[Mesa-users] Failed to create a starting model
Cao, Kaili
cao.1191 at buckeyemail.osu.edu
Thu Jul 7 13:45:21 UTC 2022
Hi Warrick,
Thank you so much for your advice.
I ran a series of simulations last night and everything went smoothly. Current I am testing whether I can roll back the other changes. It seems to me that it is still necessary to set:
pre_ms_relax_to_start_radiative_core = .false.
and other things suggested by Meridith and Thomas can be rolled back. Nevertheless I really appreciate their help.
I think it would be great if you update the documentation. Thanks.
Cheers,
Kaili
________________________________
From: Warrick Ball <W.H.Ball at bham.ac.uk>
Sent: Wednesday, July 6, 2022 5:26 PM
To: Cao, Kaili <cao.1191 at buckeyemail.osu.edu>
Cc: MESA Users List <mesa-users at lists.mesastar.org>
Subject: Re: [Mesa-users] Failed to create a starting model
Hi Kaili,
Try setting the following controls in `&controls` to match your desired abundances:
initial_y = 0.275d0
initial_z = 0.0160738558d0
where the values I've used correspond to the abundances in the inlist you attached. These changes got me to the stopping condition (D depletion) after 394 steps with 1 retry (in r22.05.1).
These controls set the abundances that the pre-MS model builder uses for the composition of the initial guess, and I suspect your problems are coming from the sudden change to the abundances between the end of the relaxation of the pre-MS model and the start of the actual evolution.
If this does consistently fix your issue, I'll see if we can document this behaviour better so others don't run into the same unexpected behaviour. You might also find you can roll back some of the other changes you made to try to work around the issue.
Cheers,
Warrick
___________
Warrick Ball
Postdoc, School of Physics and Astronomy
University of Birmingham, Edgbaston, Birmingham B15 2TT
W.H.Ball at bham.ac.uk
+44 (0)121 414 4552
On Wed, 6 Jul 2022, mesa-users at lists.mesastar.org wrote:
>
> Dear MESA Users,
>
>
> Thank you for reading this email. This is a follow-up to my inquiry last week. Thanks to Thomas and Meridith’s advice, the deuterium burning issue has been partially resolved.
>
>
> However, it still hinders my exploration of the parameter space. Specifically, I am fixing the initial Z at 0.016, varying the initial Y, and setting the initial deuterium abundance according to lodders03 isotopic mass fractions. The initial Y values
> and the resulting deuterium abundance values are tabulated as follows, together with the steps, solvers retries and time for the model to reach the central deuterium exhaustion:
>
>
> initial Y
>
> initial_h2
>
> steps
>
> solver retries
>
> time (h:mm:dd)
>
> 0.275
>
> 1.37531672d-5
>
> 3762
>
> 710
>
> 1:55:28
>
> 0.277
>
> 1.37143672d-5
>
> 1255
>
> 181
>
> 0:35:28
>
> 0.279
>
> 1.36755672d-5
>
> 653
>
> 55
>
> 0:15:04
>
> 0.281
>
> 1.36367672d-5
>
> 407
>
> 2
>
> 0:06:46
>
>
> An exemplar inlist is attached to this email.
>
>
> Since I plan to study lower Z (and hence lower Y) models, this exponential growth of time consumption with decreasing Y is unacceptable. Besides, the large number of solver retries makes the simulation results suspicious. I tried further lowering
> pre_ms_T_c, but all the simulations failed. I think that’s why Meridith said that lower pre_ms_T_c doesn’t work. I also tried to use the Eddington grey atmosphere, but it did not help.
>
>
> I plan to switch back to a previous MESA release and try to temporarily circumvent this issue, but I think eventually it must be resolved. The implementation of deuterium burning of the latest MESA release might be problematic. Thank you for your
> time.
