[aspect-devel] Geometry dimensions affect convergence?

[Eric Heien]

Hi everyone,

I'm working with a student here to test Aspect on some simple geophysics problems.  We've noticed that solver convergence is sensitive to the dimensions of the model and for some models the size of the geometry will determine whether the solver converges or not.  I've attached two parameter files - the good one runs to completion while the bad one fails to converge on the first step - note that parameters other than dimension are also changed but everything is still within ranges expected of an Earthlike mantle.  Can anyone give suggestions about why this happens and what we can do to fix it?

Thanks,

-Eric

-------------- next part --------------
A non-text attachment was scrubbed...
Name: parameters_bad.prm
Type: application/octet-stream
Size: 37921 bytes
Desc: not available
Url : http://geodynamics.org/pipermail/aspect-devel/attachments/20121105/f0b06440/attachment-0002.obj 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: parameters_good.prm
Type: application/octet-stream
Size: 37928 bytes
Desc: not available
Url : http://geodynamics.org/pipermail/aspect-devel/attachments/20121105/f0b06440/attachment-0003.obj