synthesizer.emission_models.transformers.dust_attenuation

A module containing dust attenuation functionality.

This module contains classes for dust attenuation laws. These classes provide a way to calculate the optical depth and transmission curves for a given dust model.

Example usage:

# Create a power law dust model with a slope of -1.0
dust_model = PowerLaw(slope=-1.0)

# Calculate the transmission curve
transmission = dust_model.get_transmission(
    0.33, np.linspace(1000, 10000, 1000)
)

Functions

synthesizer.emission_models.transformers.dust_attenuation.Li08(lam, UV_slope, OPT_NIR_slope, FUV_slope, bump, model)[source]

Drude-like parametric expression for the attenuation curve from Li+08.

Parameters:

  • lam (np.ndarray of float) – The wavelengths (AA units) at which to calculate transmission.

  • UV_slope (float) – Dimensionless parameter describing the UV-FUV slope

  • OPT_NIR_slope (float) – Dimensionless parameter describing the optical/NIR slope

  • FUV_slope (float) – Dimensionless parameter describing the FUV slope

  • bump (float) – Dimensionless parameter describing the UV bump strength (0< bump <1)

  • model (str) – Via this parameter one can choose one of the templates for extinction/attenuation curves from: Calzetti, SMC, MW (R_V=3.1), and LMC

Returns:

tau/tau_v at each input wavelength (lam)

Return type:

np.ndarray of float/float

synthesizer.emission_models.transformers.dust_attenuation.N09Tau(lam, slope, cent_lam, ampl, gamma)[source]

Generate the transmission curve for the Noll+2009 attenuation curve.

Attenuation curve using a modified version of the Calzetti attenuation (Calzetti+2000) law allowing for a varying UV slope and the presence of a UV bump; from Noll+2009

References

https://ui.adsabs.harvard.edu/abs/2009A%26A…499…69N

Parameters:

  • lam (np.ndarray of float) – The input wavelength array (expected in AA units, global unit).

  • slope (float) – The slope of the attenuation curve.

  • cent_lam (float) – The central wavelength of the UV bump, expected in AA.

  • ampl (float) – The amplitude of the UV-bump.

  • gamma (float) – The width (FWHM) of the UV bump, in AA.

Returns:

V-band normalised optical depth for

given wavelength

Return type:

np.ndarray of float

synthesizer.emission_models.transformers.dust_attenuation.sb26_tau(lam, B_0, B_1s, B_2s, B_3)[source]

Compute V-band normalised optical depth for Sommovigo & Bartlett 2026.

Four-parameter attenuation curve model calibrated on TNG50/TNG100 radiative transfer simulations (SKIRT) across MW, SMC, and stellar dust models. The model consists of three additive components: a UV bump at ~2175 AA, a UV slope with exponential decay toward the optical, and a curvature term controlling far-UV and optical/IR behaviour.

At the V-band wavelength (5542 AA), the curve evaluates to unity by construction.

References

Sommovigo, Bartlett et al. 2026 (https://arxiv.org/abs/2606.10027)

Parameters:

  • lam (np.ndarray of float) – The wavelengths (AA units) at which to calculate the normalised optical depth.

  • B_0 (float) – UV bump amplitude. Typical range (0, 2.5); strongly dust-model dependent: ~1.2 for MW, ~0 for SMC/stellar.

  • B_1s (float) – Linear slope term, scaled by 1e-3. Typical range (-0.15, 0.10).

  • B_2s (float) – Slope modulation term, scaled by 1e-3. Typical range (-2.5, 1.1).

  • B_3 (float) – Exponential/curvature parameter. Typical range (0.8, 5.1).

Returns:

V-band normalised optical depth (A_lambda / A_V).

Return type:

np.ndarray of float

Classes

class synthesizer.emission_models.transformers.dust_attenuation.AttenuationLaw(description, required_params=('tau_v',), require_tau_v=True)[source]

The base class for all attenuation laws.

A child of this class should define its own get_tau method with any model specific behaviours. This will be used by get_transmission (which itself can be overloaded by the child if needed).

description

A description of the type of model. Defined on children classes.

Type:

str

required_params

The name of any required parameters needed by the transformer when transforming an emission. These should either be available from an emitter or from the EmissionModel itself. If they are missing an exception will be raised.

Type:

tuple

get_extinction_curve(lam, **dust_curve_kwargs)[source]

Compute the normalised extinction curve tau(lambda)/tau(V).

Returns the wavelength-dependent extinction curve from the underlying attenuation law at the requested wavelengths. This is the tau_x_v factor used in separable attenuation:

transmission = exp(-tau_v * tau_x_v)

where tau_v is the V-band optical depth supplied separately by the caller. This method only returns the normalised curve; the caller is responsible for providing tau_v and computing the exponential.

