ehsan wrote:

Thanks Rich for looking into this.

Please find the inlist and a sample model.gyre as a zip file in the following ftp address; I cannot attach to this message

ftp://anonymous@ftp.ster.kuleuven.be/dist/ehsan/dW_dx.zip

I use GYRE v.3.0. Sorry, a bit old fashioned

Cheers

Ehsan.

Thanks Rich for looking into this.

Please find the inlist and a sample model.gyre as a zip file in the following ftp address; I cannot attach to this message

ftp://anonymous@ftp.ster.kuleuven.be/dist/ehsan/dW_dx.zip

I use GYRE v.3.0. Sorry, a bit old fashioned

Cheers

Ehsan.

Hi Ehsan --

I'm still unable to reproduce your problem, whether with 3.0 or the latest development version. Could you post a link to one of the problem output files (nad-NNNNN.h5)?

cheers,

Rich

Statistics: Posted by rhtownsend — Tue Sep 30, 2014 9:03 pm

]]>

ftp://anonymous@ftp.ster.kuleuven.be/dist/ehsan/dW_dx.zip

I use GYRE v.3.0. Sorry, a bit old fashioned

Cheers

Ehsan.

Statistics: Posted by ehsan — Tue Sep 30, 2014 12:21 pm

]]>

ehsan wrote:

Hi Rich,

Thanks for your reaction.

Indeed I use and import the "gyre" module to interact with the output HDF5 files.

That's also strange that my routine works for you, but not for me

How shall we compare stuff?

Ehsan.

Hi Rich,

Thanks for your reaction.

Indeed I use and import the "gyre" module to interact with the output HDF5 files.

That's also strange that my routine works for you, but not for me

How shall we compare stuff?

Ehsan.

To start with, could you post the inlist + model.

cheers,

R

Statistics: Posted by rhtownsend — Tue Sep 30, 2014 8:21 am

]]>

Indeed I use and import the "gyre" module to interact with the output HDF5 files.

That's also strange that my routine works for you, but not for me

How shall we compare stuff?

Ehsan.

Statistics: Posted by ehsan — Tue Sep 30, 2014 5:38 am

]]>

I can't see anything wring with your Python code -- it works for me when I apply it to one of the non-adiabatic test cases.

How are you reading the GYRE data into Python? Are you using the supplied python module, or something else?

cheers,

Rich

Statistics: Posted by rhtownsend — Mon Sep 29, 2014 4:09 pm

]]>

Looking at the two files /nad/gyre_mode.f90 and /nad/gyre_util.f90, I see how W is simply calculated:

- Code:
`W = integrate(this%x, this%dW_dx())`

and

- Code:
`function integrate_r_ (x, y) result (int_y)`

real(WP), intent(in) :: x(:)

real(WP), intent(in) :: y(:)

real(WP) :: int_y

integer :: n

! Integrate y(x) using trapezoidal quadrature

n = SIZE(x)

int_y = SUM(0.5_WP*(y(2:) + y(:n-1))*(x(2:) - x(:n-1)))

! Finish

return

end function integrate_r_

Then, I try to reproduce the same net work using a simple Python script, and just cannot get there!

- Code:
`dx = dic['x'][1 : ] - dic['x'][ : -1]`

integrand= (dic['dW_dx'][1 : ] + dic['dW_dx'][ : -1]) / 2.

W_cumul = np.zeros(nz-1)

W_net = 0.0

for i in range(nz-1):

W_cumul[i] = np.sum(integrand[0:i+1] * dx[0:i+1])

W_net += integrand[i] * dx[i]

As an example, for a specific mode that GYRE gives W=1351.65750141, I get W_net=-1.93474738483.

Am I making a mistake here?

dW_dx from gyre_nad is still a real array?

Do I need to provide any file? inlist? mode info? input model?

Thanks in advance.

Ehsan.

Statistics: Posted by ehsan — Mon Sep 29, 2014 3:53 am

]]>

This is a bug-fix release, primarily focused on addressing some issues with the MESA interface.

As usual, the source code can be downloaded from

https://bitbucket.org/rhdtownsend/gyre/downloads/

Rich

Statistics: Posted by rhtownsend — Tue Sep 02, 2014 10:35 am

]]>

Apologies for the delay in answering you, I've been trying to come up with a better answer than the one I give here: the problem is with the model, and not with GYRE.

Experience over the past few years has shown that small departures from hydrostatic equilibrium, and other equilibria, in the input stellar model can cause distortions to the mode eigenfunctions, which in turn lead to mode mis-classifications. Figuring out how to suppress these departures is an area of ongoing research; but as a possible workaround, you might want to try increasing the number of grid points MESA is using for the model, to ensure that it's properly resolving the stellar structure.

Best wishes,

Rich

Statistics: Posted by rhtownsend — Sun Jul 06, 2014 10:47 am

]]>

I am having some trouble with the eigenfunction calculations for a subgiant (MESA-) model using GYRE. During the calculation of l=9 modes I noticed some improper values for n_pg. I tried to fix this issue by playing around with the different grid options as proposed on the GYRE website 'Understanding grids', but in this case it did not help. Even worse, when keeping the model grid the same (with RESAMP_DISPERSION enabled) and just using a finer frequency scan grid (n_freq=1500 -> n_freq=3500) I got the same frequencies as a result, but for some of them different eigenfunctions.

Even for some frequencies where the frequency values are exactly the same for both cases of n_freq, and the n_pg values are also the same and seem to be correct, there are differences in the eigenfunctions depending on the parameter n_freq!

