Psi$, sequences $\{f_n\}_{n=1}^{\infty}$, $\{g_n\}_{n=1}^{\infty}$, $\{u_n\}_{n=1}^{\infty}$ and $\{v_n\}_{n=1}^{\infty}$ are bounded in $L^2(Q_T)$. So they have subsequences which converge weakly. Let $u_j\rightharpoonup u$, $v_j\rightharpoonup v$, $f_j\rightharpoonup f$ and $g_j\rightharpoonup g$ in $L^2(Q_T)$. By similar arguments to those in last section, we have that \begin{gather*} \Phi (u_n,v_n)\rightharpoonup \Phi (u,v),\\ \Psi (u_n,v_{n+1})\rightharpoonup \Psi (u,v) \end{gather*} in $L^2(Q_T)$. Taking limits of \eqref{coupleiteration1} and \eqref{coupleiteration2}, one can easily verify that $u$, $v$, $f$ and $g$ solve \eqref{couple1}-\eqref{couple4}, and satisfy \eqref{coupledcontro1}-\eqref{coupledcontro2}. We summarize our analysis as follows: \begin{theorem} \label{couplethem} Suppose that $\Phi$ and $\Psi$ are continuous functions, $\Phi$ is globally Lipschitz continuous and differentiable with respect to $u$, so is $\Psi$ with respect to $v$, and $$\Phi(0,v)=\Psi(u,0)=0, \quad \forall u,v\in\mathbb{R}.$$ Then \eqref{couple1}-\eqref{couple4} is globally approximately controllable and finite-dimensional exactly controllable. \end{theorem} \begin{remark} \label{rmk4.2} \rm By our method, all results concerning controllability of semilinear parabolic systems can be extended to coupled parabolic systems. For example, We may consider exact null-controllability of semilinear coupled parabolic systems with superlinear terms, mobile internal controls or point-controls. \end{remark} From the point view of application, it is reasonable to consider coupled parabolic system with a single control as follows: \begin{gather} u_t-\Delta u=\Phi(u,v)+f\chi_{\omega} \quad \text{in } Q_T, \label{single1} \\ v_t-\Delta v=\Psi(u,v) \quad \text{in } Q_T, \label{single2} \\ u(x,t)=v(x,t)=0 \quad \text{on } \Sigma_T , \label{single3} \\ u|_{t=0}=u_0, \quad v|_{t=0}=v_0, \quad u_0,v_0 \in L^2(\Omega). \label{single4} \end{gather} \begin{theorem} \label{onecontroltheorem} Under the same conditions as in Theorem \ref{couplethem}, system \eqref{single1}-\eqref{single4} is globally approximately controllable and finite-dimensional exactly controllable. \end{theorem} The proof is similar to that of Theorem \ref{couplethem}, but there is some difference. Let us give a sketch: Given $u_T\in L^2(\Omega)$, $\varepsilon >0$ and finite-dimensional subspace $E$ of $L^2(\Omega)$. Take any $v_1\in L^2(Q_T)$, substitute it for $v$ in \eqref{single1}, then there exist control $f_1$ and the corresponding solution $u_1$ such that $\|u_1(T)-u_T\|_{L^2(\Omega)}< \varepsilon$, and $\Pi_Eu_1(T)=\Pi_Eu_T$. Substitute $u_1$ for $u$ in \eqref{single2}, then there is an unique solution $v_2$. Substitute $v_2$ for $v$ in \eqref{single1}, there exist control $f_2$ and the corresponding solution $u_2$, which satisfy that $\|u_2(T)-u_T\|_{L^2(\Omega)}< \varepsilon$, and $\Pi_Eu_2(T)=\Pi_Eu_T$. Repeating this process yields sequences $\{u_n\}_{n=1}^{\infty}$, $\{v_n\}_{n=1}^{\infty}$ and $\{f_n\}_{n=1}^{\infty}$. Their limits solve \eqref{single1}-\eqref{single4}, and satisfy \begin{gather*} \|u(T)-u_T\|_{L^2(\Omega)}<\varepsilon ,\\ \Pi_Eu(T)=\Pi_Eu_T, \end{gather*} where $\Pi_E$ represents the orthogonal projection from $L^2(\Omega)$ into $E$. Theorem \ref{couplethem} and Theorem \ref{onecontroltheorem} can be generalized to multi-coupled parabolic systems as follows: \begin{gather*} \frac{\partial u_1}{\partial t}-\Delta u_1= \Phi_1(u_1,u_2,\dots,u_m)+f_1\chi_{\omega_1} \quad \text{in } Q_T, \\ \frac{\partial u_2}{\partial t}-\Delta