Issues with cyclone base case with zero magnetic shear

[Patrick Kim]

Hi, I’ve been trying to run (on stellar) nonlinear simulations of the cyclone base case with zero magnetic shear and periodic boundary conditions, but have been running into problems as the resulting heat flux doesn’t seem to converge. I’ve tried adjusting several parameters (box size, resolution, hyper) without much change. I was wondering if there could be any bugs that could be causing this problem? I’ve attached the input file and a plot of the heat flux. Thanks.

[David Dickinson]

Which version of the code are your using?

‌

Have you looked at the fluctuation spectrum to see where in wavenumber space things are growing? Do all the cases you’ve tried look the same? Whilst the heat flux reported does appear rather large it appears that it may be slowing down and could start to drop, have you tried running for longer?

‌

Your initial amplitude (phiinit) is maybe a little larger than recommended for nonlinear runs. Perhaps try 1e-3? I think you could probably get away with a larger time step which might help when you’re testing things out.

‌

Do you still see a failure to converge without the hyper terms and without hypermult in collisions?

‌

[Patrick Kim]

I’m using version 8.0.6.

I’ve tried reducing the initial amplitude but that did not fix the issue. Not using hyperparameters makes the convergence worse. To better explain what’s going on, I’ve attached a newer plot of the heat flux. There’s multiple periods of time where the flux grows rapidly. In each of these times, looking at the electrostatic potential plot as a function of the kx and ky modes, a new dominant mode grows that has a lower ky than the previous dominant mode. All the modes are centered around kx=0. One other thing that bothers me is that the plot seems too smooth; I would normally expected more noise.

Changing the different resolution or hyper parameters changes when exactly the heat flow may increase, but the general shape is the same.

Do you have any other suggestions or potential issues?

‌

[David Dickinson]

Thanks. Yes I agree it looks rather smooth, I suspect this may be due to being dominated by a single wavenumber.

I’ve certainly seen cases previously where one sees a transition between dominant ky and a corresponding jump in heat flux. Have you checked the linear growth rate spectrum?

How does the zonal potential compare to the non-zonal? Have you tried looking at the theta dependence of the dominant fluctuations and are these well resolved?

I don’t often run with zero shear and when I do it’s usually for simple linear tests so, whilst I think it should work, I’m not sure if there are any possible issues associated with a shearless flux tube setup here. Have you tried dropping the shear further to see if this changes anything?

Could you show phi2_by_ky vs time?

I tried running a cheaper version of this (ntheta=16, no collisions and no hyper terms) and I got something which saturated at about the same level but seemed to have some more temporal structure. I swapped from default to ‘random_sine’ (new in 8.1) and also forced the initial zonal amplitude to zero.

The only issue that I’m aware of in this area is our handling of the periodic point in our calculation of the response matrices. If we treat this point identically for both instances and included this in the response matrix we would have duplicate identical rows in the matrix so could not invert.

‌