[aspect-devel] Internal heating in aspect (Ludovic Jeanniot)

[Timo Heister]

> Does anyone know if the "artificial viscosity" of SUPG also scales with h^2? Because if it is linear, we might implement something that is better for low resolutions, but worse for high resolutions.

That depends on the size of convection and conduction. The parameter
design is actually quite complicated, see
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_geodynamics_aspect_pull_412_files-23diff-2D08f9a13515fd055c1c636c3a242b8da6R1641&d=DwIBaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=c08Btfq4m9QEScXN3ZQwLZzzWQE7S8CYq1IYuzKV_Zk&m=WueBsp2IVsSkOAJhUeiQWoucfiYfoTF0vy8BDAizrZA&s=rvzH95BkonjPPF16Nehb3nr4neH2_blQ8omWO1CyV8U&e=
for one possible strategy. I think it is linear in h for small
convection and conduction (which we are in here?), but I would need to
look into the literature again.

> - As Max mentioned: We should not need a stabilization for pure conduction problems (where velocity is 0), and should modify the algorithm accordingly.

We have this in there for a reason. I think one issue is the solution
in timestep 0 where velocity is 0 but we still need to apply
stabilization?

-- 
Timo Heister
http://www.math.clemson.edu/~heister/