[mesa-users] Calculating oscillation modes of higher degree

[Wiebke Herzberg]

Thanks Bill, Rich and Ehsan for your comments and suggestions so far.

But there is one more question bothering me: Did you, or anyone else 
reading this, ever succeed to calculate a "clean" (meaning no jumps in 
the radial order n and reasonably looking eigenfunctions) set of modes 
with high l values (let's say 10-20, which is still not that high, but 
anyway) for an evolved (off-MS, subgiant or red giant) star?

I am asking, because I would like to know if there is still hope to 
achieve this. After trying the example red giant model that comes with 
the (unofficial?) adipls.c release and running into the same problem 
there, I am quite concerned about this whole issue.

If I had an example model, for which the calculation actually works, 
then it would make things much easier. I could look at the model 
quantities you suggested and investigate similarities and/or differences 
and work my way forward from that. But right now I don't really know 
what to look for that could cause or could fix the issue.

Last week a came upon a paper by Grosjean et al.(2014): "Theoretical 
power spectra of mixed modes in low-mass red giant stars". Toward the 
end of section 2.2. they also mention numerical difficulties (although 
in the paper they only calculate modes up to l=2). I quote: "Therefore, 
we have adjusted the oscillation code to ensure the continuity of the 
Lagrangian perturbations of pressure, gravitational potential and its 
gradient."

This sounds like it is a problem I cannot solve with any settings in 
MESA, but that the oscillation code needs to be modified? So maybe this 
discussion does not belong on the MESA-mailinglist after all?

Anyway, I am digging into the model quantities in my FGONG file, the 
corresponding structure coefficients and the regularization option in 
GYRE now (which is not working for my models btw.). I will let you know 
if I find anything interesting.

Cheers,
Wiebke


Am 26.11.2014 um 12:10 schrieb Ehsan Moravveji:
> Hi Wiebke,
> Your concern is very relevant, and deserves full developer\92s attention, the way it is going on. I am very glad that Bill and Rich (after his successful Turkey hunting expedition) are on it. :-D
>
> Let me pass some ideas to you for your personal tests:
>
>> - increasing the mesh up to mesh_delta_coeff = 0.1
> Indeed that is one way to go. However, this dramatically slows you down. Instead, keep mesh_delta_coeff to a reasonable value, say 0.50, and only \93locally\94 increase your mesh resolution, around e.g. convective boundaries. MESA is extremely rich in controling such.
> Then, make your tests again, and see if your eigenfunctions are still shaky.
>> - smaller time steps, down to max_years_for_timestep = 1d7
> I suspect if this is relevant.
>> - turning diffusion on/off
> I suspect if this is relevant.
>> - starting with a pre-MS model instead of loading a ZAMS model
> I suspect if this is relevant.
>> - different convection/overshooting settings
>
> Make sure you adequately resolve these zones (see above).
>
> Additionally, these tests cross my mind:
> 1. How does increasing the mesh in your GYRE inlist influence eigenfunctions?
> 2. Have you played around with
>            tol_correction_norm = 3d-5
>            tol_max_correction = 3d-3
>      and the whole list of options in your controls inlist? They shall help you enforce HSE.
> 3. Explore the effect of setting change_lnPgas_flag off and on, and notice the difference. I bet you will see stuff going on.
> 4. Start with non-rotating models first, and get them working before going to rotating models.
> 5. Explore your input model around where you wiggly eigenfunctions first. What happens around r/R~0.017? mu-gradient zone? H-shell?
>      You may need to iron-out wiggles from your brunt profile too. That incredibly helps the numerics.
>
> Good luck with your tests, and keep us informed how you resolve the issue.
>
> Best regards,
> Ehsan
>