1 Scope

The main objective of this exercise is to test the behaviour of the oscillation frequencies of p-modes around nu_max with different oscillation codes. This exercise is performed on a given model for which HE (Hydrostatic Equilibrium) and BV (Brunt-Vaisala frequency).

1.1 The models

For this exercise we use models of two modelS, computed with GARSTEC (provided by Jørgen). Details and physics at the Wiki site.

Table 1. List of models for exercise #1
No Mass FeH alpha dov Nshell Phase
001 1.04 +0.00 2 0.00 4935 MS
002 1.04 +0.00 2 0.00 4935 MS

1.2 The oscillation codes

Here we list the all the participants and the main characteristics of their oscillation computation (code, numeric approach, etc.)

Table 2. Main characteristics of the codes involved in exercise #1: ver stands for the code version, EF stands for EigenFunction variable (Eulerian P’, Lagrangian dP); IS stands for Integration Scheme; Ri stands for Richardson extrapolation; IV stands for integration variable (r or r/P); Rm stands for remeshing; G is the Universal gravitation constant (format D-8), and Prov stands for provider of the frequencies (name acronym).
code ver EF IS Ri IV G Prov
1 ADIPLS ? P’ 2,4 y,n r 6.67232 JCD
2 GraCo R P’/dP 2,4 y,n r, r/P 6.67428 AM
3 Gyre 5.1 P’ 2,4,6 n r 6.67428 KD
5 LOSC 41.1 P’ 4 n r 6.67168 SS

1.3 Frequency differences (IS from 2 to 6)

Here we show a comparison of the frequencies computed by GraCo, LOSC, and GYRE codes, where limit conditions are set to \(\delta P=0\) at the surface and two different schemes for the integration variable: eulerian and lagrangian. We’ll analyse the behaviour of the different codes around the solar maximum frequency: \(\nu_{\mathrm{max}} = 3090\,\mu{\mathrm{Hz}}\) (around \(n=20-21\)). Note the different integration schemes in this set of computations: GraCo (2), LOSC (4), and GYRE (6).

1.3.1 Model 1-001

We see that whil LOSC and GYRE behaves very similarly, there is a signficiant difference with GraCo, in particular for \(\ell = 0\) modes. Having verified, that outer boundary conditios are well \(\delta P\) for all the computations, we conclude that, as found in ESTA/CoRoT exercises, a O(2) integration scheme produces too large differences even for 4K models.

1.3.2 Model 1-002

For model 1-002, results are equivalent to those of 1-001, except for the general behaviour of frequency differences, which are smoother (no ridges are present). In order to ensure good compatibility, i.e. \(\Delta f < 0.01\) for the whole p-mode domain (as shown in LOSC-GYRE comparisons), we recalculate the frequencies but using the same integration scheme in all the codes.

1.4 Frequency differences (IS = 4)

As in the previous section, we show here the same comparison of frequency differences for models 1-001 and 1-002 obtained with GraCo, LOSC, and GYRE codes, where outer boundary condition is set to \(\delta P=0\), and two different schemes for the integration variable: eulerian and lagrangian. However in this case, all frequencies are computed following a fourth order integration scheme. In the case of GraCo this is done using Richarson extrapolation for the eulerian case, and complete 4th-order scheme for the Lagrangian case.

1.4.1 Model 1-001

1.4.2 Model 1-002

1.5 Frequency differences (IS = 2)

In order to better understand the behavior of GraCo frequencies, we also compare the differences for the second-order scheme, between GraCo and GYRE (LOSC seem to have the option for second-order but only for 4th-order)

1.5.1 Model 1-001

Even at second order the behavior of GraCo’s \(\ell=0\) modes is quite different and have nothing to do (apparently with the outer boundary condition, which is the same as the other codes) but the implementation of the radial modes computation itself. On the other hand the largest differences for \(\ell=1,2,3\) are of the order of \(0.06\,\mu\mathrm{Hz}\) (for the highest radial order) and around \(0.04\,\mu\mathrm{Hz}\) around \(\nu_{\mathrm{max}}\). This, in return, is the average difference for the whole range betwee GYRE and LOSC.

