Start page of documentationDownload the codeDocumentation:
PDF, PS

Previous: EJIIM idea Next: Discretisation in a rectangle

Embedding into a rectangular domain

The original domain $ \Omega$ is embedded in a rectangle $ \Omega^*:=(a_x, b_x)\times(a_y, b_y)$. Parameters $ a_x$, $ b_x$, $ a_y$ and $ b_y$ are given in ASSEMBLE/problem_setup.m.

After embedding, the original domain $ \Omega$ becomes a `-' phase $ \Omega^-$ and the rest is typically denoted by `+' phase $ \Omega^+:=\Omega^*\backslash(\overline{\Omega^-})$. The boundary $ \partial\Omega$ becomes an interface and boundary conditions turn into jump conditions. The extension is done by zero in $ \Omega^+$, thus, on $ \partial\Omega^*$ homogeneous Dirichlet boundary conditions can be imposed.

Attention: The embedding rectangle has to be selected in such a way that there is ``enough'' space between $ \partial\Omega$ and $ \partial\Omega^*$. ``Enough'' depends on the mesh width used in discretisation. In any case, there should be at least several layers of grid points between $ \partial\Omega$ and $ \partial\Omega^*$.



V. Rutka, A. Wiegmann, 2005