uvex_transients.dust.log_attenuation#

uvex_transients.dust.log_attenuation(nu: Quantity, Ebv: float | Quantity | NDArray[float64], dust_law: str | None = None) → NDArray[float64][source]#

Natural log of the Milky Way foreground attenuation, \(\ln(10^{-0.4\,A(\nu)})\), at nu.

Takes an already-resolved E(B-V) – resolving one from a dust map and a sky position is a separate, prior step (resolve_ebv), deliberately kept out of this function: fixing Ebv’s shape once, up front, is what lets this stay a single, fully vectorized computation (no per-position or per-frequency Python loop, and, unlike building a ~synphot.SpectralElement per position, no repeated ~astropy.modeling.Model construction either) – the reddening law’s dimensionless shape \(A(\nu)/A(V)\) is evaluated once at every nu sample (shape (K,)), and combined with Ebv (shape S) by plain numpy broadcasting into a result of shape S + (K,).

Add this directly to a ~uvex_transients.models.core.base.SpectralModel flux calculation’s log flux (its log_attenuation keyword) – e.g., having already called ebv = resolve_ebv(dust_map(), coord) once, model.flux_log(nu, t, ..., log_attenuation=log_attenuation(nu, ebv)) – rather than multiplying a linear transmission fraction in, since everything on that side is already computed in log space.

SpectralModel.as_source_spectrum/as_astropy_model take the same shape one level removed, as a callable (their wavelength grid isn’t known until the resulting ~synphot.SourceSpectrum is actually called): use attenuation_callable rather than binding this function yourself, e.g. model.as_source_spectrum(t, ..., log_attenuation=attenuation_callable(ebv)).

Parameters:

  • nu (Quantity) – Frequency(ies) (or frequency-equivalent, e.g. wavelength) to evaluate the reddening law at, any shape (K,) (or scalar, in which case the trailing frequency axis below is squeezed away).

  • Ebv (float, Quantity, or numpy.ndarray) – Already-resolved, dimensionless \(E(B-V)\), any shape S – see resolve_ebv for turning a dust map + sky position into this.

  • dust_law (str, optional) – Passed through to get_dust_law; None (the default) uses the configured default reddening law.

Returns:

Natural log of the dimensionless attenuation, NaN wherever nu falls outside the reddening law’s native range (roughly 900 Angstrom to 32 microns for the default, dust_extinction.parameter_averages.G23). Shape S + (K,), or plain S if nu was scalar.

Return type:

numpy.ndarray