[mesa-users] stellar-structural noise

[Warrick Ball]

On Fri, 18 Oct 2013, Alfred Gautschy wrote:

>> PS: Just to be specific, the relevant line in &controls is:
>> cubic_interpolation_in_Z = .true. ! .false. by default
>
> Thanks for the hint. I checked that I am indeed on
> cubic_interpolation_in_Z = .true.
> in the treatment of opacities. So, my problem is not related to the
> opacities directly at least regarding Z-interpolation.
>
> But since EoS is also tabular, I wonder what is done there with respect
> to interpolating in Z. I cannot find any switches with respect to its
> treatment of Z-interpolation though.

Hmmm... I'm out of ideas for the time being, then.  From the papers, it 
looks like the EoSes are interpolated with cubic splines in rho and T and 
quadratic interpolations in X and Z.  (That's what I read from Section 4.2 
of Paper I, anyway.)

The only other thing I can think of now is that it might be a plotting 
error.  That is, because U is (presumably?) derived from M, R and rho, 
perhaps limitations in the precision of those numbers is causing a few 
wiggles in funny places?

> It was actually *your* PhD thesis which hinted to me that we might have
> a tabular - i.e. coarse-grain EoS - problem here. As far as I remember,
> the Cambridge stellar-evolution code always solves the Eggleton et al.
> EoS directly at each location necessary and the UV diagrams in your
> thesis are smooth like a peach.

Holy moly, someone read my thesis! :p Only too glad to discuss further. :)

Firstly, I plotted Bezier curves through those points, because the models 
used only 199 points and look slightly jagged in their raw form.  I've 
attached a variant of Fig. 6.4 from my thesis with the Bezier smoothing 
off.  As you can see, no critical difference, but the curves do look 
slightly angular.  (It's more visible in the insets.)  But I thought I 
should point that out.

You're absolutely right that the Cambridge code uses the analytic (if 
approximate) EFF EoS.  I guess that implies that any EoS-tabulation effect 
would be absent from my models, but like I say I wouldn't conclude 
anything too quickly, given that your MESA models probably have many, many 
more meshpoints.

Anyway, that's all I can think of for now.

Cheers,
Warrick

>>
>> I haven't given it much thought but are you using cubic interpolation in
>> the opacities?  I think it's enabled by default in X but not in Z.  I
>> don't know if that could be the problem but it is a grid-related option.
>>
>> Cheers,
>> Warrick
>>
>
>> On Fri, 18 Oct 2013, Alfred Gautschy wrote:
>>
>>> Aaron
>>>
>>> Thanks for the quick reply with the pointers to further playgrounds. I
>>> am on the way to follow your suggestions an will be back when I have
>>> results.
>>>
>>> For the moment just a quick answer on diffusion to inform also other
>>> possible bystanders: NO DIFFUSION is included. I cut down on
>>> astrophysical processes as much as possible to hopefully isolate the
>>> source of evil at some point [if it should turn out not to be me ;-) ]
>>>
>>> Regards,
>>> Alfred
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>
>>
>> ------------
>> Warrick Ball
>> Postdoc, Institut für Astrophysik Göttingen
>> wball at astro.physik.uni-goettingen.de
>> +49 (0) 551 39 5069
>
>
>


------------
Warrick Ball
Postdoc, Institut für Astrophysik Göttingen
wball at astro.physik.uni-goettingen.de
+49 (0) 551 39 5069
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fpopIlo-grid.pdf
Type: application/pdf
Size: 247964 bytes
Desc: 
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20131018/6e8714d5/attachment.pdf>