Monotone multigrid methods for elliptic variational inequalities. I (Q1347056)

The following elliptic optimization problem is considered: \[ u\in K: J(u)\leq J(v),\quad \forall v\in K,\tag{1} \] where \(K= \{v\in H^ 1_ 0(\Omega)\mid v(x)\leq \varphi(x)\) a.e. in \(\Omega\}\), \(J(v)= {1\over 2} \int_ \Omega \sum^ 2_{i,j= 1} a_{ij} \partial_ i v\partial_ j vdx- \ell(v)\), \(\ell\in H^{-1}(\Omega)\) and \(\Omega\) is a polygonal domain in \(\mathbb{R}^ 2\). With the using of piecewise linear finite elements the problem (1) is reduced to a discrete elliptic variational inequality. Let a regular partition of \(\Omega\) in triangles with minimal diameter of order \(2^{-j}\) be given. Using ideas of successive subspace correction, the author modifies well-known relaxation methods by extending the set of search directions. On this basis globally convergent methods for discrete analogs of (1) are derived. It is proved that the proposed methods have asymptotic convergence rates of order \(1- O(j^{-3})\). The efficiency of the methods is illustrated by a number of numerical examples.