[pmocz]

Problem

In update_rotation_mixing_info (star/private/mix_info.f90), the rotational part of D_mix is scaled by a factor f that is meant to:

• turn rotational mixing off at high temperature (D_mix_rotation_max_logT_full_on = 9.4, D_mix_rotation_min_logT_full_off = 9.5), and
• optionally turn it off at low optical depth (D_mix_rotation_min_tau_full_off/on, both 0 by default).

The tau block added in d3022fd (Jan 2025) assigns f unconditionally, overwriting the logT factor:

if (s% tau(k) >= full_on_tau) then
   f = 1d0                                        ! discards the logT factor
else if (s% tau(k) <= full_off_tau) then
   f = 0d0
else
   f = (lgT - full_on) / (full_off - full_on)     ! uses lgT, not tau
end if

Consequences:

1. The logT cutoff is never applied with the default controls. tau ≥ 0 = full_on_tau, so f = 1 everywhere and rotational mixing is never suppressed above logT 9.4–9.5 as documented. This affects rotating massive stars run through late burning stages.
2. When a tau ramp is set (e.g. dev_TDC_through_ppisn sets 1 and 10), cells in the ramp get f from the logT formula. In an envelope with lgT ≈ 5 that is (5 − 9.4)/0.1 ≈ −44, i.e. a large negative D_mix_rotation.

A separate, older issue (from the initial import): the logT ramp runs the wrong way. f = (lgT - full_on)/(full_off - full_on) goes from 0 at full_on to 1 at full_off, then jumps back to 0. It should fall from 1 to 0. The same reversed ramp is used for am_nu_factor in profile_getval.f90, and in the formula in controls.defaults.

Fix

• Compute the logT factor f and the tau factor f_tau separately, and use f*f_tau.
• Ramp f_tau in tau: (tau - full_off_tau)/(full_on_tau - full_off_tau).
• Make the logT ramp decrease: (full_off - lgT)/(full_off - full_on), in mix_info.f90 and profile_getval.f90.
• Update the documented formula in controls.defaults and describe the tau factor there.

The ramp branches are only reached when full_off_tau < tau < full_on_tau (or full_on < lgT < full_off), so the denominators are positive and the default 0/0 tau settings never divide.

Expected behaviour change

• Rotating models that reach logT > 9.4 will again have rotational mixing reduced (fully off at logT ≥ 9.5), as before d3022fd. Late-stage results of rotating massive-star test cases may change.
• Models that set the tau controls get a proper 0→1 ramp in tau, combined with the logT cutoff.
• Within 9.4 < logT < 9.5 the factor now falls smoothly from 1 to 0 instead of rising and then jumping to 0.

Testing

linters/check_defaults.py passes. The code has not been compiled or run (no MESA SDK where this was written). In the test suite, ppisn is the case most likely to change: it rotates (new_omega_div_omega_crit = 0.1) and runs to Fe-core infall, so it reaches logT > 9.4. 12M_pre_ms_to_core_collapse, 20M_pre_ms_to_core_collapse, zams_to_cc_80 and pisn turn on rotation with Ω = 0, so their rotational D_mix should be essentially zero. The dev_TDC_through_ppisn dev case, which sets the tau controls, should also be checked.

🤖 Generated with Claude Code

[Debraheem]

Thanks for fixing my errors! I half-hazardly added these controls thinking they would be useful for ppisn, and I no longer rely on them so feel they might be useless bloat. As such, I'm considering removing them entirely unless someone feels they should stay.