1.5.2 Model 1-002

As for the other IS, this second model behaves (overall) better than 1-001, with similar numerical results

1.6 Large separation

Following the same procedure we compare the large separation for the different mode degrees. Discarding an IS of order 2, and thus following ESTA/CoRoT recommendations, consider at least IS(4), the following comparision will only be done for IS(4).

1.6.1 Model 1-001

As expected the differences of large separations with GraCo is larger for radial modes (insets depicts the differences only for non-radial modes). As for the frequencies, such differences grow with \(n\). However, for the large separation, such differences are smaller than those found for the individual frequencies.

1.6.2 Model 1-002

The comparison confirms the results obtained for the individual frequency comparisons. Except for the radial modes, for which differences in the large separation reach \(0.015 \mu\mathrm{Hz}\) for the highest orders, the separations for the non-radial modes span between \(10^{-5}\) up to \(10^{-3}\mu\mathrm{Hz}\).

2 Preliminary conclusions

  1. Both frequency and large separation differences are around \(10^{-3} - 10^{-2}\) for \(\ell>0\) modes and may reach 0.15-0.2 when radial modes are included. This behavior is only found for comparisons with GraCo (see complete statistics in Appendices).
  2. IS(4) (or equivalent using Richardson extrapolation) is optimum (as suggested by ESTA/CoRoT exercises). We do thus suggest to freeze this parameter from now on.
  3. Model 1-002 shows a better overall behavior of the frequencies and large separations, thus we encourage to promote the use of full analytic CEFF in the model computations.
  4. Mode identification shows anomalies at low-\(n\) (between 1 and 15). It is not an issue for higher frequencies. This however might be a problem when solar-like stars different from the Sun (and definitely not so well tuned as modelS), and for evolved stages with the presence of mixed modes. So our recommendation is that the chosen code must be robust in mode identification.
  5. Missing modes at low-frequency range (\(n=[1,10]\)) is another important issue for some codes (in this exercise: GraCo and LOSC). This might be solved when remeshing is performed on the models.

3 Appendice

In order to have a quantitative view of the different comparisons, we have calculated for each \(\ell\) the following information

  • the mean and median of the differences in the whole range
  • the mean and median of the differences at \(\nu_{\mathrm{max}}\pm \nu(\pm 5n)\)
  • the minimum difference and the \(n_i\) where this minimum is reached
  • the maximum difference and the \(n_i\) where this minimum is reached