I then had a closer look at the differing eigenfunctions, and it turns out that for the same frequency, they show in principle the same oscillatory behavior, but the p-mode part is shifted somehow (see attached figure "xir_mode17.png"), so the oscillations are not centered around zero anymore (which actually would explain the problem with the n_pg values in some of the cases, when the shift gets so large that the whole function is shifted above/below zero). In the file "diff_xir_mode17.png" the difference of the two eigenfunctions is plotted and you can see, that in the inner part, the eigenfunctions are nearly identical, but the difference gets larger in the outer part of the model. When zooming into the g-mode region, close to the core, there is actually also a difference present, but it is very small (~1e-10, see "diff_xir_mode17_core.png").

I find it quite unsettling that this whole problem occurs only for some of the modes but not all of them: for 10 of the 17 modes that were calculated, the eigenfunctions are exactly identical for both settings of n_freq (as I would expect).

Do you know what might cause this behavior of the eigenfunctions? Or how to solve this problem? I would be grateful for any help. Attached you can find the MESA model file and GYRE inlist I used.

Best wishes,

Wiebke

Statistics: Posted by wiebke — Tue Jun 24, 2014 9:38 am

]]>

Sorry for taking so long to check back here. As a feature, don't worry about trying to make sense of stuff above the cutoff. I realise that the adiabatic calculations stop making complete sense out there, and I was more curious about how the codes were treating this situation differently. Like I said, the main reason I asked is because I was trying to fit a particular star that has some very high reported frequencies, near the cutoff. When MESA compared the observed and modelled frequencies, I think the GYRE super-cutoff results were better at steering MESA in the right direction.

The problem, it turned out, was that I wasn't including the atmosphere in the pulsation calculation. Including it increased the acoustic cutoff by enough that the fit to start behaving better again. That is, the seismically-derived model parameters are consistent with the spectroscopic parameters.

Anyway, thanks as ever for the work and support!

W

Statistics: Posted by warrick — Wed May 21, 2014 2:02 am

]]>

Try uncommenting the alpha_osc and alpha_exp lines in the RESAMP_DISPERSION grid operation. Right now, that operation does nothing (because alpha_osc and alpha_exp revert to their default values of 0).

cheers,

Rich

Statistics: Posted by rhtownsend — Fri May 16, 2014 11:38 pm

]]>

As I'm sure you're aware, above the acoustic cutoff frequency there is incomplete reflection of upwardly propagating waves at the outer boundary; thus, these waves should partially leak through the boundary and escape the star.

Mathematically, this leakage introduces imaginary terms in the outer mechanical boundary condition, which in turn means that the stellar eigenfrequencies become complex. Physically, the complex eigenfrequencies arise because, as wave energy leaks through the outer boundary, the pulsation amplitude decays with time to conserve energy.

Both ADIPLS and GYRE (when doing adiabatic calcs) neglect the imaginary terms in the boundary condition when above the cutoff frequency, and force the eigenfrequencies to be real. Exactly how this is done is different in the two codes -- meaning that they give different results. Neither result is correct!

Would you like me to modify GYRE to properly handle cases above the cutoff? Most of the machinery is already there, it will just require a bit of special casing.

cheers,

Rich

Statistics: Posted by rhtownsend — Fri May 16, 2014 10:13 pm

]]>

I'm busy toiling with some model fits and I've ended up fitting models where some of the reported frequencies are above the acoustic cutoff. That almost certainly means that there's something wrong with the fit but it led me to do some cross-checking between ADIPLS and GYRE.

The plot below shows mode inertia versus frequency for ADIPLS (blue) and GYRE (green). As you can see, the mode inertiae are very different. In fact, the frequencies above the acoustic cutoff are also shifted by about 30 uHz (with GYRE frequencies lower than ADIPLS) but that's not obvious here. FYI, the match below the cutoff is pretty good: the largest difference is about 0.27 uHz. The number of points is relatively small in this model (1500ish) and I didn't do any remeshing in either code.

GYRE_ADIPLS_above_cutoff.png

I've attached the various input tools, including the model. Like I said, I realise that the problem is with the model (the acoustic cutoff should be higher!) but I'm interested to know why there's such a difference above the cutoff. I'm not that familiar with what happens in this frequency range but the GYRE frequencies are closer to the model results, so they help steer the fit in the right direction better.

Cheers,

Warrick

Statistics: Posted by warrick — Thu May 15, 2014 5:02 am

]]>

New features in this release include:

- Split off mode parameters (l, m, X_n_pg_min and X_n_pg_max) from the &osc namelist; they are now specified in their own &mode namelist(s).
- Added support for the Doppler shift due to (possibly-differential) rotation; set m to a non-zero value to enable this.
- Added the ability to do non-adiabatic calculations in the MESA interface module.
- Added new regularize flag to the &model namelist; this regularizes an input model, to try to iron out inconsistencies in hydrostatic equilibrium and mass conservation.
- Added support for LOSC-format stellar models

As usual, the source code can be downloaded from

https://bitbucket.org/rhdtownsend/gyre/downloads/

Rich

Statistics: Posted by rhtownsend — Fri May 02, 2014 4:22 am

]]>

I have attached a figure showing two examples of this fake glitch, and also attach the input models. The x-axis is the relative radius.

Can you please help me resolve this?

Cheers

Ehsan.

Statistics: Posted by ehsan — Sat Apr 26, 2014 5:17 pm

]]>