Error with radially-defined NCC terms on a spherical shell basis

[Austin Taylor]

Hi,

I'm currently trying to reproduce the Anufriev et al 2005 paper on liquid anelastic appx geodynamos, but I'm being thrown an error while the matrices are being constructed that Azimuthal NCCs are not allowed.

The thing is, I shouldn't have any azimuthal NCCs in the first place. All of my NCCs are constructed on shell.radial_basis, and derivatives involving them are calculated explicitly.

There's definitely still debugging work in the other equations, but I've been constantly getting errors from my momentum equation about azimuthal NCCs. 

I've attached my python script below and a screenshot of the error, any insight would be great.

[J. S. Oishi]

Hi Austin,

This looks like an azimuthal NCC to me:

2*cross(ez,u)

since 

ez = dist.VectorField(coords, bases=space.meridional_basis) 

ez['g'][1] = -sin(theta) 

ez['g'][2] = cos(theta) 

Maybe try moving your rotation term to the RHS to see.

Jeff

[J. S. Oishi]

Oh dear. Clearly, that is a colatitude NCC, not an azimuthal one. Apologies for answering before being fully awake...

[Austin Taylor]

I managed to fix my code. I stripped everything down to a bare-bones Boussinesq model and then added in NCCs and terms slowly.
The matrices build fine, I was able to keep a lot more stuff on the LHS, and after tuning down some of the parameters to match the resolution I've avoided blowup.

It seems a lot of the problem was with Id in my stress tensor when it was dotted with er, causing the azimuthal NCCs not supported error.
I did have to move the coriolis term back over to the RHS due to coupling of r and theta.