[mesa-users] Overshooting

Wed Jun 3 01:08:44 EDT 2015

Hi Matteo,

On Jun 2, 2015, at 8:59 PM, Matteo Cantiello <matteo at kitp.ucsb.edu<mailto:matteo at kitp.ucsb.edu>> wrote:

Ok. Just to wrap up this interesting overshooting tread and avoid any further confusion (please correct me if this is wrong):

1) Overshooting is implemented correctly in MESA (in the sense there are no bugs, as far as we can tell).

As I mention in my admittedly somewhat lengthy message just sent, a potential inconsistency in the implementation is the fact that the diffusion coefficient with exponential overshooting is not iterated during the solution scheme. Just something to be aware of, and to be addressed at some point. Not really a bug though, just an implementation detail.

2) The physics of overshooting is complex and not fully captured by the simplified models adopted in stellar evolution codes.

As I mentioned in my other message, I fully agree with this. The model is simple and f values need to be calibrated separately at each convective boundary, in principle. That is why there are separate controls for the bottom and top of different burning shells. We do start to see some systematics though. For example, the f value that Pavel found to work well at the bottom of convective nova H-shell flashes to account for the observed levels of enhancement of C and O in the nova ejecta is basically identical to what we usually adopt at the bottom of He-shell flash convection in AGB stars to account for the observed abundances of O and C in H-deficient post-AGB stars. That is probably not pure coincidence.

More work is currently done to improve the situation (thanks to brave souls like David Arnett and collaborators)
3) Asteroseismology can help from the observational sides of things by providing a best fit to the stellar models (thanks to the work done by e.g. Ehsan Moravveji and collaborators).

yes - agreed!

F.

4) The original problem raised by  Kenny  arises because of an inaccurate initialization of overshooting (thanks to Rob and Chris for spotting this).

Cheers,
-M

Matteo Cantiello  |  http://matteocantiello.com/
Associate Specialist in Astrophysics  | Kavli Institute for Theoretical Physics
Chief Scientist and Board Member, Authorea  |  https://www.authorea.com/

On Jun 2, 2015, at 5:58 PM, Chris Mankovich <cmankovich at ucsc.edu<mailto:cmankovich at ucsc.edu>> wrote:

Hi Kenny and everyone else,

I think Rob identified the reason why M_cc looks so ratty as a function of time.  With f=f0 you're effectively not adding an overshoot region.  You can plot the mixing D's to see this, and in fact with the f=f0=0.1 case you end up with a lot of isolated, borderline-stable zones that tend to merge with each other and sporadically with the convective core.

If I run f0=0.1, f=0.2 so that there is an overshoot region extending 0.1 H_p beyond the standard boundary, things behave more like you'd expect.

Chris

On Tue, Jun 2, 2015 at 3:25 PM, Ehsan Moravveji <e.moravveji at gmail.com<mailto:e.moravveji at gmail.com>> wrote:
Along the line of Falk’s reasoning, I would like to point out that very recently, we studied the so-far richest main sequence B-type pulsator, KIC 10526294, which shows 19 consecutive dipole g-modes. We tested several input physics, among which, the choice of step-function versus exponential overshooting prescription, and examined which set of models can reproduce the observation better; indeed, our intuition dictates that the more smooth exponential presciption should be preferred over the steep step-function. However, we tried to be unbiased and let this star guide us through.
What comes out is that the models with exponential prescription — as expected — outperform the other with a factor 2 to 3 in chi-square sense. We hope to keep testing this for our other B-type dwarfs too.
More information can be found here: http://adsabs.harvard.edu/abs/2015arXiv150506902M
Indeed, we all hope that a more physically grounded model, as Dave preaches, would soon liberate us from this shortcoming of MLT.

Best regards
Ehsan.

On 02 Jun 2015, at 19:58, Falk Herwig <fherwig at uvic.ca<mailto:fherwig at uvic.ca>> wrote:

Dear Kenny,

step overshooting is provided in MESA as an option to compare with historical computations that
adopted this kind of overshooting treatment, usually at a time when it was numerically not practical
to treat convective mixing as a time dependent process (say, in the 1970’s). Now, I would say that we
have no evidence that overshooting is step-like. Any simulation of convective boundary mixing,
including overshooting that has ever been made shows some gradual decline of the mixing efficiency
in the transition from unstable to stable stratification (I admit that too much convective boundary mixing
is the easy and wrong answer to obtain with too low grid resolution, a continuing challenge of such
simulations). And any terrestrial example of overshooting or convective boundary mixing would support
this notion. Whether or not the exponential decay of mixing model is the only one or even the best
one can be debated. It does have some support from hydro simulation though. You may adopt
another mixing model, but in any case it should allow for a smooth transition between stable
and unstable layer. You can make this smooth transition as thin as you can computationally afford, but
I am confident that it will note be the diffusion length-scale! One reason for a smooth and numerically resolved
transition to be desirable comes from purely numerical considerations that it is impossible to converge
in the presence of grid face discontinuities that are not somehow modelled. This may contribute to
your results to not “seem very good” to you.

