meaning of a "integ" boundary condition in dedalus examples

[Upasana Das]

Hello Everyone,

In the 1d Rayleigh-Benard example script under EVP in Dedalus examples, can you tell me what the following boundary condition means physically, and when can 

I use a such a boundary condition?

problem.add_bc("integ(p, 'z') = 0")

I am attaching the script here for everyone's reference.

Thanks & Regards,
Upasana

[Daniel Lecoanet]

integral of p over z direction is zero

Daniel

--
You received this message because you are subscribed to the Google Groups "Dedalus Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dedalus-users+unsubscribe@googlegroups.com.
To post to this group, send email to dedalu...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dedalus-users/50ad8c84-b06a-4e66-a0fe-018ebe41aac5%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Upasana Das]

Thanks Daniel. I wanted to clarify that this condition has nothing to do with incompressibility right?

Upasana

On Thursday, 19 April 2018 18:51:39 UTC-6, Daniel Lecoanet wrote:

integral of p over z direction is zero

Daniel

On Fri, Apr 20, 2018 at 2:34 AM, Upasana Das <upasa...@gmail.com> wrote:

Hello Everyone,

In the 1d Rayleigh-Benard example script under EVP in Dedalus examples, can you tell me what the following boundary condition means physically, and when can 

I use a such a boundary condition?

problem.add_bc("integ(p, 'z') = 0")

I am attaching the script here for everyone's reference.

Thanks & Regards,
Upasana

--
You received this message because you are subscribed to the Google Groups "Dedalus Users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to dedalus-user...@googlegroups.com.

[Keaton Burns]

Hi Upasana,

It’s actually a little complicated here.  In some simulations, you’ll see this BC applied to the horizontal-mean Fourier mode (kx=0).  In that case, it’s setting the pressure gauge — the overall constant pressure, which is dynamically insignificant in incompressible simulations.

This case is different — for the eigenvalue problem, each variable takes the form f(x)e^{ikx}, where the x-dependence is implied by the substitution defining the x-derivative.  So this condition isn’t setting the pressure gauge, it’s actually a boundary condition that you can show is equivalent to setting w=0 on the other wall for the linear problem.

I think this is an outdated version of the example script though — the current version should include Dirichlet preconditioning and utilize the sparse eigenvalue solver, which is much faster when you’re only interested in finding a subset of the eigenmodes.

Best,

-Keaton

To view this discussion on the web visit https://groups.google.com/d/msgid/dedalus-users/2bd12a68-8151-406e-a355-418a99a2757c%40googlegroups.com.