Parameters:

  • lam (np.ndarray of float) – The wavelengths (with units) at which to calculate the normalised extinction curve.

  • **dust_curve_kwargs (dict) – Additional keyword arguments to be passed to the dust curve which have been defined on the emitter or model. These are applied to a copy of this curve for the computation, leaving this curve unchanged.

Returns:

The normalised extinction curve tau(lambda)/tau(V) with shape lam.shape.

Return type:

np.ndarray of float

get_tau(*args)[source]

Compute the V-band normalised optical depth.

get_tau_at_lam(*args)[source]

Compute the optical depth at wavelength.

get_transmission(tau_v, lam, **dust_curve_kwargs)[source]

Compute the transmission curve.

Returns the transmitted flux/luminosity fraction based on an optical depth at a range of wavelengths.

Parameters:

  • tau_v (float/np.ndarray of float) – Optical depth in the V-band. Can either be a single float or array.

  • lam (np.ndarray of float) – The wavelengths (with units) at which to calculate transmission.

  • **dust_curve_kwargs (dict) – Additional keyword arguments to be passed to the dust curve which have been defined on the emitter or model.

Returns:

The transmission at each wavelength. Either (lam.size,) in shape for singular tau_v values or (tau_v.size, lam.size) tau_v is an array.

Return type:

np.ndarray of float

plot_attenuation(lam, fig=None, ax=None, label=None, figsize=(8, 6), show=True, **kwargs)[source]

Plot the attenuation curve.

Parameters:

  • lam (np.ndarray of float) – The wavelengths (with units) at which to calculate transmission.

  • fig (matplotlib.figure.Figure) – The figure to plot on. If None, a new figure will be created.

  • ax (matplotlib.axes.Axes) – The axis to plot on. If None, a new axis will be created.

  • label (str) – The label to use for the plot.

  • figsize (tuple) – The size of the figure to create if fig is None.

  • show (bool) – Whether to show the plot.

  • **kwargs (dict) – Keyword arguments to be provided to the plot call.

Returns:

The figure and axis objects.

Return type:

fig, ax

plot_transmission(tau_v, lam, fig=None, ax=None, label=None, figsize=(8, 6), show=True)[source]

Plot the transmission curve.

Parameters:

  • tau_v (float/np.ndarray of float) – Optical depth in the V-band. Can either be a single float or array.

  • lam (np.ndarray of float) – The wavelengths (with units) at which to calculate transmission.

  • fig (matplotlib.figure.Figure) – The figure to plot on. If None, a new figure will be created.

  • ax (matplotlib.axes.Axes) – The axis to plot on. If None, a new axis will be created.

  • label (str) – The label to use for the plot.

  • figsize (tuple) – The size of the figure to create if fig is None.

  • show (bool) – Whether to show the plot.

Returns:

The figure and axis objects.

Return type:

fig, ax

Examples using synthesizer.emission_models.transformers.dust_attenuation.AttenuationLaw

Camels example

Camels example

SC-SAM example

SC-SAM example

Draine-Li attenuation validation

Draine-Li attenuation validation

Sommovigo & Bartlett 2026 attenuation

Sommovigo & Bartlett 2026 attenuation

Extreme model variations example

Extreme model variations example

Plot model variations example

Plot model variations example

Dust curves example

Dust curves example

Photometry example

Photometry example

Generating Lines from a Parametric Galaxy

Generating Lines from a Parametric Galaxy

Generate parametric observed SED

Generate parametric observed SED

An example showing how to scale a galaxy’s mass by luminosity/flux.

An example showing how to scale a galaxy's mass by luminosity/flux.

Generate parametric galaxy SED

Generate parametric galaxy SED

Plot line of sight diagnostics

Plot line of sight diagnostics
class synthesizer.emission_models.transformers.dust_attenuation.Calzetti2000(slope=0, cent_lam=unyt_quantity(2175, 'Å'), ampl=0, gamma=unyt_quantity(350, 'Å'))[source]

Calzetti attenuation curve.

This includes options for the slope and UV-bump implemented in Noll et al. 2009.

slope

The slope of the attenuation curve.

Type:

float

cent_lam

The central wavelength of the UV bump, expected in AA.

Type:

float

ampl

The amplitude of the UV-bump.

Type:

float

gamma

The width (FWHM) of the UV bump, in AA.

Type:

float

get_tau(lam)[source]

Calculate V-band normalised optical depth.

(Uses the N09Tau function defined above.)

Parameters:

lam (float/np.ndarray of float) – An array of wavelengths or a single wavlength at which to calculate optical depths (in AA, global unit).

Returns:

The V-band noramlised optical depth.

Return type:

float/array-like, float

get_tau_at_lam(lam)[source]

Calculate optical depth at a wavelength.

(Uses the N09Tau function defined above.)

Parameters:

lam (float/array-like, float) – An array of wavelengths or a single wavlength at which to calculate optical depths (in AA, global unit).

