[mesa-users] MESA 6749 "noisy" tracks

[Robert FARMER]

Hi
So clearly your results are sensitive to the resolution of the model which
is not a good place to be!

The question to ask is are either of the results trustworthy? for models
4589 or 6794. They may be both wrong in different ways, especially as you
sensitive to resolution, then changes in any of the solvers will cause
differences to occur or changes in the mixing or compositions or nets....
.Its really too big a gap in versions to guess whats changed and why.

I would suggest trying to find a paper/experimental results and seeing if
any models can match the data. At this point i'd be sceptical of both sets
of results

Or you could try running each version on its own at different resolutions
till they converge. Then compare the final converged results for each
version with the other version, even if that means that each version has a
different var_control or mesh_delta_coeff

Rob

On Wed, Oct 29, 2014 at 10:50 AM, Sibilla Perina <sibilla.perina at gmail.com>
wrote:

> Hi
> As a first test I did this.
> I noted that the default values of mesh_delta_coeff (1.0) and
> varcontrol_target (1e-4) are the same in the mesa versions 4589 and 6794.
> However, they produce different space/time resolution (tip1.pdf); why?
> Then, I run 6794 with mesh_delta_coeff 1.0, 0.6, 0.5, 0.4, and
> varcontrol_target 1e-2, 1e-4.
> The result most similar to the 4589 one is obtained with 6794 assuming
> mesh_delta_coeff = 0.5 and maintaining varcontrol_target as default
> (1e-4)(tip2.pdf).
> In addition to the tiny difference in logL (0.02) and logTeff (0.003),
> there is an age difference of about 450 thousand years between the two
> rgb-tip. Moreover, the 6794 execution-time is tripled.
> (tip3.pdf shows 6794 tracks assuming the two different parameter pairs).
>
> Cheers,
> Sibilla
>
> 2014-10-23 15:06 GMT-03:00 Sibilla Perina <sibilla.perina at gmail.com>:
>
>> Time ago I computed a set of evolutionary tracks using MESA 4589 (SDK
>> 20120728). Now I’m computing the same tracks with MESA 6794 (SDK 20140715).
>> I’m on OSX 10.8.5.
>>
>> I’m using the multimass method. The inlist is the same for both the
>> versions. The two run_star_extras.f have some differences I think due to
>> the code upgrade, but they should do the same things. I attached here the
>> two inlist and the two run_star_extras.f.
>> In the attached figure tip.pdf I plotted two tracks (M=1.0 Msun, Z=0.02)
>> corresponding to the different MESA versions, in a HR region very near to
>> the RGB tip.
>>
>> I’d like to know why the track from the new MESA version is so noisy
>> respect to the other and if there is a way (parameter setting) to obtain a
>> more cleaned track as the one from the older MESA version. Moreover, why
>> there is that difference in the tip location?
>>
>> Thanks for your help.
>> Cheers,
>> Sibilla
>>
>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20141029/d1a230d2/attachment.html>