>
>
> Best regards,
>
> Kaili Cao
>
> __________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> From: Mesa-users <mesa-users-bounces at lists.mesastar.org> on behalf of Cao, Kaili via Mesa-users <mesa-users at lists.mesastar.org>
> Sent: Friday, July 1, 2022 7:50 PM
> To: Meridith Joyce <meridith.joyce at gmail.com>
> Cc: MESA Users List <mesa-users at lists.mesastar.org>
> Subject: Re: [Mesa-users] Failed to create a starting model
> Hi Meridith, Could you please explain in a little more detail why any pre_ms_T_c lower than 3d5 doesn't work? Is this statement based on numerical experiments or physical arguments? Thank you. Cheers, Kaili
> Hi Meridith,
>
> Could you please explain in a little more detail why any pre_ms_T_c lower than 3d5 doesn't work? Is this statement based on numerical experiments or physical arguments? Thank you.
>
> Cheers,
> Kaili
>
> __________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> From: Cao, Kaili <cao.1191 at buckeyemail.osu.edu>
> Sent: Friday, July 1, 2022 7:48 PM
> To: Thomas Steindl <thomas.steindl at uibk.ac.at>
> Cc: MESA Users List <mesa-users at lists.mesastar.org>
> Subject: Re: [Mesa-users] Failed to create a starting model
> Hi Thomas,
> Thank you very much for telling me about release/r21.12.1 and a viable pre_ms_T_c value. A lower pre-main-sequence temperature makes good sense. I will try to run this simulation with different numerical precisions.
>
> However, I don’t think the 7Be(e-,nu)7Li is the cause of the problem. Based on the abundance plots I made with a successful run, deuterium burning seems to be the only (significant) nuclear reaction before the central deuterium exhaustion; during
> this stage, the 7Li abundance does not change, while the 7Be abundance increases from 7E-81 to 5E-39.
>
> Anyway, I really appreciate your help.
>
> Cheers,
> Kaili
>
> __________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> From: Thomas Steindl <thomas.steindl at uibk.ac.at>
> Sent: Friday, July 1, 2022 7:05 PM
> To: Cao, Kaili <cao.1191 at buckeyemail.osu.edu>
> Cc: MESA Users List <mesa-users at lists.mesastar.org>
> Subject: Re: [Mesa-users] Failed to create a starting model
> Hi Kaili, This is interesting. I can verify the convergence issues with r22.05.01, but it runs fine with r21.12.1 . The changelog states: The 7Be(e-,nu)7Li has been switched from REACLIB rate to that of Simonucci et al 2013
> Hi Kaili,
> This is interesting. I can verify the convergence issues with r22.05.01, but it runs fine with r21.12.1 . The changelog states:
>
> The 7Be(e-,nu)7Li has been switched from REACLIB rate to that of Simonucci et al 2013.
>
> As 7Be(e-,nu)7Li is part of the pp chain it seems like this results in an issue with your setup. I cannot comment on the specific differences between the rates. However, I was able to run the model successfully with pre_ms_T_c = 1d5 instead
> of pre_ms_T_c = 3d5. It seem that the with pre_ms_T_c = 3d5 the core might again already be too hot. I have only run this one inlist though, so feel free to run some more tests. :)
>
> Best,
> Thomas
> __________________________________________________________________
>
> Thomas Steindl, MSc
> University of Innsbruck | Institute for Astro- and Particle Physics
> Tel: +43 512507 52031
> Mail: thomas.steindl at uibk.ac.at
> Homepage: https://urldefense.com/v3/__http://www.thomassteindl.com__;!!KGKeukY!3qhTMW7SfPTWC6MRBRSuj49F_TGSGtOpOMwY9ot8iGxuhBQh14laqv-gnQykJPGhjsmQoD1dcjs5K2arcVrbIaIvXl9dYmB-$
>
> Am 01.07.2022 um 23:53 schrieb Cao, Kaili <cao.1191 at buckeyemail.osu.edu>:
>
> Hi Thomas,
> Thank you for your help.
>
> Attached are the inlists and the rn shell script I used in my last attempt. The rn script is based on the test suite version because I want to use the do_one function to run multistage simulations. I didn’t expect this to cause differences in
> simulations, but I realized that it might.
>
> By the way, I am using the latest release (release/r22.05.1) and running my simulations on a Red Hat machine.
