Accepted geometrical mapping for list-based post-processing
After a great success for the field mapping in list-based post-processing data, I experiment with the geometrical mapping.
According to the mappings on page 3 of the paper Efficient visualization of high-order finite elements, the geometrical approximation is as general as the field approximation (inline equation at the bottom of page 2).
However, I noticed that its implementation in Gmsh is less general. Could you please confirm that my observations below are correct? When calling setInterpolationMatrices
,
- the reference element is fixed, implying that
- the domain of the reference element and the local node ordering are given by the Node ordering section of the documentation
- the
coefGeo
array must contain the terms that correspond to the node ordering
- you are restricted to "standard" elements. E.g. you cannot define quadratic shape functions for a two-node segment:
\psi_1 = -\frac14 \xi^2 - \frac12 \xi + \frac34, \qquad \psi_2 = \frac14 \xi^2 + \frac12 \xi + \frac14, \qquad \xi \in [-1, 1]
Edited by Zoltan Csati