Carleman estimate for full-discrete approximations of the complex Ginzburg-Landau equation with dynamic boundary conditions and applications to controllability

Xu Zhu, Wenwen Zhou, Bin Wu · 2025

Read Carleman estimate for full-discrete approximations of the complex Ginzburg-Landau equation with dynamic boundary conditions and applications to controllability by Xu Zhu, Wenwen Zhou, Bin Wu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-20628.txt.

Introduction

Xu Zhu, Wenwen Zhou, and Bin Wu proved a Carleman estimate in twenty twenty-five for a fully discrete version of the complex Ginzburg-Landau equation. This equation models systems from superconductivity to fluid patterns, and here it carries dynamic boundary conditions, where the boundary itself evolves. Carleman estimates are weighted energy bounds that power results about control and observability.

Working on a fully discrete scheme — a grid in both space and time — the authors first prove the estimate for the adjoint system. From it they derive a relaxed observability inequality, saying the whole state can be partly recovered from boundary readings. That yields a controllability result for the discrete equation. Such results matter when control must be designed on the same grid a computer will use.

Catalog record

CREATORXu Zhu, Wenwen Zhou, Bin Wu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE128 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-20628

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Carleman estimate for full-discrete approximations of the complex Ginzburg-Landau equation with dynamic boundary conditions and applications to controllability Xu Zhu, Wenwen Zhou, Bin Wu Abstract In this paper, we investigate Carleman estimate and controllability result for the fully-discrete approximations of a one-dimensional Ginzburg-Landau equation with dynamic boundary conditions. We first establish a new discrete Carleman estimate for the corresponding adjoint system. Based on this Carleman estimate, we obtain a relaxed observability inequality for the adjoint system, and then a controllability result for the fully-discrete Ginzburg-Landau equation with dynamic boundary conditions. Keywords: Carleman estimates, fully-discrete Ginzburg-Landau equation, dynamic boundary conditions, null controllability 1 Introduction 1.1 Motivation Let T>0 , \Omega=(0,1) and \omega be a nonempty subset of \Omega . Given the parameters \alpha>0 , c\in\mathbb{R} and \beta,\gamma\in\mathbb{R}\backslash\{0\} , we consider the following linear Ginzburg-Landau equation with dynamic boundary conditions [ 16 ]: \left\{\begin{aligned} &\partial_{t}y-(\alpha+i\beta)\partial_{x}^{2}y+(c+i\gamma)y={\bf 1}_{\omega}v,&&(x,t)\in\Omega\times(0,T),\\ &\partial_{t}{y}_{\Gamma_{0}}(t)-(\alpha+i\beta)\partial_{x}y(0,t)+(c+i\gamma)y_{\Gamma_{0}}=0,&&t\in(0,T),\\ &\partial_{t}{y}_{\Gamma_{1}}(t)+(\alpha+i\beta)\partial_{x}y(1,t)+(c+i\gamma)y_{\Gamma_{1}}=0,&&t\in(0,T),\\ &y_{\Gamma_{0}}(t)=y(0,t),\quad y_{\Gamma_{1}}(t)=y(1,t),&&t\in(0,T),\\ &y(x,0)=y^{0}(x),&&x\in\Omega,\\ &y_{\Gamma_{0}}(0)=y_{\Gamma_{0}}^{0},\quad y_{\Gamma_{1}}(0)=y_{\Gamma_{1}}^{0},&&x\in\Omega.\end{aligned}\right. (1.1) Here (y_{\Gamma_{0}},y,y_{\Gamma_{1}})=(y_{\Gamma_{0}}(t),y(x,t),y_{\Gamma_{1}}(t)) is the state of the system, (y^{0}_{\Gamma_{0}},y^{0},y^{0}_{\Gamma_{1}}) is the initial condition, and the control function v is acting on an open subinterval \omega\subset\Omega through {\bf 1}_{\omega}(x) , the characteristic function of \omega . Recently, a variety of mathematical models with dynamic boundary conditions have attracted much attention owing to their extensive applications, such as in population dynamics [ 17 ], heat transport in different media [ 22 ] and oscillations under strong friction [ 8 ]. We refer to [ 13 ] for the well-posedness of system ( 1.1 ) under suitable assumptions in the class of (y_{\Gamma_{0}},y,y_{\Gamma_{1}})\in C([0,T];\mathbb{H}^{1}) , where \displaystyle\mathbb{H}^{1}:=\{(u_{\Gamma_{0}},u,u_{\Gamma_{1}})\in\mathbb{R}\times L^{2}(\Omega)\times\mathbb{R};\ u\in H^{1}(\Omega),u(0)=u_{\Gamma_{0}},u(1)=u_{\Gamma_{1}}\}. The null controllability for ( 1.1 ) is formulated as follows: for any initial state (y^{0}_{\Gamma_{0}},y^{0},y^{0}_{\Gamma_{1}}) \in\mathbb{H}^{1} , one can find a control v\in L^{2}(\omega\times(0,T)) such that the associated solution (y_{\Gamma_{0}},y,y_{\Gamma_{1}}) of ( 1.1 ) satisfies \displaystyle y(x,T)=0,\quad{x}\in\Omega,\quad{\rm and}\quad\left(y_{\Gamma_{0}}(t),y_{\Gamma_{1}}(t)\right)=(0,0),\quad t\in(0,T). In [ 13 ], the null controllability for a cubic Ginzburg-Landau equation with dynamic boundary conditions was studied. More precisely, consider the following cubic Ginzburg-Landau equation with dynamic boundary conditions \left\{\begin{aligned} &\partial_{t}y-(\alpha+i\beta)\Delta y+(c+i\gamma)|y|^{2}y={\bf 1}_{\omega}v,&&(x,t)\in\Omega\times(0,T),\\ &\partial_{t}y_{\Gamma}+(\alpha+i\beta)\partial_{\nu}y-(\alpha+i\beta)\Delta_{\Gamma}y_{\Gamma}+(c+i\gamma)|y_{\Gamma}|^{2}y_{\Gamma}=0,&&(x,t)\in\Gamma\times(0,T),\\ &y=y_{\Gamma},&&(x,t)\in\Gamma\times(0,T),\\ &(y(0),y_{\Gamma}(0))=(y^{0},y^{0}_{\Gamma}),&&(x,t)\in\Omega\times\Gamma,\end{aligned}\right. (1.2) where \Omega\subset\mathbb{R}^{d} (d\geq 1) is a bounded domain with boundary \Gamma of class C^{2} , \partial_{\nu}y is the normal derivative associated to the outward normal \nu of \Omega and \Delta_{\Gamma} denotes the Laplace-Beltrami operator on \Gamma . Firstly the authors investigated the null controllability for a linearized system by the duality approach and an appropriate Carleman estimate. Further, the local null controllability of the nonlinear system was obtained by an inverse mapping theorem. From proposition 4.3 in [ 13 ] the null controllability for ( 1.1 ) could be deduced directly. There are also some known controllability and observability results about the deterministic and stochastic complex Ginzburg-Landau equations in continuous settings [ 19 , 21 , 32 ]. The main objective of this paper is to investigate discrete Carleman estimate and controllability for the full discrete approximations of system ( 1.1 ). It is well known that the null controllability result does not necessarily remain valid in the discrete setting. For example, a specific 2-D counterexamples was proposed for which a spatial discrete parabolic equation is not approximately controllable for discrete parameter \Delta x in [ 37 ]. Thus, we are interested in the \phi(\Delta x) -null controllability of the system ( 1.1 ). More precisely, we want to obtain uniformly bounded controls such that the norm of the semi-discrete solution at time T , y(T) , is approximately of the size \sqrt{\phi(\Delta x)} with a function \phi that tends to zero exponentially as space discretization parameter tends to zero. In order to formulate our discrete controllability, let us consider M,N\in\mathbb{N}^{*} , the space and time-discretization parameters \Delta x=\frac{1}{M+1} and \triangle t=\frac{T}{N} . Furthermore, we introduce an equidistant spatial mesh of (0,1) , 0=x_{0}<x_{1}<\cdots<x_{M+1}=1 , and an equidistant time mesh of (0,T) , 0=t^{0}<t^{1}<\cdots<t^{N}=T . Let x_{j}=jh , t^{n}=n\triangle t and \displaystyle g_{j}=\left\{\begin{array}[]{ll}y^{0}_{\Gamma_{0}},&j=0,\\ y^{0}(x_{j}),&1\leq j\leq M,\\ y^{0}_{\Gamma_{1}},&j=M+1,\end{array}\right.\quad{\rm and}\quad y^{n}_{j}=y(x_{j},t_{n}),\quad 1\leq j\leq M,1\leq n\leq N. Then the full-discrete approximations of ( 1.1 ) can read as follows \left\{\begin{array}[]{ll}\frac{y^{n+1}_{j}-y^{n}_{j}}{\Delta t}-(\alpha+i\beta)\frac{y^{n+1}_{j+1}-2y^{n+1}_{j}+y^{n+1}_{j-1}}{(\Delta x)^{2}}\\ \hskip 122.34692pt+(c+i\gamma)y_{j}^{n+1}={\bf 1}_{\omega}v^{n+1}_{j},&1\leq j\leq M,\ 0\leq n\leq N-1,\\ \frac{y^{n+1}_{0}-y^{n}_{0}}{\Delta t}-(\alpha+i\beta)\frac{y^{n+1}_{1}-y^{n+1}_{0}}{\Delta x}+(c+i\gamma)y_{0}^{n+1}=0,&0\leq n\leq N-1,\\ \frac{y^{n+1}_{M+1}-y^{n}_{M+1}}{\Delta t}+(\alpha+i\beta)\frac{y^{n+1}_{M}-y^{n+1}_{M-1}}{\Delta x}+(c+i\gamma)y_{M+1}^{n+1}=0,&0\leq n\leq N-1,\\ y^{0}_{j}=g_{j},&0\leq j\leq M+1.