[mesa-users] stellar-structural noise

[Aaron Dotter]

Hi Alfred,

Thanks for such an interesting email.  I haven't seen such a thing
before but that is only because I have not looked.  I have three
suggestions for further exploration.

1.  There are some controls in the star_job inlist (see eos controls) for
shifting the boundaries between the different EOS options.  These don't
appear to be near the region where you see the wiggles, but might be worth
trying to shift the boundaries anyway.

2.  Have you tried to adjust the model mesh?  There is a wonderfully
detailed section of the controls inlist called "mesh adjustment" with all
sorts of options for refining the mesh.  The broad brush approach is to
reduce mesh_delta_coeff, but don't stop there!

3.  These are solar mass models.  Have you tried computing models with
element diffusion enabled?

I hope these suggestions will at least give you further insight.

Cheers,
Aaron



On Friday, October 18, 2013, Alfred Gautschy wrote:

> I attach a plot showing spatial profiles (measured in log T) of
> homology invariant U in the deeper interior of a sequence of solar-mass
> models evolving off the main sequence.
> Problem: Whatever I do to the steering of MESA, I cannot get rid of the
> "noise" in the U profiles.
>
> I played with various opacity tables (gn93, gs98, a09, OP): to no
> avail. I played with the two EoS choices (mesa, macdonald): to no
> avail. I played with different nuclear networks: to no avail.
> Changing num_threads, even going down to 1 - no parallelism at all - did
> make no difference. Using P rather than rho as principle dependent
> variable makes no difference either. I required tighter tolerances in
> the convergence of the Newton-Raphson iterations: to no avail at all. It
> is interesting to note that the ripples have a systematics which
> repeats, slightly shifted in temperature, from one model to the next
> (indicated in the attached plot in two cases with light yellow lines).
> The regular pattern of the wiggles is somewhat reminiscent of an
> underlying grid (not that of the stellar models though!).
>
> Since the "noise" is more expressed in U and much less in V and since
> the choice of opacity has essentially no effect, in particular not on
> the regularity pattern, I presume the problem is associated with the
> density.
>
> The wiggliness of U (in regular PopI or PopII stars) is mostly
> confined to roughly 6.5 and 7.3 in log T; a higher temperature window,
> however, appears to be involved in helium stars. So it is not unlikely
> that this hints at compositional profiles being involved. The amplitudes
> of the wiggles diminish considerably outside of the mentioned domain.
>
> I wonder what I am doing wrong here or what point I miss. Is there
> anybody around out there in the MESAverse who encountered the same and
> who found a way to avoid it. To me, currently, the scattering of the U
> profiles are only a minor annoyance, as I cannot achieve the perfectly
> smooth-looking models. However, should I aim at *high-precision*
> asteroseismology, this behavior likely affects/disturbs some mode
> frequencies.
>
> Or is what I plotted here the price we must pay for using tabular EoS
> and having to interpolate in them?
>
> BTW: I am using MESA version 4723; relevant parts of an inlist as
> usually going into the computations is attached. Star models and
> tracks look at first sight as we are used to them from our textbook
> experience.
>
> Alfred Gautschy
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20131018/5428435c/attachment.html>