Nonadiabatic computation of mixed modes

Bug/problem reports for any of the GYRE executables (gyre_ad, gyre_nad, etc)
Post Reply
benfernando
Posts: 6
Joined: Mon Jun 18, 2018 10:10 am

Nonadiabatic computation of mixed modes

Post by benfernando » Wed Mar 27, 2019 11:21 am

Hi all,

I'm looking at whether non-adiabatic effects are important in a particular class of helium core-burning stars. In the adiabatic case the calculations look fine, but in the non-adiabatic case I get very small chi values (~1e-20) and in some cases, non-monotonic mode orders.

I understand that this might be an issue with the numerical discretisation, but am not sure how to fix it (increasing the grid points to 100 in the oscillation regions and 20 in the evanescent regions doesn't fix the problem).

In short: Does anyone have any idea whether this is likely to be an issue with the model (I've already set double point at density jumps to true in MESA) or with the search parameters I'm using? My aim is to identify whether the differences between the adiabatic and non-adiabatic modes are genuine or not.

If it helps, the models were computed a few years ago using a rather old version of MESA. I'm using the latest version of GYRE though.

Thanks
Ben
Attachments
gyre.in
(2.22 KiB) Downloaded 32 times
he_core_burn_unstripped_TII_10000K_double.GYRE
(714.08 KiB) Downloaded 33 times

benfernando
Posts: 6
Joined: Mon Jun 18, 2018 10:10 am

Re: Nonadiabatic computation of mixed modes

Post by benfernando » Mon Apr 01, 2019 11:15 am

As requested: initial model (which I inherited), an example inlist (I've tried uncommenting the last two lines of &controls to change the grid), and the final model from which I take the pulsation data. Thanks!
Attachments
HE_core_burn.zip
(323.62 KiB) Downloaded 16 times

jacquelinegoldstein
Posts: 3
Joined: Tue Nov 11, 2014 12:37 pm

Re: Nonadiabatic computation of mixed modes

Post by jacquelinegoldstein » Wed Apr 10, 2019 1:05 pm

Hi Ben,

The non-monotonicity is the result of a weakness in the root solver when searching for non-adiabatic frequencies, which are complex. The root solver needs an initial trial frequency, and right now those are the adiabatic frequencies, which are purely real. The more non-adiabatic a frequency, the further it is from the adiabatic frequency, and the harder it is for the root solver to find it from an adiabatic initial trial frequency. We're aware of this weakness and are in the process of developing a new method for finding initial trial frequencies that is more robust.

Also, if you use the magnus GL2 scheme you'll see the monotonicity is better, but it is still incomplete.

I've attached an image comparing the two methods on the complex frequency plane. Here the grey filled circles are the adiabatic frequencies and the black filled circles are the non-adiabatic frequencies found using your inlist. Each non-adiabatic frequency is connected with a line back to the adiabatic frequency that was used as its initial trial frequency. Some adiabatic frequencies without lines didn't converge to a non-adiabaic frequency at all; the root solver wandered off into the woods. Some non-adiabatic frequencies were found by subsequent adiabatic frequencies (resulting in the non-monotonicity), and some non-adiabatic frequencies were never found.

The red open circles are the non-adiabatic frequencies found using our new method. They are monotonic in order and include the non-adiabatic frequencies that were previously missed. We are in the process of developing the new method for release in GYRE and it is in preparation for publication.

Sincerely,
Jacqueline
Attachments
l0_ad_nad_collocgl2_map.pdf
(162.27 KiB) Downloaded 27 times

benfernando
Posts: 6
Joined: Mon Jun 18, 2018 10:10 am

Re: Nonadiabatic computation of mixed modes

Post by benfernando » Wed Apr 24, 2019 1:21 pm

Hi Jacqueline,

Thanks very much for your reply - I look forward to the new release!

Best
Ben

Post Reply