Best, Falk.

--
Falk Herwig
Dept of Physics & Astronomy, U of Victoria
fherwig at uvic.ca<mailto:fherwig at uvic.ca>, tel: +1 (250) 721-7743<tel:%2B1%20%28250%29%20721-7743>

On Jun 2, 2015, at 8:34 AM, Kenny Van <kvan at ualberta.ca<mailto:kvan at ualberta.ca>> wrote:

Hi, I was trying to produce some results using overshooting and it seems as though regardless of timestep the plot produced doesn't seem very good. I have attached the three plots I created using my data. The overshooting and changes in timestep are said in the title. But just in case I will list all the properties of the star.

&star_job

   mesa_dir = '/home/kvan/mesa_svn'

   create_pre_main_sequence_model = .true.

/ !end of star_job namelist

&controls

! initial parameter

 ! initial mass of star
   initial_mass = 20.0

 ! initial metallicity of star, MESA only goes from 0 to 0.04
   initial_z = 0.014

 ! initial helium mass fraction of star
   initial_y = 0.266

 ! changing the MLT to match that in Meynet
   mixing_length_alpha = 1.6

! overshooting

 ! all exponential overshooting is by default zero
 ! setting step overshooting
   step_overshoot_f_above_burn_h = 0.1
   overshoot_f0_above_burn_h = 0.1
   step_overshoot_f_below_burn_h = 0.1
   overshoot_f0_below_burn_h = 0.1
   step_overshoot_f_above_burn_he = 0.1
   overshoot_f0_above_burn_he = 0.1
   step_overshoot_f_below_burn_he = 0.1
   overshoot_f0_below_burn_he = 0.1
   step_overshoot_f_above_burn_z = 0.1
   overshoot_f0_above_burn_z = 0.1
   step_overshoot_f_below_burn_z = 0.1
   overshoot_f0_below_burn_z = 0.1

! Mass loss

 ! Changing mass loss so system acts like that of Ekstrom et al.
 ! Use Vink mass loss rate
   RGB_wind_scheme = 'Vink'
   Vink_wind_eta = 1

! timestep limits
 ! Limit magnitude of max change in log10 total pressure in any cell
   delta_lgP_limit = 0.2

 ! Limit magnitude of max change in log10 temperature in any cell
   delta_lgT_limit = 0.2

! when to stop

 ! max timestep size
   max_years_for_timestep = 500

 ! stop when we reach AGB to produce similar results to Maynet et al.
   stop_at_TP = .true.

 ! max age
   max_age = 1.0d10

/ ! end of controls namelist
<20M_Low_Lim_noOS_5d5.png><20M_Low_Lim_OS_5d5.png><20M_Low_Lim_OS_500.png>

Thanks
------------------------------------------------------------------------------
_______________________________________________
mesa-users mailing list
mesa-users at lists.sourceforge.net<mailto:mesa-users at lists.sourceforge.net>
https://lists.sourceforge.net/lists/listinfo/mesa-users

------------------------------------------------------------------------------
_______________________________________________
mesa-users mailing list
mesa-users at lists.sourceforge.net<mailto:mesa-users at lists.sourceforge.net>
https://lists.sourceforge.net/lists/listinfo/mesa-users

------------------------------------------------------------------------------

_______________________________________________
mesa-users mailing list
mesa-users at lists.sourceforge.net<mailto:mesa-users at lists.sourceforge.net>
https://lists.sourceforge.net/lists/listinfo/mesa-users

<f=0p1_mixing_ds.png><f=0p1_mass_conv_core.png><f=0p2_mixing_ds.png><f=0p2_mass_conv_core.png>------------------------------------------------------------------------------
_______________________________________________
mesa-users mailing list
mesa-users at lists.sourceforge.net<mailto:mesa-users at lists.sourceforge.net>
https://lists.sourceforge.net/lists/listinfo/mesa-users

------------------------------------------------------------------------------
_______________________________________________
mesa-users mailing list
mesa-users at lists.sourceforge.net<mailto:mesa-users at lists.sourceforge.net>
https://lists.sourceforge.net/lists/listinfo/mesa-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20150603/3adee508/attachment.html>