Degraded precision of evolutionary models when Ledoux switched on
parallelpro opened this issue · 2 comments
@mjoyceGR and I noticed that a 1Msun evolution track (r230501) is a bit bumpy on the HR diagram (Lum vs Teff) around the first acent RGB, with use_Ledoux_criterion = .true.. Here I attached the inlist to reproduce this track. For comparision, with use_Ledoux_criterion = .false. the track is much smoother. It would be great to gather some insights on this issue and improve it in future releases.
Does this persist if you include predictive mixing or convective premixing in your model? Without one of those options or some overshoot, it's very likely that convective boundary evolution will be incorrect, and that might be the cause of some of these features.
I was able to reproduce this issue.
You can get good results with use_Ledoux_criterion = .true. if you also set num_cells_for_smooth_gradL_composition_term=0 (the default value is 3; changing this parameter should not affect your results when ledoux criterion is false).
An explanation of this parameter and why setting it to 0 helps is found in the docs: https://docs.mesastar.org/en/latest/reference/controls.html#num-cells-for-smooth-gradl-composition-term
which I reproduce below:
Number of cells on either side to use in weighted smoothing of gradL_composition_term. gradL_composition_term is set to the “raw” unsmoothed brunt_B and then optionally smoothed according num_cells_for_smooth_gradL_composition_term. In cases where the Ledoux criterion is used to evaluate the boundary for burning convective cores, you may need to set num_cells_for_smooth_gradL_composition_term = 0 to avoid smoothing the stabilizing composition jump into the convection zone and unphysically causing it to shrink. See section 3.2 in Moore, K., & Garaud, P. 2016, APJ, 817, 54