[aspect-devel] Refinement Strategy

[Timo Heister]

Hi Thomas,

> i never looked at rho*c_p*temperature
> what do you mean by "resolve it"

Well, if this product can not be represented with a small enough error
by the finite element space, cells will get refined. This is done by
looking at an approximation of the gradient. This way we see steep
gradients of the temperature and jumps in the density. If both things
happen at the same place (plume hitting a density gradient), it is
seen as more important than just a density jump. You will see, that
the mesh does not keep the density jump on the maximum refinement
level at all times (which is also not necessary if nothing is
happening there if you think about it).

> i will test a few of these options tomorrow.

Thanks.

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