DG fields in Paraview

[Mark]

I am working on some cohesive element formulation, relying heavily on the ‘Fenicsx Tours’, Cohesive zone modeling restricted to an interface — Computational Mechanics Numerical Tours with FEniCSx

My script is running, but I am totally unable to visualise the results in Paraview. Referring to the script czm_interface_only.py, it saves a czm.pvd file, plenty of .vtu files, as well as pvtu.

I can open the czm.pvd, but completely unable to find a field to ‘Warp by Vector’ by, to see the interface opening.

I appreciate this is a Paraview-centered question, but I hope somebody can help, as it also relates to how Dolfinx outputs DG fields.

Thanks

[Stein]

Have you taken a look at the procedure in Divergence conforming discontinuous Galerkin method for the Navier–Stokes equations # noqa — DOLFINx 0.10.0.post5 documentation ? I.e, export with VTXWriter. That DG export loads into paraview without issues, and allows for warp by vector.

[dokken]

VTKFIle should be able to store DG values, as illustrated with the following MWE:

from mpi4py import MPI
import dolfinx

mesh = dolfinx.mesh.create_unit_square(MPI.COMM_WORLD, 10, 10)
V = dolfinx.fem.functionspace(mesh, ("DG", 1, (mesh.geometry.dim,)))
u = dolfinx.fem.Function(V)

right_cells = dolfinx.mesh.locate_entities(mesh, mesh.topology.dim, lambda x: x[0]>0.5 - 1e-14)
u.interpolate(lambda x: (x[0], x[1]), cells0=right_cells)

with dolfinx.io.VTKFile(mesh.comm, "u.pvd", "w") as vtk:
    vtk.write_function(u)

yielding:

where the wireframe grid is the unwarped grid, while the solid bits is the warped version.

[Mark]

Thanks to all, much appreciated

[Stein]

Questions on different output formats (and their restrictions) pop up every now and then. Do you think you could give a short run down on the options and best practices?

[dokken]

Things wrt IO never seem to stabilize, there is always the N+1 formats. Here is a rundown as of Feb 2026:

• XDMF: currently the preferred format for reading in meshes in parallel, as it uses hdf5 to store heavy data. However, the format has not been maintained since ~2021, and supports limited visualization of higher order fields/DG fields.
• VTX (.bp): Scalable output format similar to xdmf, but uses adios2 for the IO. Not good for reading in meshes, ref: VTK support for unstructured grids with global arrays. · Issue #3688 · ornladios/ADIOS2 · GitHub. Supports higher order visualization of continuous and discontinuous fields.
• VTK (.pvd): store data in many files making it tediuous for large problems and reading in data. Supports the same visualization as VTX.
• VTKHDF (.vtkhdf): the «new» format from kitware. Uses HDF5 with a specific ordering to be possible to visualize in Paraview. Hopefully the future format. Slightly less flexible with respect to multiple grids with different time stepping than xdmf, but supports higher order fields. An initial implementation exists in DOLFINx with a more complex reader/writer (including meshtags) at: Support writing time dependent meshes to vtkhdf by jorgensd · Pull Request #199 · jorgensd/adios4dolfinx · GitHub
• Exodus: old-school format that has implemented some new abstractions in VTK: Discontinuous Galerkin elements and other novel cell-types/function-spaces - Development - VTK . We don’t have a writer for this, as we hope that the DGCell will be implemented in VTKHDF.

The latest update of vtkhdf from kitware is at https://www.kitware.com/vtkhdf-file-format-2025-status-update/