[WIP] Fix density used for thermal conductivity in visco_plastic and diffusion_dislocation models

[naliboff]

In the visco_plastic and diffusion_dislocation models the thermal diffusivity is specified as an input file parameter instead of the thermal conductivity. The thermal conductivity is then calculated from the thermal diffusivity, density and heat capacity.

This leads to an inconsistency when set Temperature equation = reference density profile.

The same inconsistency is present in any heating model that uses the material model density output. @gassmoeller @jdannberg - I was planning to use the structure to enforce that the heating models use the correct density. Any objections before I do this in another commit?

Thanks and credit to Phillip Heron and Jeroen van Hunen for identifying an issue by running the Blankenbach benchmarks with these two material models!

[jdannberg]

The same inconsistency is present in any heating model that uses the material model density output.

This should already be implemented correctly. If you look in assembly.cc, around line 859, if the formulation for the temperature equation is reference_density_profile, then the heating models get the reference density as an input instead of the real density.

[jdannberg]

Thanks for fixing this! Your changes look good, but I've found a few more places where the diffusivity is computed, and it would be great if you could fix those too (otherwise, let me know and I'll merge this pull request and we can open another pull request for these new fixes):

• the visualization postprocessor thermal_diffusivity.cc
• in the basic_statistics postprocessor in line 75
• in the initial temperature plugin adiabatic (line 92)

If we want to be really consistent, we could also use the reference density when computing the conduction time step, but I think we can leave that for now.

The changes in these other files would probably be very similar to the ones you did here, so I hope that's not too much work.

[naliboff]

@jdannberg - Thanks of for the review and glad that the heating model formulation is already implemented correctly. Happy to fix the diffusivity in the other files you listed as well ... seems logical to do this in the same PR.

[naliboff]

@jdannberg - I've fixed the diffusivity in all the places above. Let me know if you would like further documentation or reformatting in these sections.

[jdannberg]

Looks good, thank you!