[aspect-devel] Free-Surface in 3d

[Ian Rose]

​Hi Avery,

So, when I change free-surface.prm to 3d things seem to be working okay.
Perhaps take another crack at that one? However, when I try to give
convection-box-3d.prm a free surface​ I can reproduce the error.  I think I
may have an idea of what is going on, which comes down to the free surface
stabilization term and the conditioning of the Stokes matrix.

If we consider the typical size of entries in the normal system matrix, we
get something like [image: \eta L], where [image: \eta] is the viscosity
and [image: L] is a lengthscale for the cells.  For free surface problems,
the surface should be stabilized.  The stabilization term is typically of
order [image: \rho g \Delta t L^2].  The ratio of the size of the
stabilization term to the normal matrix elements is then [image: \rho g L
\Delta t / \eta]. For most problems of geophysical relevance this term will
be on the small (but not negligible, since we need it for stabilization!)
side, something like [image: 10^{-1} - 10^{-2}],

However, the convection-box-3d parameter file has some odd values which
change the story.  In particular, the value for gravity is large ([image:
10^{16}]!!) and the value for viscosity is small (1).  So the stabilization
term absolutely dominates the matrix entries by around nine or ten orders
of magnitude.  You can imagine that that would be bad.  I would note that
the unusual values for the convection-box parameter files have caused
problems in the past, see, for example issue #94.  Perhaps Wolfgang or Timo
would have some further insight here.

Okay, so the TL;DR version of this is (1) try free-surface.prm again, I
think it may work, and (2) try convection-box-3d.prm with more reasonable
model parameters.

Hope this helps!
Ian


On Tue, Jul 21, 2015 at 10:47 AM, Avery Noonan <
avery.noonan at mail.utoronto.ca> wrote:

>
>  Hello Ian,
>
> Thank you for the quick response.  I have managed to run models in 2d that
> are identical to the ones I am trying to run in 3d (except, of course, for
> the necessary changes in order to incorporate a third dimension).
>
>  I have attached a screen shot of the exact output I am given
> corresponding to the error.  I am told that the stokes solver does not
> converge.
>
> The input file I was using in this case, was the "convection-box-3d.prm"
> cookbook with the minimum number of changes required to set a free-surface
> (pressure normalization = no / free surface boundary = top).  I also took
> the 2-dimensional "free-surface.prm" cookbook and made the minimum required
> changes to make it 3-dimensional and was confronted with a near identical
> error.
>
>  I will definitely take a look at the crustal deformation cookbook you
> mentioned.  Hopefully I will manage to find the solution to my problem
> there.
>
> Thanks again,
>
> Avery
>
>  ------------------------------
> *From:* Aspect-devel <aspect-devel-bounces at geodynamics.org> on behalf of
> Ian Rose <ian.rose at berkeley.edu>
> *Sent:* July 21, 2015 1:18 PM
> *To:* aspect-devel
> *Subject:* Re: [aspect-devel] Free-Surface in 3d
>
>   Hi Avery,
>
>  Could you be a little more specific about what is failing in your
> model?  What is the input file you are using, and what does the output look
> like? If you try it in 2D, do things appear to be working correctly? In
> general, it would be useful for us to have a minimal example that
> reproduces your problem.
>
>  There is a crustal deformation cookbook that is in 3D (contributed by
> Cedric Thieulot and Anne Glerum) which you could run running to verify
> things are working. At the moment, the material model on which it relies is
> waiting to be merged, but you could try checking that out.
>
>  Best,
>  Ian
>
> On Tue, Jul 21, 2015 at 9:22 AM, Avery Noonan <
> avery.noonan at mail.utoronto.ca> wrote:
>
>>  Hello,
>>
>>
>>  My name is Avery Noonan and I was awarded a research grant to work in
>> Russ Pysklywec's lab at the University of Toronto.
>>
>> I am attempting to model the tectonic environment around/beneath
>> New-Zealand, in order to see the extent to which mantle convection can
>> explain the topographical low (considering the densities of the continental
>> crust, oceanic crust and mantle).
>>
>> To do this, I, of course, need a free surface on the upper boundary of my
>> model.  I have not been successful in running a 3d model with a
>> free-surface, despite variations to the solver tolerances and a CFL number
>> as small as 0.005.
>>
>>
>>  I was wondering if anyone else had been successful in running 3d models
>> with a free-surface and whether they might have some advice for me.
>>
>> Thank you very much,
>>
>> Avery
>>
>> _______________________________________________
>> Aspect-devel mailing list
>> Aspect-devel at geodynamics.org
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20150722/fedd27ea/attachment-0001.html>