[aspect-devel] cookbooks/composition-active.prm

[Thomas Geenen]

he Timo,

this is the result of higher order basis functions no? with sharp contrasts
this typically happens when the high value in the nodal point is multiplied
with the local negative value of the basis function.
we could chop up the elements in linear elements and interpolate to the
integration points with linear basis functions of these smaller elements.

of course this should be an option to the user i guess.
this might also allow for larger elements in the thermal boundary layer
(prevent negative temperatures there).

cheers
Thomas


On Mon, Nov 4, 2013 at 5:32 PM, Timo Heister <heister at clemson.edu> wrote:

> Hey,
>
> it turns out cookbooks/composition-active.prm broke a long time ago
> (r1370, more or less one year ago). Today I finally managed to take a
> look at it. The change modified how the entropy viscosity is computed:
> since then we are computing and using max_density and
> max_specific_heat on each cell. Before, the entropy viscosity was at
> the very end multiplied by the local density and specific_heat. This
> change makes a lot of sense because it fixed some inconsistencies with
> respect to the scaling with density and specific_heat.
>
> Now to the problem. It turns out that the density becomes negative in
> some quadrature points, while the max_density is still a positive
> number. Example:
>
> max dens = 69.7217
> loop 69.7217
> loop 69.7217
> loop 69.7217
> loop 0.992031
> loop 0.992031
> loop 0.992031
> loop -7.73768
> loop -7.73768
> loop -7.73768
>
> The density becomes negative, because the composition has a value of
> about -0.08 here.
>
> Suggestions:
> 1. Density should always be positive and we should enforce this in
> debug mode with an Assert. What to do in release mode? Should we
> restrict the density to something sensible?
> 2. The material model 'simple' should simply clamp the compositional
> fields to [0,1].
> 3. If the user specifies that a compositional field is supposed to be
> in [0,1], should we postprocess the field and restrict it to [0,1]? I
> think not, because we would lose mass this way (we could restrict it
> only for output and material models but use the original one in the
> next timestep).
>
> Opinions?
>
> --
> Timo Heister
>
> https://urldefense.proofpoint.com/v1/url?u=http://www.math.clemson.edu/~heister/&k=dsHy%2FVymniCD0osh6tze%2Fw%3D%3D%0A&r=WLGUy1n%2F1Vzsqown7KFPcAEzR5DIdbP771jYw4Spy%2Bk%3D%0A&m=S11NmahkaA8sPelibOGff%2FjUvWqX3xTwDWFpjCX9Hko%3D%0A&s=1a44e5483f646199cc1b0d3d4f2c55799e3902613ce8c863e9bc3b4350fad75c
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://geodynamics.org/pipermail/aspect-devel/attachments/20131104/377a735a/attachment.html>