scientific article; zbMATH DE number 4026009
The author considers the celebrated (and ill-posed) Cauchy problem for the Laplace equation in an open bounded set \(D\subset {\mathbb{R}}^ 2\), whose boundary \(\partial D\) is a smooth closed curve with outward normal n. If \(\Gamma_ 1\) denotes an open subset in \(\partial D\) with positive length, he looks for a solution u to the problem \[ (1)\quad \Delta u=0\quad in\quad D;\quad u=f\quad on\quad \Gamma_ 1;\quad D_ nu=g\quad on\quad \Gamma_ 1, \] f and g being prescribed functions. He states that a solution u to (1) can be constructed by solving (in the weak sense) the following B. V. P., where \(\Gamma_ 2=\partial D\setminus \Gamma_ 1:\) \[ (2)\quad \Delta^ 2v=0\quad in\quad D;\quad v=f\quad on\quad \Gamma_ 1;\quad D_ nv=g\quad on\quad \Gamma_ 1;\quad \Delta v=0\quad on\quad \Gamma_ 2;\quad D_ n\Delta v=0\quad on\quad \Gamma_ 2. \] The test functions are assumed to belong to a Hilbert space H(D) consisting of functions \(\phi\) such that \[ (3)\quad \phi =D_ n\phi =0\quad on\quad \Gamma_ 1;\quad \Delta \phi =D_ n\Delta \phi =0\quad on\quad \Gamma_ 2. \] Though the author does not enter specifically the technical details, the reviewer thinks that a reasonable choice for H(D) is the following: \[ (4)\quad H(D)=\{\phi \in W^{2,2}(D): \phi =D_ n\phi =0\quad on\quad \Gamma_ 1\}, \] \(W^{k,p}(D)\) denoting the usual Sobolev spaces related to D. In that case the weak formulation of (2) is the following: \[ (5)\quad \int_{D} \Delta v\Delta \phi dx=0,\quad \forall \phi \in H(D). \] Under the assumptions \(f\in W^{3/2,2}(\Gamma_ 1)\) and \(g\in W^{1/2,2}(\Gamma)\) problem (5) (i.e. (4)) admits a unique solution \(v\in W^{2,2}(D)\) which satisfies the estimate \[ (6)\quad \| v\|_{W^{2,2}(D)}\leq C[\| f\|_{W^{3/2,2}(\Gamma_ 1)}+\| g\|_{W^{1/2,2}(\Gamma_ 1)}] \] for some positive constant C independent of (f,g). Then the author observes that the function \(w=\Delta v\) solves problem (1) with \((f,g)=(0,0)\) and \(\Gamma_ 1\) replaced by \(\Gamma_ 2\). Using uniqueness results for problem (1), the author concludes that \(w=\Delta v=0\) in D. Hence v solves problem (1) and satisfies estimate (6). This result contrasts with the following example, where D is an open bounded set with a \(C^{\infty}\) boundary \(\partial D\) containing the segment \(\Gamma_ 1\) with endpoints (0,0) and (0,1) and the arc of circle \(r=r_ 0\), \(\theta_ 1\leq \theta \leq \theta_ 2\) \((r_ 0>1,\) \(0< \theta_ 1< \pi/4< 3\pi/4< \theta_ 2< \pi).\) Then the functions \(u_ n(r \cos \theta,r \sin \theta)=(n+1)^{- 1}r^{n+1}\sin [(n+1)\theta]\) (n\(\in {\mathbb{N}})\) solve problem (1) with \(f_ n=0\) and \(g_ n(r)=r^ n\). Yet \[ \| u_ n\|_{W^{2,2}(D)}\| g_ n\|^{-1}_{W^{1/2,2}(\Gamma_ 1)}\geq C\| u_ n\|_{C(\bar D)}\| g_ n\|^{- 1}_{W^{1/2,2}(\Gamma_ 1)}\to \infty \quad as\quad n\to \infty, \] \(C_ 1\) being the Sobolev embedding constant related to the pair \((W^{2,2}(D)\), \(C(\bar D))\). This contradicts (6).
- scientific article; zbMATH DE number 4013190
- scientific article; zbMATH DE number 4040453
- Embedding methods for solving variational inequalities
- A variational method and approximations of a Cauchy problem for elliptic equations
- Variational Methods for the Numerical Solution of Nonlinear Elliptic Problems
- scientific article; zbMATH DE number 66473
- scientific article; zbMATH DE number 2144564
- scientific article; zbMATH DE number 428286
- scientific article; zbMATH DE number 3905318
- An embedded corrector problem to approximate the homogenized coefficients of an elliptic equation
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3767768)