Semigrid

Semigrid is a term used in computational science to describe a grid approach that combines elements of structured and unstructured meshing. In a semigrid, the computational domain is partitioned into multiple blocks that are internally structured, such as Cartesian layouts, while the connections between blocks are arranged in an overall unstructured manner. This arrangement aims to balance the efficiency of structured grids with the geometric flexibility often required by complex domains.

Form and generation. Each block typically contains a regular, grid-like arrangement of cells or nodes, which facilitates fast data access and straightforward stencil operations. The blocks are linked at interfaces to ensure continuity of the solution across the domain. Interfaces may involve nonuniform alignments or alternative connectivity, and grid quality is monitored using standard metrics to maintain stability and accuracy.

Applications. Semigrid concepts appear in fields such as computational fluid dynamics, heat transfer, and electromagnetics, where problems involve intricate boundaries that benefit from localized refinement within blocks without sacrificing the efficiency of structured computations. It is related to, and sometimes implemented alongside, multi-block structured grids, block-structured grids, and overset or Chimera grid techniques.

Advantages and limitations. The semigrid approach offers improved data locality and easier implementation of solvers within blocks, while still accommodating complex geometries through flexible inter-block interfaces. Challenges can include ensuring high-quality interfaces between blocks, managing grid conformity, and implementing refinement or adaptation across block boundaries.

See also. Structured grids, unstructured grids, multi-block grids, block-structured grids, overset/Chimera grids.