4 Stats for frequency differences

Table 3. 1-001. Frequencies. IS(4)
Mean Median n_min Df(n_min) n_max Df(n_max) L
GYRE-GRACO (Eul)
8.212e-02 6.131e-02 5 7.347e-03 34 2.320e-01 0
2.879e-04 1.643e-04 5 1.665e-05 34 1.039e-03 1
2.948e-05 2.975e-05 30 1.459e-06 23 5.366e-05 2
3.171e-05 3.359e-05 31 2.868e-06 24 5.360e-05 3
GYRE-GRACO (Lag)
1.161e-01 1.076e-01 18 2.324e-02 34 2.304e-01 0
9.563e-03 6.659e-03 30 7.648e-04 18 3.092e-02 1
9.245e-03 4.786e-03 32 1.828e-04 17 2.874e-02 2
9.345e-03 6.629e-03 22 1.381e-03 17 3.251e-02 3
LOSC-GRACO (Eul)
7.964e-02 6.580e-02 16 7.820e-03 33 2.034e-01 0
1.704e-02 9.860e-03 21 1.300e-04 5 5.814e-02 1
1.641e-02 9.740e-03 20 1.600e-04 5 5.424e-02 2
1.530e-02 9.110e-03 13 3.000e-04 5 5.556e-02 3
LOSC-GRACO (Lag)
1.168e-01 1.086e-01 17 4.584e-02 33 2.153e-01 0
3.900e-04 3.300e-04 17 1.100e-04 33 9.000e-04 1
2.412e-05 3.000e-05 28 0.000e+00 33 5.000e-05 2
2.588e-05 3.000e-05 27 1.000e-05 33 5.000e-05 3
LOSC-GYRE (Eul)
2.102e-02 1.345e-02 27 1.270e-03 5 6.864e-02 0
2.017e-02 1.050e-02 21 3.208e-04 4 6.410e-02 1
2.001e-02 1.179e-02 20 2.043e-04 4 6.819e-02 2
1.934e-02 1.027e-02 13 2.674e-04 4 6.306e-02 3
LOSC-GYRE (Lag)
2.275e-02 1.547e-02 3 2.376e-04 7 6.786e-02 0
2.388e-02 1.730e-02 30 1.385e-03 7 6.922e-02 1
2.392e-02 1.741e-02 32 2.128e-04 6 6.921e-02 2
2.440e-02 1.901e-02 22 1.351e-03 6 7.230e-02 3
Table 3. 1-002. Frequencies. IS(4)
Mean Median n_min Df(n_min) n_max Df(n_max) L
GYRE-GRACO (Eul)
8.806e-02 6.567e-02 5 7.848e-03 34 2.489e-01 0
3.252e-04 1.817e-04 7 1.792e-05 34 1.196e-03 1
3.389e-05 3.211e-05 5 3.454e-06 22 6.323e-05 2
3.601e-05 3.744e-05 8 7.895e-06 20 6.933e-05 3
GYRE-GRACO (Lag)
1.322e-01 1.218e-01 17 4.867e-02 34 2.497e-01 0
1.140e-03 1.047e-03 24 2.781e-04 17 3.117e-03 1
5.700e-04 4.167e-04 27 9.450e-06 18 1.688e-03 2
5.670e-04 4.178e-04 34 5.689e-05 17 2.050e-03 3
LOSC-GRACO (Eul)
8.112e-02 5.857e-02 5 1.830e-03 33 2.304e-01 0
1.517e-03 1.060e-03 28 3.000e-05 6 4.640e-03 1
1.441e-03 8.100e-04 31 5.000e-05 12 3.870e-03 2
1.407e-03 8.700e-04 12 2.000e-05 6 4.730e-03 3
LOSC-GRACO (Lag)
1.249e-01 1.162e-01 17 4.892e-02 33 2.307e-01 0
4.347e-04 3.600e-04 17 1.200e-04 33 1.010e-03 1
2.529e-05 3.000e-05 28 0.000e+00 19 4.000e-05 2
2.824e-05 3.000e-05 28 0.000e+00 33 6.000e-05 3
LOSC-GYRE (Eul)
1.791e-03 1.010e-03 32 9.146e-05 3 6.602e-03 0
1.760e-03 1.251e-03 18 3.081e-05 6 4.694e-03 1
1.598e-03 9.891e-04 31 4.062e-05 4 4.395e-03 2
1.552e-03 1.015e-03 12 6.624e-05 6 4.722e-03 3
LOSC-GYRE (Lag)
2.179e-03 8.858e-04 28 4.420e-05 3 1.310e-02 0
1.669e-03 9.928e-04 24 4.194e-05 6 6.344e-03 1
1.579e-03 9.171e-04 27 1.055e-05 2 6.207e-03 2
1.456e-03 9.568e-04 19 5.795e-05 2 6.181e-03 3

