[CIG-LONG] Oscillating velocity at bottom free boundary

[Matthias Schmiddunser]

Hi all,

I try to model a convergent crust with isostatic compensation of the 
thickening slab and encounter numerical instabilities along the bottom 
boundary (see attached .png).
I get very high velocity values, oscillating with the grid spacing. When 
using a constant viscosity, this doesn't happen, but with all 
temperature-dependent viscosities I tried (FrankKamnetskii, Arrhenius 
and NonNewtonian), the same error appears.
I tried increasing the mesh resolution, reducing the timestep with 
dtFactor and a different scaling of my input values without much 
success. The only thing that helped so far was using the NonNewtonian 
viscosity and limiting the viscosity range, but this only reduces the 
oscillations and doesn't get rid of them.
The temperature profile and viscosity function I adopted from another model.

Do you have an idea how to get rid of this instability?

I attached my input file and the scaling for the input file I used. The 
thermal configuration is chosen such that the initial thermal field is 
preserved in case of no motion and for isostasy a supporting material of 
higher density (3300 kg/m^3) is chosen.

Thank you very much,

Matthias

-- 
Matthias Schmiddunser
Department of Geosciences
University of Tübingen
Wilhelmstrasse 56
D-72074 Tübingen

E-Mail: matthias.schmiddunser at uni-tuebingen.de
Telephone: +49(7071)29-73157
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Input_oscillations.json
Type: application/json
Size: 5333 bytes
Desc: not available
URL: <http://geodynamics.org/pipermail/cig-long/attachments/20131021/55919c76/attachment-0001.json>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Velocity_Oscillation.png
Type: image/png
Size: 320130 bytes
Desc: not available
URL: <http://geodynamics.org/pipermail/cig-long/attachments/20131021/55919c76/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Scaling.xls
Type: application/vnd.ms-excel
Size: 36352 bytes
Desc: not available
URL: <http://geodynamics.org/pipermail/cig-long/attachments/20131021/55919c76/attachment-0001.xls>