[Mesa-users] [EXT] Fragmented HR diagram from pisn test suite
Amar Aryan
amararyan941 at gmail.com
Mon Sep 14 09:21:01 UTC 2026
Thank you for suggesting the references, Mathieu. Since my goal is to study
the resulting light curves, I think starting from a helium star would be a
wise idea.
With regards,
Amar
On Sat, Sep 12, 2026 at 2:12 AM Mathieu Renzo <mrenzo at arizona.edu> wrote:
> Hi Amar,
>
> Good practice to always do the resolution test with your specific setup --
> in general there is no "guarantee" that any value works for your setup and
> science case!
>
> On the HRD: first thing that really catches the eye is you're not
> reverting your T_eff axis! I now understand better *when* the HRD jumps
> occur, and looking at the ppisn test case, you can see that they correspond
> to when the test case saves a *.mod file, reads another inlist, and then
> loads the *.mod file to restart.
> This numerical procedure is of course not how nature works, presumably
> the physics is the same throughout each evolutionary phase. The test case
> tries to build the initial conditions for (p)pisn *fast* (in cpu and
> wallclock time) for testing purposes, not *right* for science, and I think
> those jumps are a manifestation of that.
> While it is true that the test_suite exemplifies capabilities of MESA, and
> it has been suggested as starting point for science in the past, it
> now-a-days is more optimized to run fast at the expenses of science. Most
> of the time I try to start my project from published setups rather than the
> test cases, personally.
>
> In the published (p)pisn works I am aware of, we did not take this
> approach (on the other hand, we used pure-He balls as initial conditions
> with few exceptions), and for your science I would suggest the same.
> Try to have one single run smoothly going from your initial conditions
> (appears that you want to start from pre-MS) to the pulses/pisn regime. I
> would recommend starting from a clean $MESA_DIR/star/work directory, and
> building your own inlist/run_star_extras.f90, copying the relevant parts
> from the pisn test case or from the papers by Marchant+19, Farmer+19, +20,
> my own papers, Farag+22
> (all have zenodo repositories with code and output, but also, MESA has
> evolved in the past ~6-7 years and they will likely not be "plug-an-play",
> you'll need to identify things to update and update them yourself).
>
> Good luck!
> Mathieu
>
>
>
> On 9/11/26 01:04, Amar Aryan wrote:
>
>
> * External Email *
> ------------------------------
> Dear Mathieu,
>
> Thank you for your prompt reply. The suggestions were indeed very helpful.
>
> Since I kept mesh_delta_coeff and mesh_delta_coeff_at_high_T values <1, I
> considered them satisfactory for initial runs. Of course, I usually run
> the convergence test before any serious study. I rerun my model at a much
> higher spatial resolution by setting mesh_delta_coeff=0.3 and
> mesh_delta_coeff_at_high_T=0.4 (the models converged; a slightly coarser spatial
> resolution with mesh_delta_coeff=0.4 and mesh_delta_coeff_at_high_T=0.5
> gave almost identical results). The HRD was still fragmented. Please see
> the attached image. The jumps were during the following stages:-
>
> 1. First, when the models go from make-late-pre-zams to zams
> 2. Second, core-He-exhaustion to remove-envelope
> 3. remove-envelope to core-C-exhaustion
>
> [image: Test_PISN_HR.png]
>
> Thank you for detailing the possible reason for the jumps in my HRD.
> Unfortunately, *the jumps are during the early stages of evolution* when
> the model switches from one stage to another (as mentioned above and
> indicated in the attached plot). My goal is to study the resulting
> lightcurves. I have been using radiative transfer codes with some
> preliminary success in generating the initial light curves already from the
> MESA pisn models; the light curves looked encouraging! But now I am a
> little worried if the models were actually physical.
>
> *In both your cases, if they are *pisn* (which the termination code
> suggests), then I would say you should expect the remnant mass to be zero.*
> I noticed that at the end of the evolution, total_energy_foe > 0, but the
> remnant mass is ~110 Msun even when I simply run the test suite without any
> changes. I deal with this later when calculating the light curves, where I
> assume all of this mass is the ejecta and that nothing remains as a central
> remnant.
>
> Please provide your suggestions.
>
> With best regards,
> Amar
>
> On Fri, Sep 11, 2026 at 1:18 AM Mathieu Renzo <mrenzo at arizona.edu> wrote:
>
>> Hi Amar,
>>
>> great results! A few comments:
>>
>> - you say "satisfactory temporal and spatial resolution" and quote the
>> varcontrol_target and mesh_delta_coeffs. The values you use seem ok
>> intuitively, but that does not guarantee "satisfactory resolution" in
>> general. What do you mean exactly by *satisfactory*? To me that usually
>> means that the physics I want to study has a bigger impact than the
>> numerics, so it's a statement dependent on what physics I want to study.
>> The only way to be sure is to run a resolution test by running the same
>> model twice, the second time with smaller `mesh_delta_coeff` and
>> `mesh_time_coeff` to force more mesh points/more timesteps and check that
>> the results agree (see
>> https://docs.mesastar.org/en/latest/using_mesa/best_practices.html#explore-numerical-convergence
>> ).
