[Mesa-users] STELLA calculation fail to converge for massive stars

Thu Dec 17 19:54:54 EST 2020

Hi Amar,

I don't have a perfect answer for you. The speed of the STELLA calculations
does not track "ease of getting the model to core-collapse in MESA."

For your reasonably-normal-looking lightcurves, it should be noted that
CCSN Lightcurves are dominated by 2 factors: the Ni mass, and the mass of
the Hydrogen-rich ejecta (which, for a normal IIP, is just the total ejecta
mass, since RTI mixes H into the deep interior of the ejecta). If you cut
out extra material near the inner boundary, or have fallback (and don't add
accretion-powered luminosity or anything like that, which STELLA does not
incorporate), then the lightcurves will likely still look very "normal"
upon inspection.  As a sanity check, it may be useful to check the mass
coordinate of the inner boundary in STELLA for both models.

a) In  inlist_to_lgT9.9, If the lgT=9.85, then the Fe core infall limit is
> reached before the lgT=9.85 condition satisfied (I missed to mention this
> in previous mail), and the calculation stops. Is it possible to delay the
> Fe core infall so that lgT=9.85 is achieved or vice-versa? I had tried the
> *T_function1_weight *but still failed to reach the prescribed temperature
> limit before the Fe core infall limit reached.

If the Fe core infall limit is reached before lgT = 9.85 condition is
satisfied, then I guess the core did collapse! My work is mostly modeling
red supergiants 10 < M/M_\odot < 25, so I don't have perfect intuition
about the exact regions of parameter space where you'll get core-collapse
for something like 300-400Msun.  My guess is this isn't causing the issues
in STELLA.

b) With remove_initial_center_by_entropy set less than 18.0, the following
> error message (for the case when remove_initial_center_by_entropy =4) is
> displayed in the terminal:

<image>

This error message is saying that the value for the entropy at the center
of the star is greater than 4. This probably, but not necessarily, means
that the innermost material which is to be excised in inlist_infall is not
a degenerate Fe core. That may be fine if you're just looking at
lightcurves or ejecta structure, depending on what your specific science
goals are.
More things to think about, which may help from a scientific perspective:
What is the value of m_center after inlist_infall? What is the value of
m_center after inlist_shock_part5? Are they the same or was there any
fallback?
To figure out if these values make sense, you can plot the composition
profiles of the model before inlist_infall, in order to know the
composition of the material that was excised from the MESA model.

Again, STELLA running slowly may not indicate any problems in MESA, even if
there is a separate question of "how can I use MESA to create a model that
will be easy for STELLA to use". In my experience there are two "problem
regions" of the ejecta which may lead to STELLA may take a long time: The
inner boundary (especially if material there is moving very slowly), and
the surface if there is interaction between the shock and the surface. Here
it will be very helpful to compare to what did seem to work in STELLA.
What properties of the ejecta structure are different for the 300 and 400
Msun models? Perhaps take a look at a density profile and a velocity
profile near core-collapse as a function of mass coordinate, and maybe even
log(radius).

Other questions which may lend you some insight: What are the velocities of
the inner ejecta? Did you use the same explosion energy for both the 300
and 400 Msun models? How does the total final energy you used in
inlist_edep compare to the binding energy of the star (total energy of the
star before you put the thermal bomb in in inlist_edep)?

Hope this rambling is somewhat helpful,

Cheers,
~Jared

On Mon, Dec 14, 2020 at 3:05 AM Amar Aryan <amararyan941 at gmail.com> wrote:

> Dear Jared,
>
> Many thanks for your prompt response. I am really overwhelmed. Before
> going to the references suggested by you, I wish to mention that a similar
> situation arose for a 300M_sun case, but  lgT = 9.61  in
> inlist_to_lgT9.9  and remove_initial_center_by_entropy = 18.0  in
> inlist_to_lgT9.9 worked nicely, that's why I am troubled with the very slow
> STELLA calculations for the case of 400M_sun. The STELLA produced
> lightcurve and FeII line velocity evolution with these settings for
> 300M_sun progenitor are inscribed below.
> [image: luminosity.png][image: FeII5169_vel_300M.png]
> a) In  inlist_to_lgT9.9, If the lgT=9.85, then the Fe core infall limit is
> reached before the lgT=9.85 condition satisfied (I missed to mention this
> in previous mail), and the calculation stops. Is it possible to delay the
> Fe core infall so that lgT=9.85 is achieved or vice-versa? I had tried the  *T_function1_weight
> *but still failed to reach the prescribed temperature limit before the Fe
> core infall limit reached.
>
> b) With remove_initial_center_by_entropy set less than 18.0, the following
> error message (for the case when remove_initial_center_by_entropy =4) is
> displayed in the terminal:
> [image: terminal_err.png]
> So, I do not know how to proceed further. Please suggest.
>
> With best regards,
> Amar
>
> On Mon, Dec 14, 2020 at 12:29 PM Jared Goldberg <goldberg at physics.ucsb.edu>
> wrote:
>
>> Hi Amar,
>>>
>>>
>>> a) I could notice that after finishing the *inlist_to_si_burn* stage in
>>> the *example_make_pre_ccsn* directory, the model fails to achieve the
>>> prescribed temperature (lgT) of 9.85 while running *inlist_to_lgT9.9*,
>>> so I decreased the limit to 9.61 and rerun the calculation. This time the
>>> evolution successfully proceeded. Please note that even a lgT of 9.62 was
>>> not achieved so I tried 9.61.
>>>
>> Sounds like the model never made it to the end of Si burning/the
>> generation of the Fe core, which should happen around log(Tcenter) = 9.9.
>>
>>
>>> b) Thereafter, the shock is evolved using the *example_ccsn_IIp* directory
>>> with few changes like, in *inlist_infall*, we required to change the *remove_initial_center_by_entropy
>>> = 19.6*   (default value is 4). Few timestep adjustments were also
>>> done. Then the calculation proceeded nicely and the *mesa.abn* and
>>> *mesa.hyd* (please see attached ) files were successfully created.
>>> Please take a look to *inlist_shock_part5* also.
>>>
>>
>> This makes sense, given point (a). The entropy profile tells you where
>> the various core boundaries are, and the Fe core boundary crosses an
>> entropy value of 4. So by going up in entropy you're effectively removing
>> more than what would have been the Fe core, likely closer to the CO
>> boundary.
>>
>> c) These two files were then used as input for STELLA calculations. The
>>> STELLA calculation started nicely but seems that it fails to converge. On a
>>> 16GB RAM and octa-core processor machine, the calculations ran up to a
>>> period of more than 24hours but the light curve only up to 7days was
>>> produced with the calculation still going on.
>>
>> Since this is a fundamentally STELLA issue, I'm happy to discuss as a
>> MESA+STELLA user but note that it seems like everything you're doing in
>> actual MESA is working.
>> It might just be the case that more runtime is simply what STELLA needs.
>> Ultimately, you are solving radiation transport through a lot of
>> slow-moving, dense material, with an order of magnitude more mass than a
>> normal IIP calculation. Things to consider: What are the velocities you
>> get? What explosion energy do you use? What do you set for the innermost
>> velocity (your fallback cut)? Do you get any fallback (see Appendix D of MESA
>> instrument paper V
>> <https://ui.adsabs.harvard.edu/abs/2019ApJS..243...10P/abstract> and
>> appendix A of this paper
>> <https://ui.adsabs.harvard.edu/abs/2019ApJ...879....3G/abstract>)? For a
>> 400 Msun star, the inner ejecta definitely could be causing the
>> computational slowdown.
>>
>> Cheers,
>> ~Jared
>>
>> On Sun, Dec 13, 2020 at 10:37 PM <amar at aries.res.in> wrote:
>>
>>> Dear MESA team,
>>>
>>> MESA Version : 11701
>>>                   OS : ubuntu 20.04
>>>
>>> I am trying to study the evolution of very massive stars (in particular
>>> 400M_sun without any rotation and mass loss) with the help of MESA. I could
>>> evolve the star up to the stage of core-collapse. Here, I wish to discuss
>>> and understand few points :
>>>
>>> a) I could notice that after finishing the *inlist_to_si_burn* stage in
>>> the *example_make_pre_ccsn* directory, the model fails to achieve the
>>> prescribed temperature (lgT) of 9.85 while running *inlist_to_lgT9.9*,
>>> so I decreased the limit to 9.61 and rerun the calculation. This time the
>>> evolution successfully proceeded. Please note that even a lgT of 9.62 was
>>> not achieved so I tried 9.61.
>>>
>>> b) Thereafter, the shock is evolved using the *example_ccsn_IIp*
>>> directory with few changes like, in *inlist_infall*, we required to
>>> change the *remove_initial_center_by_entropy = 19.6*   (default value
>>> is 4). Few timestep adjustments were also done. Then the calculation
>>> proceeded nicely and the *mesa.abn* and *mesa.hyd* (please see attached
>>> ) files were successfully created. Please take a look to
>>> *inlist_shock_part5* also.
>>>
>>> c) These two files were then used as input for STELLA calculations. The
>>> STELLA calculation started nicely but seems that it fails to converge. On a
>>> 16GB RAM and octa-core processor machine, the calculations ran up to a
>>> period of more than 24hours but the light curve only up to 7days was
>>> produced with the calculation still going on.
>>>
>>> Please suggest me some possible methods to get convergence of STELLA
>>> calculation. Also, suggest if I am doing something wrong.
>>>
>>> With thanks and best regards,
>>> Amar
>>>
>>>
>>> _______________________________________________
>>> mesa-users at lists.mesastar.org
>>> https://lists.mesastar.org/mailman/listinfo/mesa-users
>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20201217/f1fa3593/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: terminal_err.png
Type: image/png
Size: 62131 bytes
Desc: not available
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20201217/f1fa3593/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: FeII5169_vel_300M.png
Type: image/png
Size: 5294 bytes
Desc: not available
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20201217/f1fa3593/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: luminosity.png
Type: image/png
Size: 4779 bytes
Desc: not available
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20201217/f1fa3593/attachment-0002.png>