kinextract.build_template_matrix_fortran

kinextract.build_template_matrix_fortran(xg: ndarray, template_paths: list[str]) → tuple[ndarray, ndarray, ndarray, ndarray][source]

Build the full (npix, ntemplates) stellar template matrix.

Loads every template listed in template_paths, median-normalizes each to a flux level of approximately 1.0, and interpolates it onto the galaxy wavelength grid xg. The resulting matrix is what kinextract.numerics convolves with a trial LOSVD (one column at a time, weighted by a fitted per-template amplitude) to build the model spectrum for chi-squared minimization.

Median normalization is required when templates are supplied in physical flux units (e.g. MUSE spectral library .dat files in erg cm⁻² s⁻¹ Å⁻¹) so their scale matches the galaxy spectrum, which is itself close to unity after ALS or pre-normalization.

Parameters:

  • xg (ndarray) – Galaxy wavelength grid, in Angstroms, onto which every template is resampled.

  • template_paths (list of str) – Paths to individual template spectrum files (2- or 3-column wavelength/flux[/flux_err] text files), typically produced by kinextract.io.read_template_list().

Returns:

  • T (ndarray, shape (npix, ntemplates)) – Template matrix on the galaxy grid, each column normalized so its flux values are of order 1.

  • T_err (ndarray, shape (npix, ntemplates)) – Per-pixel template flux uncertainty, normalized identically to T. Columns are all-zero for templates whose file had no error column.

  • outside_each (ndarray of bool, shape (npix, ntemplates)) – True where a given template has no wavelength coverage at a given pixel – used to exclude just that template (not the whole pixel) from the per-pixel mixture when template libraries have slightly different native wavelength ranges (the common case for real libraries assembled from more than one source/run).

  • outside_all (ndarray of bool, shape (npix,)) – True at pixels that fall outside the wavelength coverage of every template in the list; such pixels carry no template information at all and are typically masked out of the fit (gerr set to BIG).