Appending boundary conditions by Lagrange multipliers: Analysis of the LBB condition (Q5943697): Difference between revisions

The authors address the Dirichlet problem for a second-order elliptic partial differential operator in arbitrarily-dimensioned space. The problem is addressed in its weak form over the bounded domain \(\Omega\) and boundary conditions are introduced as Lagrange multipliers on its Lipschitz continuous boundary, \(\partial\Omega\). The domain \(\Omega\) is assumed to be embedded in a possibly larger domain \(\Upsilon\) which can be, for example, a cube. Solutions will be approximated by elements from a hierarchy of trial spaces \(S_j\subset H^1(\Upsilon)\), with \(S_j\) characterized by a mesh size \(2^{-j}\). The Lagrange multipliers will be approximated by elements from a hierarchy of trial spaces \(Q_\ell\subset L_2(\partial\Omega)\), with \(Q_\ell\) characterized by a mesh size \(2^{-\ell}\). The spaces \(Q_\ell\) need not be related to the spaces \(S_j\) and, in particular, need not be traces of \(S_j\) on \(\partial\Omega\). A major result in the paper is the proof that there is an integer \(L\) so that if \(j=\ell+L\) then the Ladyšenskaja-Babuška-Brezzi (LBB) condition holds uniformly for \(\ell\rightarrow\infty\). As part of the proof, the authors introduce a second multiresolution \(\widetilde{Q}_\ell\) of nested subspaces of \(L_2(\partial\Omega)\) that satisfy an inf-sup condition. The spaces \(\widetilde{Q}_\ell\) play no direct role in constructing the solutions, but judicious choice can improve the estimates needed for the LBB condition. Later in the paper, the spaces \(Q_\ell\) and \(\widetilde Q_\ell\) are constructed via dual wavelet bases. Following the general proof, the authors consider the case that the spaces \(S_j\) and \(Q_\ell\) are generated via wavelets. In this setting, the constants appearing in the LBB condition depend on properties of the wavelet representation of the trace operator and are not dependent on the mesh away from \(\partial\Omega\). The paper concludes with numerical solution of the Dirichlet problem for the Helmholtz equation \(\Delta u+u=1\) on \(\Omega\), where \(\Omega\) is the interior of a circle of radius \(R\) centered at the point \((1/2,1/2)\) in two dimensions. The fictitious domain \(\Upsilon\) is taken as \((0,1)\times(0,1)\). The LBB condition influences the matrix condition numbers of the Uzawa ``outer'' iterations and conjugate gradient ``inner'' iterations. The numbers of these iterations for various cases are used to infer deterioration of the LBB condition when \(\ell\) becomes equal to or larger than \(j\).