rd1042

Just to check - the appendix of the stella paper shows a damped numerical mode (negative growth rate). Numerical instability only occurs if an explicit numerical scheme is used, and...

I've been having a fiddle around with the example input file and have noticed: - The instability gets worse as dt increases (e.g. from dt=0.01 to 0.1 in the input...

`The instability disappears when include_mirror = .false. in the physics_flags namelist. I had expected this to give the same behaviour as the case with set_bmag_const = .true. (if bmag's constant,...

@DenSto I'll take a look, but some additional (possibly relevant) info: - I've found phi^2 goes large (from phi^2 (t=0) = 1 to phi^2 (t=20) ~ 1E9) when only the...

> Does this instability occur with ky = 0 in your set-up? Using Michael's cbc input file and setting ky.rho_ref = 0. , kx.rho_ref = 0.1, dt = 0.1, I...

> Also, what does the very-long-time behaviour with include_mirror = .false. look like? Is it actually stable, or is it unstable with a much reduced growth rate? I've tested up...

> Re: Bob's theory about the mirror term and finite vpa grid, do you find that increasing vpa_max alleviates the initial phi blowup? - In the "mirror term only" simulations,...

Running the same simulation in the branch feature/Davies-linear-EM-terms (commit 39822afe19a2244fd32cd71d6e3e493e46c3bdf4 ) (keeping beta=0), I find that we still have an instability, but this time, it gets worse as dt gets...

@mabarnes can you remind me why stella uses (vpa, mu) as coordinates rather than (e.g.) (energy, mu)? Does this simplify how trapped particles are simulated, or similar?

This sounds sensible to me. I usually use Edmund Highcock's thesis for the geometric quantities in a sheared slab.