DomainDecompositionParametersΒΆ
| Type: | section |
|---|---|
| Appearance: | optional |
The section is used to steer the iterative domain decomposition solver. This solver is applied when the geometry is split up into sub-domains. The sub-domain problems are coupled iteratively to converge to the solution for the entire domain.
For example, a domain decomposition approach can be applied in the presence of layered media definition, c.f., LayeredMedia. In this case the layered media stack is treated analytically and coupled to the numerical solution in the actual computational domain.
A decomposition of the computational domain may also be used when computing a large 3D problem. Then the geometry is split automatically into smaller sub-domains to reduce the memory demands of the solver. The automatic 3D domain decomposition solver is currently implemented only for geometries which are periodic in the 
-plane, with an extruded height profile in the 
-direction (see JCMgeo, Extrusion).
We give a (very) rough sketch on how the the domain decomposition solver works. This is necessary to understand the role of the numerical tuning parameters:
As the initial step the solution on each sub-domain is computed separately with impressed currents restricted to the domain and with the prescribed illumination. The gives rise to secondary, scattered fields leaving the separated domains.
Now, the scattered fields of the sub-domains are coupled into neighboring sub-domains. This give again rise the scattered fields and the process is repeated until a specified maximal number of steps has been reached, or until a specific accuracy has been reached.