[Mesa-users] Implementation of the Canuto-Mazzitelli Model

Wed Nov 19 16:29:09 UTC 2025

Hello Vishal, (please keep mesa-users cc'd)

I would suggest turning turbulent pressure off during the pre-ms relaxation, and then reenabling it when the model begins evolving.

If you're keen enough to try, (now that you have some experience modifying mesa to fix your mlt gradT bug), you're welcome to try adding said function into the auto_diff variable types in $MESA_DIR/star/auto_diff, to see if this alleviates your issue with adopting autodiff. I can say for certain that the numerical stability of your routines will likely be affected by your choice to include or not include auto_diff, or to return analytic derivatives. The mlt_vc_ad and gradT_ad returned by mlt to the star newton solver typically carry with them derivatives that aid in convergence.

-EbF
________________________________
From: Vishal Vetrivel <vishalsvetrivel at gmail.com>
Sent: Wednesday, November 19, 2025 5:43 AM
To: Farag, Ebraheem <ebraheem.farag at yale.edu>
Subject: Re: [Mesa-users] Implementation of the Canuto-Mazzitelli Model

Thank you for your replies.
I read the paper on the auto_diff variables and I get how it works now. I am not sure if I can even implement it the way it is written now since I am using the safe_root for the Gamma value (which seems to accept only a real value). Unless there is a similar utility for auto_diff values I think I might just stick to what I have right now.
As for the turbulent pressure I tried it out but I wasn't able to get a model to get beyond the pre main sequence model creating for C values from Table 1 in the 1991 CM paper (~0.2). The solver seems to give up in the pre main sequence model generation. To check I tried lowering the value of C and it seems to work up to around 0.1. I am unsure exactly what the issue is but the final errors state that it stopped due to dt < min_timestep_limit. I can provide more information from the output but I'm not sure what would be relevant so I have included the last few lines of the MESA output I got when I ran a max_Pturb_factor above 0.1 (even 0.12 doesn't seem to work).

stopping because of problems dt < min_timestep_limit
 failed in relax num steps
 failed in do_relax_num_steps
 star_create_pre_ms_model ierr          -1
 do_load1_star ierr          -1
 do_before_evolve_loop ierr          -1
 do_before_evolve_loop ierr          -1

On Mon, Nov 17, 2025 at 9:36 PM Farag, Ebraheem <ebraheem.farag at yale.edu<mailto:ebraheem.farag at yale.edu>> wrote:
Hello Vishal,

Is the question, where to implement the turbulent pressure in other structure equations such as the energy and momentum equation?

Following https://adsabs.harvard.edu/pdf/1991ApJ...370..295C, I see (eq40) Pt = 2*C*(1/2*rho*v_turb^2). Where C is given in Table 1.

I believe in your implementation  v_turb -> conv_vel -> (in mesa star) mlt_vc.

In that case, the turbulent pressure will take similar form to the mlt_turbulent_pressure (see https://docs.mesastar.org/en/24.08.1/reference/controls.html#mlt-pturb-factor), the documentation here is actually slightly outdated, as we mass weight on to faces instead of simple averaging (see hydro_energy):

where Pt_mlt = 1/3 * rho * v_turb^2 = 1/3 * rho * mlt_vc^2.

To implement turbulent pressure in the momentum equation, all you need to do is change 'mlt_Pturb_factor' to something that matches the CM model.

mlt_Pturb_factor = 3*C.

-EbF
________________________________
From: Mesa-users <mesa-users-bounces at lists.mesastar.org<mailto:mesa-users-bounces at lists.mesastar.org>> on behalf of Philip Mocz via Mesa-users <mesa-users at lists.mesastar.org<mailto:mesa-users at lists.mesastar.org>>
Sent: Monday, November 17, 2025 9:08 AM
To: Vishal Vetrivel <vishalsvetrivel at gmail.com<mailto:vishalsvetrivel at gmail.com>>
Cc: mesa-users at lists.mesastar.org<mailto:mesa-users at lists.mesastar.org> <mesa-users at lists.mesastar.org<mailto:mesa-users at lists.mesastar.org>>
Subject: Re: [Mesa-users] Implementation of the Canuto-Mazzitelli Model

Hi Vishal,

I'm excited you are trying to implement this into MESA!

Section 2 of the MESA VI paper shows how to use the autodiff variables to automatically get derivatives of a function with respect to it's input variables. https://arxiv.org/abs/2208.03651  . It's no problem to avoid using autodiff and use analytic derivatives instead if you have those expressions.

I'll have to dig more to find where to connect/update turbulent pressure. Does anyone know off the top of their head?

Best,
Philip

On Sun, Nov 16, 2025 at 5:42 AM Vishal Vetrivel via Mesa-users <mesa-users at lists.mesastar.org<mailto:mesa-users at lists.mesastar.org>> wrote:
Hello all,
I have been attempting to implement the Canuto-Mazzitelli and Canuto-Goldman-Mazzitelli model into MESA for the past few months. I managed to find an old MESA user list thread from Dr. Warrick Ball (https://sourceforge.net/p/mesa/mailman/message/33552129/<https://urldefense.com/v3/__https://sourceforge.net/p/mesa/mailman/message/33552129/__;!!DSb-azq1wVFtOg!S9dz0uQsHWzxET8X62meLlV0zU_BkQ1c-CzGSBs9bFVEgpiQ5_VJ91WGHYR0ekhNw7xjYbzDdCXwQW3qRKb55XXBe2eGUWET$>) detailing a similar implementation. However, since this was a while ago a lot of the code has since changed. I initially hoped to implement it in the same way by using a `run_stars_extra.f90` file. However, a bug in the MESA prevented me from doing so as the routine get_gradT does not call the custom mlt hook. Instead I have been implementing this into a local copy of the MESA source code (I am very grateful that you can run multiple versions of MESA so easily).

In this process I have managed to implement all the modifications made in the original thread. However, I am facing some issues figuring out where to implement the turbulent pressure. I have managed to replace the convective velocity with what I believe should be the equivalent expressions for CM and CGM. I have not yet implemented setting the mixing length to the distance to the convective surface yet (I believe there is already an option to limit the mixing length to the depth so this might not be necessary). A major issue I had was I am not sure how the new auto_diff variables are supposed to work so I have manually assigned the derivatives using the old reference implementation. This is a pretty bad way to do it so I'd be grateful if anyone could help me with how these variables are meant to be used.

I have attached my modified mlt.f90 file from the turb module which I believe should be most if not all the modifications I have made to source code. I am using version 24.08.1 of MESA. Thank you in advance for your help.

--
Regards
Vishal Sai Vetrivel
5th Year Integrated Masters
School of Physical Sciences
UM DAE CEBS, Mumbai

_______________________________________________
mesa-users at lists.mesastar.org<mailto:mesa-users at lists.mesastar.org>
https://urldefense.com/v3/__https://lists.mesastar.org/mailman/listinfo/mesa-users__;!!DSb-azq1wVFtOg!S9dz0uQsHWzxET8X62meLlV0zU_BkQ1c-CzGSBs9bFVEgpiQ5_VJ91WGHYR0ekhNw7xjYbzDdCXwQW3qRKb55XXBe8VerUaX$

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20251119/09128439/attachment.htm>