Choosing BASTA run controls

Fitting options for global seismic and classical surface parameters

Choose the columns from your stellar data file to include in the fit. Note that all fit parameters use the same names as in the stellar data file except for the large separation, Δν. If your uploaded data contains a generic “dnu” column, you must choose here which Δν definition to use: dnufit, dnufitMos12, dnuAsf, dnuscal, or dnuSer. See §4.1.1 of BASTA II for definitions.

Fitting options for individual oscillation frequencies

Default fit — Fit the raw individual frequencies: “freqs”. Automatically fits and applies a surface correction term to the frequencies of the model, following the prescription chosen by the user.

Surface-insensitive diagnostics — Instead of applying an ad-hoc surface correction to the frequencies of the model to deal with the inadequacies of the model near the surface, one can instead use combinations of frequencies that cancel out the contributions from the surface layers. There are two types of these combinations:

Epsilon differences — “e01”, “e02”, “e012” - are frequency phase shift differences, utilizing that the difference between the phase shifts of individual frequencies of different spherical degree effectively cancels out all contributions from the surface. Developed by Ian Roxburgh, it is an improved version of the frequency ratios fitting method, and thus the recommended option. The numbers in the keys refer to the spherical degrees of the observed frequencies to be used to construct the phase shift differences, thus if \(\ell=0,1,2\) is available, it is recommended to use the full set of phase shift differences ”e012”. To be fully documented in Winther et al. in preparation.
Frequency ratios — “r02”, “r01”, “r10”, “r012””, r102” or “r010” - are the frequency ratios between the small and large frequency separations formed by combinations of frequencies that cancel out the surface contribution. BASTA uses the 5-point formulation of the small frequency separation. The numbers in the keys refer to the spherical degree of the observed frequencies to be used to construct the frequency ratios, thus if \(\ell=0,1,2\) is available, it is recommended to use either ”r012” or ”r102”. Using ”r010” is discouraged as it can lead to overfitting (Roxburgh 2018).

Glitch options — Glitches are small variations from regularity in pressure mode frequencies, caused by localized structural transitions (e.g., He partial-ionization layers in the outer envelope). BASTA can deal with “glitches” directly by fitting an oscillatory term on a smooth polynomial background to the observed pressure mode frequencies. This returns the glitch’s mean amplitude, acoustic width, and acoustic depth.

Glitches can also be fit jointly with five-point frequency ratios, using a single likelihood with full covariance: “gr010”, “gr02”, “gr01”, “gr10”, “gr012”, “gr102”.

Search range limits

Search range limits can be applied to speed up runtime, so that likelihoods will be calculated only for models within a certain range from the observed values of each star. Useful when fitting to an extensive grid of models that covers a large span in evolutionary phases.

Grid selection

Select the theoretical models you want to fit against. BASTA uses grids of stellar tracks or isochrones stored in HDF5 format. Note that not all parameters are available in every grid. For example, the BaSTI isochrone grids do not include individual oscillation frequencies whereas most stellar track grids do. A summary of a grid’s content and parameters is displayed when you select the grid. Note that grid resolution for Sobol sampling can not be meaningfully compared to that of uniform sampling (See MNRAS,525,1416 (2023) Table1+Appendix C for an example of Sobol sampling of resolution 3.8). If you would like other grids included on the site, please contact us.

Asteroseismic options

Specify the solar reference values of Δν and ν*max you wish to use. By default, BASTA adopts the SYD asteroseismic pipeline values of \(\Delta\nu=135.1\mu\mathrm{Hz}\) and \(\nu*\mathrm{max}=3090\mu\mathrm{Hz}\).

The following options appears only if frequency files were uploaded on the first page.

Choose the surface-effect correction: (i) the two-term Ball & Gizon (2014) correction, (ii) the one-term (“cubic”) Ball & Gizon (2014) correction, (iii) the Kjeldsen et al. (2008) power law, or (iv) no correction. It is recommended to use (i) or (ii) depending on the data. (i) Is a versatile formulation that makes it possible to fit most stars, but can sometimes over-correct. (ii) Is the most realistic formulation that can not over-correct, but requires accurate frequencies and a dense grid. Use “no correction” only for testing/validation, or if your input frequencies have already been corrected.
You can also enable correlations between frequencies/ratios – this should always be set to “Yes” for surface-independent fitting (ratios or epsilon differences).
Define the seismic frequencies weighting to balance the fit, so classical and global seismic observables still carry weight. Typically, the seismic χ² term (from individual frequencies) is divided by a factor before being added to the total χ². Options are (i) 1/N (default; divide/normalize by the total number of frequencies or ratios), or (ii) 1/1 (no scaling; each frequency counts as one classical and global seismic parameter). Specifically, in the case of no correlations, where the classical observations \(i\) are fitted along with the individual frequencies/surface independent measures \(j\), the total χ² is calculated as \(\chi^2 = \sum_i \left(\frac{O_{i,\mathrm{obs}}-O_{i,\mathrm{mod}}}{\sigma_i}\right)^2 + w\sum_j^N \left(\frac{O_{j,\mathrm{obs}}-O_{j,\mathrm{mod}}}{\sigma_j}\right)^2\) where \(w\) is the seismic weight specified here.

Statistical Output

BASTA is designed to work with the median and corresponding Bayesian credibility intervals or quantiles (16th and 84th percentile). Thus, by default BASTA will report the median of the posterior distribution as the parameter value and the quantiles as the (asymmetric) errors.
For some applications (e.g. to compare with other results), it can be useful to instead report the mean and standard deviation of the distributions. Note that this only is reasonable for normal distributions – please inspect/verify with the corner plots.