>
> Thanks again,
> Kaili
>
> __________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> From: Thomas Steindl <thomas.steindl at uibk.ac.at>
> Sent: Friday, July 1, 2022 5:15 PM
> To: Cao, Kaili <cao.1191 at buckeyemail.osu.edu>
> Cc: MESA Users List <mesa-users at lists.mesastar.org>
> Subject: Re: [Mesa-users] Failed to create a starting model
> Hi Kaili,
> I just tried to reproduce the error messages but the inlists you sent with the modifications suggested by Meridith seem to run pretty stable for me. Can you send the inlist that leads to the error messages you describe?
>
> Thomas
> __________________________________________________________________
>
> Thomas Steindl, MSc
> University of Innsbruck | Institute for Astro- and Particle Physics
> Tel: +43 512507 52031
> Mail: thomas.steindl at uibk.ac.at
> Homepage: https://urldefense.com/v3/__http://www.thomassteindl.com__;!!KGKeukY!3qhTMW7SfPTWC6MRBRSuj49F_TGSGtOpOMwY9ot8iGxuhBQh14laqv-gnQykJPGhjsmQoD1dcjs5K2arcVrbIaIvXl9dYmB-$
>
> Am 01.07.2022 um 21:27 schrieb Cao, Kaili via Mesa-users <mesa-users at lists.mesastar.org>:
>
> Thank you for reviewing my inlist. I misthought that I needed to turn on relax_to_radiative_core to have a radiative model, which is indispensable for MESA to check_radiative_core_Lnuc_div_L_limit.
> Now that I turn that off, I steadily got:
> retry: max residual jumped -- give up in solver
> twice or thrice every 10 models before model 1479 (a fraction large enough to make the results suspicious); and since then I got:
> retry: max resid -- give up in solver
> for almost every single model; eventually I got:
> stopping because of problems dt < min_timestep_limit
> terminated evolution: cannot find acceptable model
> termination code: min_timestep_limit
> again after model 1508.
>
> The H-R diagram is shown as follows:
> <0_H-R_diagram.png>
>
> Could you please provide with me some hints about possible causes and solutions? Thank you.
>
> __________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> From: Meridith Joyce <meridith.joyce at gmail.com>
> Sent: Friday, July 1, 2022 2:39 PM
> To: Cao, Kaili <cao.1191 at buckeyemail.osu.edu>
> Cc: MESA Users List <mesa-users at lists.mesastar.org>
> Subject: Re: [Mesa-users] Failed to create a starting model
> It looks like this inlist does not have relax_to_radiative_core turned off (it is on by default). Can you tell me what happens when you include this?
>
> On Fri, Jul 1, 2022 at 2:26 PM Cao, Kaili <cao.1191 at buckeyemail.osu.edu> wrote:
> Hi Meridith,
> I merged all the settings you mentioned into my inlist (except the ones with default values; see the attached lasted version), but unfortunately, things got worse. Not only got I:
> retry: max residual jumped -- give up in solver
> retry: avg+max corr+resid -- give up in solver
> retry: avg+max resid -- give up in solver
> for almost every single model, I also got:
> stopping because of problems dt < min_timestep_limit
> terminated evolution: cannot find acceptable model
> termination code: min_timestep_limit
>
> The documentation says: “If you have initial convergence problems creating a pre-ms model, you might try different values for pre_ms_T_c – that sometimes helps.” Do you have any suggestions regarding this parameter? I think it is
> more of a numerical choice. Thank you.
>
> I really don’t want to use the MESA code as a black box, however there are too many mysterious, unexplained factors showing up as obstacles. Specifically, due to my limited understanding of the relaxation method, I have no idea why
> the seemingly insignificant initial deuterium abundance can make such a huge difference.
>
> Cheers,
> Kaili
>
> __________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> From: Mesa-users <mesa-users-bounces at lists.mesastar.org> on behalf of Cao, Kaili via Mesa-users <mesa-users at lists.mesastar.org>
> Sent: Friday, July 1, 2022 1:42 PM
> To: Meridith Joyce <meridith.joyce at gmail.com>
> Cc: MESA Users List <mesa-users at lists.mesastar.org>
> Subject: Re: [Mesa-users] Failed to create a starting model
> Hi Meridith,
> Thanks a lot for sharing your valuable experience. I am studying whether setting
> pre_ms_check_radiative_core_Lnuc_div_L_limit = 1d-6
> works, and will test the numerical precisions and different atmosphere choices.