5 Stats for large separations

Table 3. 1-001. Large spacings. IS(4)
Mean Median n_min Df(n_min) n_max Df(n_max) L
GYRE-GRACO (Eul)
7.533e-03 6.826e-03 5 1.377e-03 34 1.687e-02 0
3.513e-05 2.366e-05 13 4.077e-07 33 1.193e-04 1
1.202e-05 1.188e-05 7 1.168e-07 32 3.544e-05 2
1.038e-05 9.007e-06 22 4.166e-07 33 3.217e-05 3
GYRE-GRACO (Lag)
1.353e-02 1.271e-02 26 1.793e-03 17 2.633e-02 0
6.661e-03 5.465e-03 32 1.375e-04 17 2.301e-02 1
5.742e-03 5.468e-03 22 6.314e-04 17 1.629e-02 2
5.678e-03 5.218e-03 31 2.832e-04 17 1.238e-02 3
LOSC-GRACO (Eul)
1.311e-02 1.158e-02 21 1.640e-03 12 3.963e-02 0
9.202e-03 5.600e-03 9 4.600e-04 14 3.448e-02 1
9.506e-03 6.550e-03 32 1.900e-04 11 3.503e-02 2
9.450e-03 5.220e-03 18 1.600e-04 11 3.339e-02 3
LOSC-GRACO (Lag)
1.028e-02 1.028e-02 17 5.300e-03 33 1.606e-02 0
4.765e-05 4.000e-05 18 1.000e-05 33 1.000e-04 1
5.882e-06 0.000e+00 19 0.000e+00 31 2.000e-05 2
5.882e-06 1.000e-05 17 0.000e+00 29 2.000e-05 3
LOSC-GYRE (Eul)
1.010e-02 7.008e-03 33 1.258e-04 12 3.640e-02 0
9.931e-03 6.016e-03 9 4.738e-04 14 3.449e-02 1
1.054e-02 7.105e-03 32 1.546e-04 2 3.563e-02 2
9.940e-03 5.579e-03 18 1.663e-04 2 3.361e-02 3
LOSC-GYRE (Lag)
9.998e-03 7.329e-03 23 5.047e-05 14 3.598e-02 0
1.059e-02 7.885e-03 32 3.745e-05 14 3.775e-02 1
1.009e-02 6.890e-03 8 3.512e-04 13 3.482e-02 2
1.023e-02 8.351e-03 31 2.932e-04 13 3.899e-02 3
Table 3. 1-002. Large spacings. IS(4)
Mean Median n_min Df(n_min) n_max Df(n_max) L
GYRE-GRACO (Eul)
8.088e-03 7.256e-03 5 1.622e-03 34 1.815e-02 0
4.588e-05 3.615e-05 17 8.254e-07 34 1.582e-04 1
2.307e-05 2.458e-05 20 2.901e-06 5 3.947e-05 2
2.192e-05 2.150e-05 30 1.144e-06 29 4.019e-05 3
GYRE-GRACO (Lag)
1.127e-02 1.163e-02 17 1.761e-03 34 1.878e-02 0
1.227e-03 9.742e-04 30 5.567e-05 18 3.879e-03 1
1.056e-03 8.711e-04 26 1.922e-06 19 3.124e-03 2
1.159e-03 7.670e-04 30 3.117e-05 17 3.557e-03 3
LOSC-GRACO (Eul)
8.655e-03 8.600e-03 16 3.800e-04 32 1.758e-02 0
2.471e-03 2.000e-03 20 1.300e-04 7 7.780e-03 1
2.398e-03 1.490e-03 26 1.500e-04 5 7.604e-03 2
2.293e-03 1.460e-03 30 2.000e-05 7 8.030e-03 3
LOSC-GRACO (Lag)
1.103e-02 1.100e-02 17 5.740e-03 33 1.719e-02 0
5.353e-05 5.000e-05 20 1.000e-05 33 1.100e-04 1
6.471e-06 1.000e-05 17 0.000e+00 32 2.000e-05 2
7.647e-06 1.000e-05 17 0.000e+00 32 2.000e-05 3
LOSC-GYRE (Eul)
2.822e-03 1.641e-03 27 4.095e-05 4 7.635e-03 0
2.726e-03 2.155e-03 20 1.246e-04 7 7.816e-03 1
2.704e-03 2.569e-03 26 1.881e-04 5 7.643e-03 2
2.670e-03 1.646e-03 30 2.114e-05 7 8.066e-03 3
LOSC-GYRE (Lag)
3.206e-03 2.237e-03 27 1.396e-04 2 1.192e-02 0
2.802e-03 1.837e-03 16 1.453e-04 2 1.096e-02 1
2.715e-03 1.479e-03 26 1.922e-06 3 9.771e-03 2
2.609e-03 1.545e-03 30 3.117e-05 6 7.871e-03 3