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

[Scott King]

Hi;

Looking at Max’s results, could there be a 1/r^n or r^n term missing in the artificial diffusivity calculation?   My impression is the 2D Cartesian cases are spot on.  If you look at my 2009 G^3 paper, ConMan results continue to evolve noticeably even up to 200x200 elements, so I am not as concerned about the level of refinement needed in Cartesian geometry results.  I know of no magic bullet to avoid resolution if you want accurate results.  I felt the 2D Blankenbach problems in the cookbook were excellent (and we repeated them to be sure something had not gone wrong with our installation and got the same results) and what this suggested to us is that something was wrong with the spherical geometry.  We even did the cylindrical results in the cookbook to see if that aspect of curvature was o.k.  I recall this behaved somewhat better than the sphere but with frightening need for resolution.

I’ll keep trying to use the aspect-devel list.   I’ve never actually seen one of my own posts (maybe that’s a feature of the listserver setting) and it seems like some of other peoples posts may be getting filtered out.    Or it could be a feature of VT’s e-mail filtering (which sometimes marks VT e-mail sent to VT people as spam).

We are currently running A1 and C1 on a number of grids with beta=cR=0.  We should know today (I hope) and I’ll pass along results.

Chaos abounds in the king household and the Lyapunov exponent is greater than one.

Cedric, we have all the CitcomS results for A1-9 and C1-4 on various grids.  We used a uniform mesh but that didn’t change much.   Happy to share, happy to collaborate with any and all as long as we include Grant and Shangxin ‘cuz they did most of the work.

Scott

> On Aug 28, 2018, at 11:01 PM, Wolfgang Bangerth <bangerth at colostate.edu> wrote:
> 
> On 08/28/2018 05:33 PM, Max Rudolph wrote:
>> From this, it is very obvious why the solution to the convection problem at low resolution is very diffusive and also why the interior temperature is much closer to the surface temperature than to the CMB temperature because the artificial viscosity is on the order of 20 times larger than the thermal conductivity near the surface.
> 
> Would it be easy to verify whether the artificial viscosity ("artificial conductivity") decreases at the expected rate with mesh refinement?
> 
> 
>> For the conduction problem, the default values of the artificial viscosity are also much larger than the thermal conductivity.
> 
> I think that's the point worth investigating. Since in this case the velocity is zero, one would expect the artificial viscosity to also be at least quite small. Why is it not?
> 
> Best
> W.
> 
> -- 
> ------------------------------------------------------------------------
> Wolfgang Bangerth          email:                 bangerth at colostate.edu
>                           www: http://www.math.colostate.edu/~bangerth/
>