>>
>> - the point above is general, and has nothing to do with pisn
>> specifically. For the HRDs of the (p)pisn models you have to make sure
>> about what you are plotting as luminosity. The pisn termination condition
>> is when *all the mass in the domain is moving faster than the local escape
>> velocity*, which means we compute a good chunk of the explosion itself to
>> make sure nothing aborts the explosion. *However*, during the explosion,
>> especially in the outer layers, the model stops being physical: optically
>> thin low density ejecta are treated like a stellar interior. If you want to
>> study the interior structure this is ok (those layers are faster than the
>> escape and sound speed and don't feed back on the interior), but if you
>> want the light curve, then this can and usually is a problem. The (p)pisn
>> setup is meant to study the interior structure, more than the light curve.
>> It is plausible to me that the jumps in the HRD are a manifestation of the
>> non-physical nature of the outer layers during the explosion. Check *when*
>> the jump happens. Is it before the explosion (check T_c or eps_nuc)? If
>> during or after, while or after a shock propagates through the star, I
>> would say you shouldn't look at the HRD from MESA (or you need to make sure
>> the outer layers are physical, which they likely aren't).
>>
>> - changing overshooting changes the structure of the star. How that
>> results in a higher final mass is not a trivial question: by making the He
>> core (or CO core) bigger, it takes less mass loss for the wind to remove H
>> (or He) envelopes, and change the wind driving, resulting in a different
>> mass loss. In both your cases, if they are *pisn* (which the termination
>> code suggests), then I would say you should expect the remnant mass to be
>> zero. The pre-pisn mass is known to be sensitive to winds and other
>> assumptions and I think your experiments re-affirm this. In terms of BH
>> masses from ppisn in Rob Farmer's 2019 paper
>> <https://ui.adsabs.harvard.edu/abs/2019ApJ...887...53F/abstract> we
>> claimed that overshooting does not have a large impact, but we didn't look
>> at the pre-explosion masses (see fig. 4, relevant panel attached):
>>
>>
>>
>> Hope this helps!
>> Mathieu
>> On 9/10/26 09:59, Amar Aryan via Mesa-users wrote:
>>
>>
>> * External Email *
>> ------------------------------
>> Dear MESA users,
>>
>> MESA Version: mesa-r26.04.1 with MESA-SDK version 26.6.1
>> OS: MacOS 26.6
>>
>> I am using the 'pisn' test suite to study the evolution of a 200 M_sun
>> ZAMS star. The evolution proceeds successfully with the termination code of
>> "Star is unbound". When I plot the resulting HR diagram for the entire
>> evolution (from inlist_make_late_pre_zams to inlist_finish), I get a
>> fragmented curve (please see the attached plots), although I am using
>> satisfactory temporal (varcontrol_target=1d-4) and spatial resolutions
>> (mesh_delta_coeffs of 0.7 and 0.8). The final mass of the star in this case
>> was ~ 110 M_sun.
>>
>> Later, I slightly experimented with the overshooting parameters (similar
>> to 20M_pre_ms_to_core_collapse) by adding a) step overshooting in H core,
>> b) exponential in the He core, c) inward overshooting in the shell across
>> compositions boundaries, and d) a small amount of overshooting on top of
>> any other convective core to avoid spurious numerical behaviour. Please see
>> the attached inlist_common file. In this case, the evolution proceeds
>> successfully with the termination code of "Star is unbound". The final
>> mass of the star in this case was ~ 130 M_sun!
>>
>> I have two basic queries:
>>
>> 1. Why are the HR-diagram and the rho_core vs T_core curves fragmented
>> despite satisfactory temporal and spatial resolution?
>>
>> 2. Why does the model with the modified overshooting parameters have a
>> higher final mass? Since the model terminates with a similar termination
>> code of "Star is unbound", can we assume that it is another successful
>> "pisn" model?
>>
>> [image: Test_PISN_HR.png]
>>
>> [image: Test_PISN_Rhoc_vs_Tc.png]
>>
>> I shall appreciate any comments/suggestions on these queries and the
>> choice of modified overshooting parameters.
>>
>> With best regards,
>> Amar
>>
>> _______________________________________________mesa-users at lists.mesastar.orghttps://lists.mesastar.org/mailman/listinfo/mesa-users
>>
>> --
>> Mathieu Renzo <https://lavinia.as.arizona.edu/~mrenzo/>
>> Assistant Professor
>> University of Arizona - Steward Observatory
>>
> --
> Mathieu Renzo <https://lavinia.as.arizona.edu/~mrenzo/>
> Assistant Professor
> University of Arizona - Steward Observatory
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20260914/d52c6236/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Test_PISN_HR.png
Type: image/png
Size: 74371 bytes
Desc: not available
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20260914/d52c6236/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EVzzPqfyCA1zXl9E.png
Type: image/png
Size: 77958 bytes
Desc: not available
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20260914/d52c6236/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Test_PISN_HR.png
Type: image/png
Size: 56545 bytes
Desc: not available
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20260914/d52c6236/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Test_PISN_Rhoc_vs_Tc.png
Type: image/png
Size: 54841 bytes
Desc: not available
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20260914/d52c6236/attachment-0003.png>
More information about the Mesa-users
mailing list