>
> Cheers,
> Kaili
>
> __________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> From: Meridith Joyce <meridith.joyce at gmail.com>
> Sent: Friday, July 1, 2022 11:17 AM
> To: Cao, Kaili <cao.1191 at buckeyemail.osu.edu>
> Cc: MESA Users List <mesa-users at lists.mesastar.org>
> Subject: Re: [Mesa-users] Failed to create a starting model
> Hi Kaili,
>
> I spent a few months last year looking at a similar problem. There are, in my opinion, some still-unresolved issues when it comes to the pre-main sequence relaxation procedure (in particular, that it causes "proper" evolution for a
> low-mass model to begin at too hot a temperature).
> The following settings have helped me with low-mass stellar models using deuterium burning:
>
> &star_job
> pre_ms_T_c = 3d5 ! default -- any lower doesn't work
> pre_ms_check_radiative_core_min_mass = 0.3d0 ! default is 0.3d0
> pre_ms_check_radiative_core_Lnuc_div_L_limit = 1d-6
>
> &controls
>
> atm_option = 'table'
> atm_table = 'photosphere'
>
> In the inlist I'm pulling from, I also use weaker mesh and time restrictions than you do:
> mesh_delta_coeff = 0.8
> time_delta_coeff = 1.0
>
> Let me know if that works.
>
> Cheers,
>
> On Fri, Jul 1, 2022 at 10:17 AM Cao, Kaili via Mesa-users <mesa-users at lists.mesastar.org> wrote:
> Sorry, I forgot to attach the new version of my inlist.
>
> __________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> From: Mesa-users <mesa-users-bounces at lists.mesastar.org> on behalf of Cao, Kaili via Mesa-users <mesa-users at lists.mesastar.org>
> Sent: Friday, July 1, 2022 10:12 AM
> To: MESA Users List <mesa-users at lists.mesastar.org>
> Subject: [Mesa-users] Failed to create a starting model
> Dear MESA Users,
> Following Thomas’s advice, I set pre_ms_relax_to_start_radiative_core = .false. to build my starting model with specified initial deuterium abundance. Although this enabled MESA to produce some results, the issue has not been
> resolved, and a more severe problem has occurred when I tried to use higher numerical precision.
>
> Using the attached inlist (updated in this email), I repeatedly got:
> retry: max residual jumped -- give up in solver
> retry: avg+max resid -- give up in solver
> retry: max resid -- give up in solver
> Quantitatively, I got the “give up in solver” message 206 times for 1137 models. Although the resulting H-R diagram (see below; the blue curve corresponds to this inlist) seems reasonable, such a large fraction of “give up in
> solver” models make the results suspicious.
>
> H-R diagram with mesh_delta_coeff = 1.0d0:
> <0_H-R_diagram.png>
>
> Besides, it took 21.0 minutes for the starting model to reach central deuterium exhaustion; in contrast, the entire evolution from the central deuterium exhaustion to the RGB bump (the orange curve in the above plot) took only
> 23.4 minutes. Such a high time consumption was not expected for a starting model.
>
> The time consumption becomes unacceptable when I reran the same simulation with a higher spatial resolution (mesh_delta_coeff = 0.5d0). Usually when I halve mesh_delta_coeff, I double the time consumption. However, this time
> it took MESA 16.5 hours to simulate a small fraction of the previous case. See the H-R diagram below.
>
> H-R diagram with mesh_delta_coeff = 0.5d0:
> <0_H-R_diagram.png>
>
> Could you please provide me with some hints about the large fraction “give up in solver” models and the high time consumption of my starting models? Thank you for your time and I look forward to your response.
>
> Best regards,
> Kaili Cao
>
> __________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> From: Mesa-users <mesa-users-bounces at lists.mesastar.org> on behalf of Cao, Kaili via Mesa-users <mesa-users at lists.mesastar.org>
> Sent: Thursday, June 30, 2022 3:18 PM
> To: Thomas Steindl <thomas.steindl at uibk.ac.at>
> Cc: MESA Users List <mesa-users at lists.mesastar.org>
> Subject: Re: [Mesa-users] Failed to create a starting model
> Hello Thomas,
>
> Thank you for your quick and insightful reply. I really appreciate it.
