[aspect-devel] Negative pressure
FELIPE ORELLANA ROVIROSA
f_orellana at berkeley.edu
Fri Mar 11 13:11:51 PST 2016
Hi Lev,
I have gotten negative pressures too. This is all related to what
Wolfgang explained us. Other softwares do this too.
I think Julianne Dannberg made a key point for your results. I would
like to further comment:
The first pressure figure you provided seems physically plausible to
me, at least from a qualitative point of view: with pressure normalization
at the surface (where you have a spike), you might get something like that.
In fact, your pressure field exhibits uncountable orders of magnitude of
variation (goes from - to +), but that's not a problem for relative
pressure. Imagine that the total pressure were to be found by adding 10^10
units, in that case you would get a pressure range from 3.5e9 to 7.2e10
approx.
The overall pressure field variations seem physically plausible: the
pressure increases with depth inside the asthenospheric mantle, that's ok
for downward gravity. And the pressure has a maximum at the lithosphere due
to horizontal convergence. Think that given a stress sij
p = sum(sii) /3
And because your viscosities are so high at the lithosphere, the
horizontal convergence implies high compressive stress sxx, giving rise to
a high pressure, larger than the deep pressures.
I would just try to find (i) what is the reasonable pressure constant
you need to add, and (ii) whether the viscosities give rise to such large
horizontal stress by convergence considering the velocities you prescribed.
For your second figure (negative presure2), I see 10 orders of
magnitude variation. The pressure field looks very odd and it might have be
an error in the overall calculation, I would advice you to start over.
cheers, good luck,
Felipe
On Fri, Mar 11, 2016 at 1:28 AM, Lev Karatun <lev.karatun at gmail.com> wrote:
> Hi Max,
>
> the viscosity plot is attached. The box dimensions are 512*1024km, the
> rest is in the .prm file attached to the previous email.
>
> Best regards,
> Lev Karatun.
>
> 2016-03-10 15:34 GMT-05:00 Max Rudolph <maxwellr at gmail.com>:
>
>> Lev,
>> Can you post plots that show viscosity? It would also be useful since
>> this is a dimensional calculation to show us the dimensions of the box.
>>
>> Max
>>
>> On Thu, Mar 10, 2016 at 10:00 AM, John Naliboff <jbnaliboff at ucdavis.edu>
>> wrote:
>>
>>> Hi Lev,
>>>
>>> Following from Julianne’s points below, it would be quite helpful if you
>>> could make separate plots of hydrostatic vs dynamic pressure and provide a
>>> bit more detail about the model (bc, material model, etc).
>>>
>>> Cheers,
>>> John
>>>
>>> > On Mar 10, 2016, at 9:26 AM, Juliane Dannberg <dannberg at gfz-potsdam.de>
>>> wrote:
>>> >
>>> > Hi Lev,
>>> >
>>> > I see your point that the pressure is positive at the top of the model
>>> and then decreases with depth, which normally shouldn't be the case.
>>> > But just from seeing the pictures it is difficult for us to find out
>>> what the problem is.
>>> >
>>> > If your gravity is positive (which I assume it is), other reasons for
>>> negative pressures I sometimes see in my models are prescribing velocities
>>> at the boundaries. For example, if you prescribe convergent velocities at
>>> the top boundaries, there is a point somewhere in the middle of the top of
>>> the domain, where velocities point inwards from both sides, and so you get
>>> a very high spike in dynamic pressure in this place. It looks like this
>>> could be the case in your model. If you then normalize your pressure with
>>> the values at the surface, they might become negative in a layer below.
>>> >
>>> > How does your pressure gradient look like? Is that basically density *
>>> gravity once you are a few cells away from the top, or is it different? If
>>> you find that the problem is only because of prescribed velocities at the
>>> surface, you can just use a different value for the surface pressure, one
>>> that you think is reasonable for your model.
>>> >
>>> > Another point to think about is the inflow: is the sum of your in- and
>>> outflow zero?
>>> >
>>> > Best,
>>> > Juliane
>>> >
>>> >
>>> > On 03/10/2016 09:14 AM, Wolfgang Bangerth wrote:
>>> >> On 03/09/2016 11:45 PM, Lev Karatun wrote:
>>> >>>
>>> >>> thank you for the quick reply. The pressure normalization was
>>> actually
>>> >>> set to "no". I tried changing it to "surface", but it made made it so
>>> >>> that the pressure across the entire model domain except for the very
>>> lop
>>> >>> layer became negative =(
>>> >>
>>> >> But the point remains true: the Stokes equations only determine the
>>> pressure up to a constant. If you want to add 100 GPa to the pressure
>>> everywhere, it will still solve the equations. In other words, whether the
>>> pressure is negative or positive matters from a physical perspective, but
>>> has no mathematical meaning in the context of the equations you are solving
>>> because you can make the pressure positive everywhere or negative
>>> everywhere by just adding a constant. Mathematically, what matters are only
>>> pressure *differences*, not the overall pressure.
>>> >>
>>> >> Best
>>> >> W.
>>> >>
>>> >
>>> > _______________________________________________
>>> > Aspect-devel mailing list
>>> > Aspect-devel at geodynamics.org
>>> > http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>>
>>> _______________________________________________
>>> Aspect-devel mailing list
>>> Aspect-devel at geodynamics.org
>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>>
>>
>>
>> _______________________________________________
>> Aspect-devel mailing list
>> Aspect-devel at geodynamics.org
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>
>
>
> _______________________________________________
> 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/20160311/5a475a51/attachment.html>
More information about the Aspect-devel
mailing list