Composite overshoot

[matthiasfabry]

add functionality to model a step + exponential overshoot.

you can enable with:

      overshoot_scheme(1) = 'step+exponential'
      overshoot_zone_type(1) = 'burn_H'
      overshoot_zone_loc(1) = 'core'
      overshoot_bdy_loc(1) = 'top'
      overshoot_f(1) = 0.335   ! step part, slightly less than brott
      overshoot_f2(1) = 0.01  ! exponential part, replace outer part of brott with this
      overshoot_f0(1) = 0.01

[Debraheem]

I just cleaned up some of the white space diff so this touches less files, and tweaked the docs slightly. Now we just need testing https://testhub.mesastar.org/composite_overshoot/commits/27af2e7.

[wmwolf]

Running on Hercules: https://testhub.mesastar.org/composite_overshoot/commits/a84c8d0. It'll take a while to get through things since it's running serially, though.

[wmwolf]

Nearly done. Looks like make_pre_ccsn_13bvn is stalling... currently at over 10k models and log_dt ~ –10 in inlist_cc.

[wmwolf]

Hmm.. the more I watch make_pre_ccsn_13bvn, the more concerned I am. It seems to have no smooth failure mode like a maximum model number. I'll let it keep going, but I think you can probably count it as a failure. The remaining missing tests are from binary and astero, and thus haven't even started yet, so I doubt there are issues with them.

[Debraheem]

It definitely needs a max_model_number. I think you should kill the test, if it's going on indefinitely. I'm a little surprised it is failing thought, given this branch did not touch it.

[Debraheem]

@wmwolf https://testhub.mesastar.org/main/commits/aa27a08/test_cases/star/make_pre_ccsn_13bvn, it should only do ~4k timesteps. I can't see the log since it doesn't update testhub until the run is complete, but maybe this is helpful.

[wmwolf]

It surprised me, too. I killed it and the remaining tests should roll in soon.

[Debraheem]

I'm surprised the BVN test is even failing, since it is untouched by this pr except via the change to

mesa/star/private/overshoot_utils.f90

Line 219 in a84c8d0

if (s%alpha_mlt(k)*Hp > dr) Hp = dr/s%alpha_mlt(k)

My lack of sleep from last week has turned into an illness, but when i'm back up to speed, i'll investigate and try to fix this while i go through all the other branches from the hackathon.

[wmwolf]

Feel better soon, @Debraheem! Hercules has finished testing. Everything passed, except for make_pre_ccsn_13bvn, which stalled at tiny timesteps and had to be manually killed.

[warrickball]

I've just started this on bluebear, which does full tests. make_pre_ccsn_13bvn hasn't been presenting any issues in the full tests, so it might just be that an intermediate model needs updating.

[warrickball]

All clear on bluebear.

[Debraheem]

Would you be open to updating the mod files from your test? It’s okay if they’re deleted, i can re run it locally!

…

-EbF

On Tue, May 26, 2026 at 9:19 AM Warrick Ball ***@***.***> wrote: *warrickball* left a comment (MESAHub/mesa#1003) <#1003 (comment)> All clear on bluebear <https://testhub.mesastar.org/composite_overshoot/commits/a84c8d0>. — Reply to this email directly, view it on GitHub <#1003?email_source=notifications&email_token=APBTKAELIRZCGOANJWIHAV344WKW7A5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINJUGQZTQMRRGEZ2M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-4544382113>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/APBTKAH535YWC5MBTMF6N5344WKW7AVCNFSM6AAAAACZE3JMZCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DKNBUGM4DEMJRGM> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[warrickball]

Incidentally, does this not also allow one to refactor pure step/exponential overshoot into special cases of step+exponential? As in, is step overshoot the same as step+exp with the exponential coefficients set to zero? (I haven't yet had a chance to look closely at the code.)

[Debraheem]

Incidentally, does this not also allow one to refactor pure step/exponential overshoot into special cases of step+exponential? As in, is step overshoot the same as step+exp with the exponential coefficients set to zero? (I haven't yet had a chance to look closely at the code.)

I think in principle, yes. It might be much simpler as well. Something like just turning on the exponential tail if using step and f2 is non-zero.

[Debraheem]

@warrickball following your comments, I've refactored the implementation into the step overshoot routine, and significantly reduced the diff. Thanks for the feedback! I've also regenerated the mod files from the bvn case, so it should pass normal testing now.