\end{array}\right. (1.3) Now, we state our discrete null controllability problem in this paper. To formulate it, we introduce the notation g=(g_{0},g_{1}\cdots,g_{M+1})^{\rm T}\quad{\rm and}\quad y^{n}=(y^{n}_{0},y^{n}_{1},\cdots,y^{n}_{M+1})^{\rm T},\ n=0,1,\cdots,N to denote discrete functions. Discrete \phi(\Delta x) -null controllability problem. For any initial data g\in L^{2}(\overline{\mathcal{M}}) , find control v\in{L}^{2}(\omega\times\mathcal{N}^{*}) such that the corresponding solution y satisfies \left\|y^{N}\right\|_{{L}^{2}(\overline{\mathcal{M}})}\leq\sqrt{\phi(\Delta x)}\left\|g\right\|_{{L}^{2}(\overline{\mathcal{M}})} with a function \phi(\Delta x) , where the notations of discrete domains \mathcal{M} , \mathcal{N}^{*} , discrete function spaces L^{2}(\overline{\mathcal{M}}) and {L}^{2}(\omega\times\mathcal{N}^{*}) , and discrete norm \|\cdot\|_{{L}^{2}(\overline{\mathcal{M}})} will be presented in Section 1.2. There has been much work about the study of controllability for various partial differential equations with dynamic boundary conditions. Such sort of works mainly focus on the continuous cases. We refer to [ 24 ] for parabolic equations, [ 30 ] for Schrödinger equations, [ 3 , 4 ] for stochastic parabolic equations and [ 13 ] for Ginzburg-Landau equations. Moreover, as for Carleman estimates for complex partial differential equations, we refer to [ 19 ] for parabolic equation with a complex principal part, [ 5 ] for Schrödinger equations and [ 20 , 21 ] for complex Ginzburg–Landau equations. There are rich references on discrete Carleman estimates for various partial differential equations. In [ 6 ], a discrete Carleman estimate for the spatial semi-discrete hyperbolic equation was established and subsequently applied it to obtain the stability of an inverse problem to recover a potential term in this discrete hyperbolic equation. Discrete Carleman estimate for one-dimensional elliptic equation has been developed in [ 11 ] to establish a relaxed observability for the associated semi-discrete parabolic equation. As for discrete Carleman estimates for parabolic equations, we refer to [ 10 ] for time semi-discrete case, [ 12 , 31 ] for spatial semi-discrete case, and [ 14 ] for full-discrete case. These discrete Carleman estimates were applied to obtain relaxed controllability results for the associated discrete parabolic equations. In [ 15 ] and [ 23 ], discrete Carleman estimates for fourth-order differential operators were established. Moreover, there are some works about discrete Carleman estimates for stochastic differential equations, e.g. [ 34 , 35 ] for stochastic second-order semi-discrete parabolic equations and [ 33 ] for stochastic fourth-order semi-discrete parabolic equations. To the best of our knowledge, there is only one paper to discuss Carleman estimates and controllability for discrete differential equations with dynamic boundary conditions [ 25 ], in which Carleman estimate for a fully discrete approximations of the parabolic equation with dynamic boundary conditions was established and then applied to obtain \phi(\Delta x) -null controllability. However, to the author’s knowledge, discrete Carleman estimates and controllability for the full discrete approximations of Ginzburg-Landau equation with dynamic boundary conditions have not been studied thoroughly yet. In order to handle dynamic boundary conditions, we need to carefully combine all boundary terms in proving the Carleman estimates. It is well-known that Ginzburg-Landau equation is closely related to both parabolic equations and Schrödinger equations. In order to reveal the distinctions between our Carleman estimate and the ones for these equations, we also decouple \alpha and \beta from the related constants appearing

Read the complete text