Stokes and 2nd order elements

[fgillet]

Dear all,

attached a small test case of the Stokes Solver with second order elements;
There is a sharp discontinuity on the bottom BC orientation (but this is common in ice flow modelling and real applications).
This results in noisy results when using the "Bubbles" stabilisation flag (as shown on the plot) but it is ok with "Stabilized".

I get these noisy results when the discontinuity fall on a middle node of a 408 element;
It is ok also with Bubbles if this discontinuity is on a side-node (for example you can try with "Element Divisions 1 = 200" in the grd file).

By the way, my main motivation to go to 2nd order elements is that I need a good estimate of the second derivatives of my velocities.
In principle It should be possible to write a Solver to do this estimation also with linear elements. Do you have any experience on this? would it be robust?

Thanks

Fabien

[Juha]

Hi,

i think the "FluxSolver" can compute nodal fitted first derivatives for you,
from which you could directly compute the second derivatives, whether
using 1st or nth order elements. Don't really know any better way anyway...

Don't have an immediate answer for the bubbles misbehaviour when using
2nd order elements, will have to investigate.

Regards, Juha

[fgillet]

Hi Juha,

Thanks, I will try the flux Solver;

I forget to say that p2/p1 does the same as Bubbles.

Cheers

Fabien