[CIG-LONG] Initial temperature conditions

[Robert Gray]

Hey Walter,
The more realistic thermal gradient model with dtFactor=0.01 will still
produce unrealistically high temperatures if the model goes beyond ~2000
time steps.  I'm reluctant to decrease the dtFactor even further because of
the very long model run times, which are undesirable.  Also, the resolution
of the Eulerian grid doesn't seem to change the results much if the model
runs for a few hundred time steps.
Rob

On Fri, Feb 24, 2012 at 2:51 PM, Walter Landry <walter at geodynamics.org>wrote:

> Robert Gray <graywacke at gmail.com> wrote:
> >>> Robert Gray <graywacke at gmail.com> wrote:
> >>> > Hello everyone,
> >>> > I'm trying to set up an initial temperature profile in Gale that
> >>> increases
> >>> > steadily from 273 K to 1623 K in the top 120 km of the box, while
> >>> > everything below that is 1623 K.  To do this I've used the following
> >>> > condition:
> >>> >
> >>> >  "value": "(y>=0.48e6) ? -1.125e-2*y + 7023.0 : 1623.0"
> >>> >
> >>> > It works great for the first time step, but after that temperatures
> jump
> >>> > several orders of magnitude to unrealistic values.  That said, the
> >>> > incremental increase :
> >>> >
> >>> > "value": "(-1350.0/0.6e6)*y + 1623.0"
> >>> >
> >>> > works fine but does not produce the initial temperature profile I
> need.
> >>>
> >>> Can you send the input file to the mailing list?
>
> The resulting viscosity structures are very different.  For the linear
> temperature gradient, the mantle is still relatively high viscosity.
> So with the boundary conditions you have, pushing at the top right,
> you get strain dissipated throughout the mantle.
>
> For the more realistic temperature profile, the mantle is very low
> viscosity.  This allows the crust to slide along the top of the
> mantle.  Since there is also a low viscosity block in the center, the
> right side slides as a unit and you get strong vertical movement in
> the middle at the top.
>
> With the realistic temperature, you also have a very strong thermal
> gradient.  It turns out that the step size that Gale automatically
> computes is too large.  If you reduce the step size by a factor of 4
> by running it with the options
>
>  ./build/Gale wedgeThermal.json --dtFactor=0.25
>
> then the temperature in the next step is not so bad.  However, I think
> that there are "drunken seaman" problems after that.  To fix that, you
> will have to reduce the step size much farther.  I tried
> --dtFactor=0.01 and it seemed stable.  You should experiment to see
> how large you can make it.
>
> Cheers,
> Walter Landry
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://geodynamics.org/pipermail/cig-long/attachments/20120306/67c15e3e/attachment.htm