[aspect-devel] Aspect-devel Digest, Vol 12, Issue 3

[John Naliboff]

Hi Wolfgang,

Happy Thanksgiving!

That was Ian who sent that input file over.  The spherical shell models I tested actually all crashed a little bit after I last emailed you, with all of them reporting very high (> 1e20) residuals at there final time step.  The crash occurred over a range (~ 10^8 - 10^4).

I'm a bit surprised at the high resolutions, considering my mesh setup: 4 initial global refinements, 3 initial adaptive refinements, refinement fraction of 0.3 and coarsening fraction of 0.05.  I'll try modifying the mesh to see if that helps with convergence.

In terms of resolving basal verse surface heat flux, would it be difficult to include a parameter that modifies the spatial resolution as a function of depth, in addition to temperature, cp, etc?

Cheers,
John




On Nov 19, 2012, at 3:27 PM, Wolfgang Bangerth wrote:

> 
> All,
> I know it's terrible -- I don't recall who exactly sent me this input file (attached) and whether I modified it, but in any case I ran it over the weekend to see what happens to the boundary fluxes in the shell geometry. Attached is what I get from simply running this on 16 processors for a *very* long time.
> 
> The two heat fluxes never reach a steady state -- rather, the jump up and down every time a plume detaches from the inner or outer boundaries. Looking at the solution picture (also attached), I'm not surprised that the heat fluxes don't coincide: it looks like the boundary layers are less than one cell width, and so we can not expect to resolve them accurately. Rather, the heat flux given by these numerical approximations is determined by the smallest length scale that can be resolved by the mesh -- and because this length scale is different for the inner and outer boundaries, the heat fluxes is different. (Of course, had we chosen a mesh where the radial resolution is the same at the two boundaries, we may have gotten results where the two heat fluxes match -- but are both completely wrong.)
> 
> Best
> W.
> 
> -- 
> ------------------------------------------------------------------------
> Wolfgang Bangerth               email:            bangerth at math.tamu.edu
>                                www: http://www.math.tamu.edu/~bangerth/
> 
> <test_flux.prm><heatflux.png><visit0001.png>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://geodynamics.org/pipermail/aspect-devel/attachments/20121122/58261488/attachment.htm