>
> Cheers,
> Kaili
>
> __________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
> From: Thomas Steindl <thomas.steindl at uibk.ac.at>
> Sent: Thursday, June 30, 2022 1:58 PM
> To: Cao, Kaili <cao.1191 at buckeyemail.osu.edu>
> Cc: MESA Users List <mesa-users at lists.mesastar.org>
> Subject: Re: [Mesa-users] Failed to create a starting model
> Hello Kaili,
> I was able to run your inlist to completion by setting
>
> pre_ms_relax_to_start_radiative_core = .false.
>
> This will start the evolution at the Hayashi track instead of first evolving the initial model until a radiative core develops. I am speculating here, but the issues could be related to putting deuterium into the stellar core
> with central temperature that far exceeds the temperature at which deuterium burning starts.
>
> From your inlist it seems like you are interested in finishing the evolution during the phase of deuterium burning. For this you will have to start before the creation of the radiative core anyway.
>
> Best,
> Thomas
> __________________________________________________________________
>
> Thomas Steindl, MSc
> University of Innsbruck | Institute for Astro- and Particle Physics
> Tel: +43 512507 52031
> Mail: thomas.steindl at uibk.ac.at
> Homepage: https://urldefense.com/v3/__http://www.thomassteindl.com__;!!KGKeukY!3qhTMW7SfPTWC6MRBRSuj49F_TGSGtOpOMwY9ot8iGxuhBQh14laqv-gnQykJPGhjsmQoD1dcjs5K2arcVrbIaIvXl9dYmB-$
>
> Am 30.06.2022 um 19:19 schrieb Cao, Kaili via Mesa-users <mesa-users at lists.mesastar.org>:
>
> Dear MESA Users,
> Thank you for reading this email.
>
> I am trying to create a starting model using the attached inlist. The only difference between this version and a previous version that worked is the way I set the initial composition. However I got:
> stopping because of problems dt < min_timestep_limit
> terminated evolution: cannot find acceptable model
> termination code: min_timestep_limit
> Could you please point out the problem(s) with my inlist? Thanks.
>
> Besides, I tried to solve this problem by using a simpler atmosphere. However, by examining the source code atm_support.f90, it seems to me that none of the fixed_something choices presented in the documentation is still
> available. Is there any available atmosphere simpler than the default Eddington grey atmosphere, which is easier for MESA to yield a starting model?
>
> Thank you for your time and I look forward to your response.
>
> Best regards,
> Kaili Cao
> <inlist_start>_______________________________________________
> mesa-users at lists.mesastar.org
> https://urldefense.com/v3/__https://lists.mesastar.org/mailman/listinfo/mesa-users__;!!KGKeukY!3qhTMW7SfPTWC6MRBRSuj49F_TGSGtOpOMwY9ot8iGxuhBQh14laqv-gnQykJPGhjsmQoD1dcjs5K2arcVrbIaIvXmCpXeX8$
>
>
> _______________________________________________
> mesa-users at lists.mesastar.org
> https://urldefense.com/v3/__https://lists.mesastar.org/mailman/listinfo/mesa-users__;!!KGKeukY!3qhTMW7SfPTWC6MRBRSuj49F_TGSGtOpOMwY9ot8iGxuhBQh14laqv-gnQykJPGhjsmQoD1dcjs5K2arcVrbIaIvXmCpXeX8$
>
>
>
> --
>
> Dr Meridith Joyce
> Lasker Fellow
> Space Telescope Science Institute
>
>
>
> --
>
> Dr Meridith Joyce
> Lasker Fellow
> Space Telescope Science Institute
> _______________________________________________
> mesa-users at lists.mesastar.org
> https://urldefense.com/v3/__https://lists.mesastar.org/mailman/listinfo/mesa-users__;!!KGKeukY!3qhTMW7SfPTWC6MRBRSuj49F_TGSGtOpOMwY9ot8iGxuhBQh14laqv-gnQykJPGhjsmQoD1dcjs5K2arcVrbIaIvXmCpXeX8$
>
>
> <inlist_start_header><rn><inlist_start>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20220707/12a29628/attachment.htm>
More information about the Mesa-users
mailing list