sheap.Profiles.profiles_templates module

Template-Based Profiles

This module provides template-driven spectral components used in sheap:

  • Fe II templates (UV, optical, combined) read from ASCII files and broadened

    via FFT convolution.

  • Balmer high-order blends represented as fixed templates.

  • Host galaxy templates based on E-MILES SSP cubes, sub-selected in metallicity,

    age, and wavelength, and combined with free weights.

Functions

  • make_feii_template_function :

    Factory for Fe II template models by name. Supports optional wavelength cuts and returns a JAX-ready profile function plus template metadata.

  • make_host_function :

    Factory for host galaxy models from a precomputed SSP cube. Uses efficient memory mapping and a single FFT-based convolution of the weighted template sum.

Constants

  • TEMPLATES_PATH : Path to the bundled template data directory.

  • FEII_TEMPLATES : Registry of available Fe II template definitions.

Notes

  • All returned models are decorated with @with_param_names and are JAX-compatible.

  • FFT-based Gaussian broadening quadratically subtracts the intrinsic template

    resolution before applying user-defined FWHM.

  • Host models build parameter names dynamically as weight_Z{Z}_age{age}

    for each included SSP grid point.

make_host_function(filename=PosixPath('/home/docs/checkouts/readthedocs.org/user_builds/sheap/checkouts/latest/sheap/SuportData/templates/miles_cube_log.npz'), z_include=[-0.7, 0.22], age_include=[0.1, 10.0], xmin=None, xmax=None, verbose=None, **kwargs)[source]

Factory for host-galaxy SSP models stored on a log-sampled wavelength grid.

The returned model builds a weighted sum of SSP templates, then applies a single FFT-based Gaussian broadening in velocity space (LOSVD-like), and finally interpolates to the user grid x.

Parameters:

  • filename (str | Path) – Either "miles", "xsl", a path to a *_cube_log.npz file, or None (defaults to MILES).

  • z_include (tuple[float, float] | list[float] | None) – Metallicity selection; either (min,max) or list/array.

  • age_include (tuple[float, float] | list[float] | None) – Age selection; either (min,max) or list/array.

  • xmin (float | None) – Optional wavelength window in Angstrom. A ±50 Å guard band is applied.

  • xmax (float | None) – Optional wavelength window in Angstrom. A ±50 Å guard band is applied.

  • verbose (bool | None) – Print selected grid size.

Returns:

{"model": callable, "host_info": dict}

Return type:

dict

Notes

  • Important: The template wavelength axis in the NPZ file must be sampled approximately uniformly in \(\ln(\lambda)\) (i.e. log-spaced in wavelength). The convolution kernel is constructed in \(\ln(\lambda)\), so the model assumes a constant velocity step per pixel.

  • The input x passed to model(x, params) is in Angstrom, but should be sampled on a grid that is also approximately uniform in \(\ln(\lambda)\) if you want the broadening to behave exactly like a Gaussian LOSVD on your evaluation grid. (Interpolation to arbitrary x is allowed, but the kernel is defined on the template log-grid.)

  • Velocity shift is applied as \(\lambda \rightarrow \lambda \exp(v/c)\).

The parameter order is: [logamp, logFWHM, vshift_kms, weight_0, weight_1, ...].