[Mesa-users] Understanding why 'retry: get_T_tau -- L <= 0' occurs and ways of stopping it terminating my simulations

Wolf, William M WOLFWM at uwec.edu
Thu Aug 18 20:11:23 UTC 2022

Hi Arman,

I don’t have a quick solution for you. In my experience, mass loss from novae is always fraught with numerical difficulties (but if you don’t remove the mass, you get worse issues later on in the evolution). Often you’ll have low density matter with locally super-eddington luminosities with material just on top keeping it bound in a “pressure cooker” situation. Mass loss can then upset this delicate balance.

Some things to experiment with:

  *   atmosphere controls: Assuming you’re using the “Eddington” option for atm_T_tau_relation, you might try experimenting with the three values of atm_T_tau_opacity (“varying”, “fixed”, and “iterated”). “varying” should in theory be the most physically consistent, while “fixed” is the simplest. Opting for “fixed” may make things simpler and more stable.
  *   Adding “extra” pressure at the surface. The control Pextra_factor adds a multiplier to the surface pressure, which seems to help numerical convergence at the cost of physical accuracy.   You could try a value like 2, but if you have a high value of tau_factor, surprisingly high values of Pextra_factor can be numerically helpful (see wd_nova_burst for an example).
  *   Speaking of that, you can try increasing tau_factor to a higher value. This effectively moves the model-atmosphere deeper into the model, which sometimes can sidestep some numerical nastiness at the surface

None of these are bulletproof, and they don’t give a satisfying answer to what precisely is happening (perhaps an atmosphere expert can comment on that particular error), but they might just help you limp through the mass loss phase. Also, I would advise against manually changing the values of s% L, as that value may have been used both before and after you set it, potentially setting up some inconsistencies. In general, don’t modify any code in private unless there is a bug with a known fix that you are applying.


On Aug 16, 2022, 2:49 PM -0500, Arman Aryaeipour via Mesa-users <mesa-users at lists.mesastar.org>, wrote:
Dear MESA-users,

I'm using MESA revision 21.12.1 to model classical novae outbursts with ejection velocity mass loss. To do this, I use the other_adjust_mdot hook to eject mass from the model as the envelope expands and the outer material exceeds its local escape velocity. However, before any significant mass is lost from the model (~2% of the envelopes mass) I encounter the 'retry: get_T_tau -- L <= 0' error. Every retry that is attempted due to this error gets the same subsequent 'retry: get_T_tau -- L <= 0' error, entering the simulation into a cycle of retries that eventually stop the simulation with the 'terminated evolution: hydro_failed' code.

If I turn off mass loss a few timesteps before I get this error, then the envelope continues in its expansion with no problems. If I limit the timestep size so as to limit the amount of mass lost per timestep during the mass loss phase, I still eventually run into the same error and simulation termination. Therefore, it appears that the mass loss is causing the 'retry: get_T_tau -- L <= 0' error.

In an attempt to stop this error, I changed the luminosity values of the surface cells to be positive values extrapolated from the interior when they are assigned to the s% L luminosity pointer in the solver_support.f90 routine, but this only made a host of other convergence issues stop the simulation instead. I have also fiddled with various MESA parameters in an attempt to stop this error too, but none have made a difference.

I'm confused as to why I suddenly get this error, and what is causing it. Can anyone kindly explain why this error can suddenly become such a big issue, and the best ways of dealing with it?

I have attached an example of the inlist I use. I would be very grateful for any help or suggestions and thank you in advance!

Kindest regards,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20220818/3f2f2243/attachment.htm>

More information about the Mesa-users mailing list