u_2= \Phi_2(u_1,u_2,\dots,u_m)+f_2\chi_{\omega_2} \quad \text{in } Q_T, \\ \dots \\ \frac{\partial u_m}{\partial t}-\Delta u_m= \Phi_m(u_1,u_2,\dots,u_m)+f_m\chi_{\omega_m} \quad \text{in } Q_T, \\ u_i|_{\Sigma_T}=0,\quad u_i|_{t=0}=u_{i0},\quad u_{i0}\in L^2(\Omega), \quad i=1,2,\dots,m. \end{gather*} \section{Examples} The following example, from combustion theory \cite{Temam}, illustrates our results. \begin{gather} \frac{\partial u_1}{\partial t}-d_1\Delta u_1-(u_2)^ph(u_1)=f\chi_{\omega_1} \quad \text{in } Q_T, \label{examp1e1} \\ \frac{\partial u_2}{\partial t}-d_2\Delta u_2-(u_2)^ph(u_1)=g\chi_{\omega_2} \quad \text{in } Q_T, \label{examp1e2} \\ u_1(x,t)=u_2(x,t)=0 \quad \text{on} \quad \Sigma_T, \label{example3} \end{gather} where $h(s)=|s|^\gamma\exp(-\alpha/|s|)$, and $p$, $\alpha$, $\gamma$ are positive constants. $u_1$ represents a temperature, while $u_2$ represents a concentration. The nonlinearity vanishes at $u_1=0$ and $u_2=0$. It is Lipschitz continuous, provided that $p=\gamma=1$. It follows from section 4 that the coupled system \eqref{examp1e1}-\eqref{example3} is globally approximately controllable and exactly null-controllable. \begin{thebibliography}{00} \bibitem{Escobedo} M. Escobedo, M. A. Herrero; Boundedness and blow up for a semilinear reaction diffusion systems,'' \emph{J. Diff. Eqs.}, 89(1991), 176-202. \bibitem{Fabre} C. Fabre, J.-P. Puel, and E. Zuazua; Approximate controllability for the semilinear heat equation,'' \emph{Proc. Roy. Soc. Edinburgh}, 125(1995), 31-61. \bibitem{Fernandez} L. A. Fernandez and E. Zuazua; Approximate controllability of the semilinear heat equation involving gradient terms,'' \emph{J. Opti. Appl.}, 101(1999), 307-328. \bibitem{Henry} D. Henry; \emph{Geometric Theory of Semilinear Parabolic Equations}, Springer, Berlin, 1981. \bibitem{Khapalov} A. Y. Khapalov; A class of globally controllable semilinear heat equations with superlinear terms,'' \emph{J. Opti. Appl. }, 110(2001), 245-264. \bibitem{Khapalov2} A. Y. Khapalov; Exact null-controllability for semilinear heat equation with superlinear term and mobile internal controls,'' \emph{Nonlinear analysis}, 43(2001), 785-801. \bibitem{Khapalov3} A. Y. Khapalov; \emph{Mobile point controls versus locally distributed ones for the controllability of semilinear parabolic equation,} Technical report, Washington State University, 1999. \bibitem{Ladyzhenskaja} O. H. Ladyzhenskaja, V. A. Solonnikov, N. N. Ural'ceva; \emph{Linear and Quasi-linear Equations of Parabolic Type.} AMS, Providence, Rhode Island, 1968. \bibitem{Lavanya} R. Lavanya, K. Balachandran; Null controllability of nonlinear heat equation with memory effects,'' \emph{Nonlinear analysis: Hybrid Systems}, 3(2009), 163-175. \bibitem{Robinson} J. C. Robinson; \emph{Infinte-Dimensional Dynamical Systems}, Cambridge University Press, Cambridge, 2001. \bibitem{Sakthivel} K. Sakthivel; G. Devipriya, K. Balachandran, J.-H. Kim, Exact null controllability of a semilinear parabolic equation arisining in finance,'' \emph{Nonlinear analysis:Hybrid Systems}, 3(2009), 565-577. \bibitem{Tang} Shanjian Tang, Xu Zhang; Null controllability for forward and backward stochastic parabolic equations,'' \emph{SIAM J.Control Optim.}, 48(2009), 2191-2216. \bibitem{Temam} R. Temem; \emph{Infinite-Dimensional Dynamical Systems in Mechnics and Physics}, Springer, New York, 1998. \bibitem{Zuazua} E. Zuazua; Finite dimensional null controllability for the semilinear heat equation,'' \emph{J. Math. Pures Appl.}, 76(1997), 237-264. \end{thebibliography} \end{document}