Returns:

float/array-like, float

The optical depth.

class synthesizer.emission_models.transformers.dust_attenuation.DraineLiGrainCurves(grid_name, grid_dir, grain_dict=None, lam=None)[source]

Draine and Li extinction curves.

Draine and Li extinction curves obtained from pre-processing the extinction efficiencies for the required grain size distribution. This is done in the syncretize repo under ‘syncretize/src/synthesizer_grids/dust/ create_dustextcurve_draine_li.py’ for the required dust parameters. Currently only implemented for 2 grain sizes of graphites and silicates, and 1 size of PAHs.

grid_name

Name of the extinction curve grid (without hdf5 extension)

Type:

string

grid_dir

Location of the grid

Type:

string

grain_dict

Dictionary containing the grain type (‘graphite’ or ‘silicate’ or ‘pahionised’ or ‘pahneutral’) and their corresponding centre of the grain size distribution in microns. E.g. grain_dict = {‘graphite’: [0.01, 0.1]}

Type:

Dict

get_tau(lam, sigmalos_H=None, **sigmalos_dust)[source]

Calculate optical depth normalised at V-band at a wavelength.

Parameters:

  • lam (float/array-like, float) – An array of wavelengths or a single wavlength at which to calculate optical depths (in AA, global unit).

  • sigmalos_H (unyt array) – Line-of-sight H density in units of Msun/pc^2

  • (Dict (sigmalos_dust) – unyt_array): Dictionary containing the different line-of-sight dust density of the dust components. This should follow the format ‘sigmalos_{grain_type}_a{grain_bin_centre}um’: unyt_array (units of Msun/pc^2). The function will unpack the contents.

Returns:

optical depth/optical depth at V-band (unyt_array, float)

The optical depth at the given wavlength ((N_dtg, N_lambda) if sigmalos input is array-like, otherwise shape (N_lambda,)). Dimensionless

get_tau_at_lam(lam, sigmalos_H, **sigmalos_dust)[source]

Calculate optical depth at a wavelength.

Parameters:

  • lam (unyt_array, float) – An array of wavelengths or a single wavelength at which to calculate optical depths (in AA, global unit).

  • sigmalos_H (unyt array) – Line-of-sight H density in units of Msun/pc^2

  • (Dict (sigmalos_dust) – unyt_array): Dictionary containing the different line-of-sight dust density of the dust components. This should follow the format ‘sigmalos_{grain_type}_a{grain_bin_centre}um’: unyt_array (units of Msun/pc^2). The function will unpack the contents.

Returns:

optical depth (unyt_array, float)

The optical depth at the given wavlength ((N_dtg, N_lambda) if sigmalos input is array-like, otherwise shape (N_lambda,)). Dimensionless

get_transmission(tau_v=None, lam=None, **dust_curve_kwargs)[source]

Compute the transmission curve.

Returns the transmitted flux/luminosity fraction based on an optical

depth at a range of wavelengths.

Parameters:

  • tau_v (None) – Optical depth at V-band, not required here.

  • lam (np.ndarray of float) – The wavelengths (with units) at which to calculate transmission.

  • **dust_curve_kwargs (dict) – Additional keyword arguments to be passed to the dust curve which have been defined on the emitter or model.

Returns:

The transmission at each wavelength

Return type:

np.ndarray of float

class synthesizer.emission_models.transformers.dust_attenuation.GrainModels(model='WD01', submodel='SMCBar')[source]

Grain model dust attenuation curves.

These models are based on dust grain size, composition, and shape distributions constrained by observations of extinction, abundances, emission, and polarization. Some of these models can be used to estimate extinction at wavelengths inaccessible to observations (e.g., extreme UV below 912 Å). These models are taken from the astropy affiliated dust-extinction package (https://dust-extinction.readthedocs.io/en/latest/ for details). By default, we will use the Weingarter & Draine 2001 (WD01) models, and the submodel for the SMC bar.)

model

The dust grain model used. Available models are:

DBP90: Desert, Boulanger, & Puget 1990, A&A, 237, 215 WD01: Weingartner & Draine 2001, ApJ, 548, 296 D03: Draine 2003, ARA&A, 41, 241; Draine 2003, ApJ, 598, 1017 ZDA04: Zubko, Dwek, & Arendt 2004, ApJS, 152, 211 C11: Compiegne et al. 2011, A&A, 525, 103 J13: Jones et al. 2013, A&A, 558, 62 HD23: Hensley & Draine 2023, ApJ, 948, 55 Y24: Ysard et al. 2024, A&A, 684, 34

Type:

str

submodel

The submodel to use within the main grain model. The submodels available for the different models listed below. All of them are self-explanatory with the RV defining the normalisation of the extinction curve, where RV = AV / E(B-V). DBP90: MWRV31 WD01 MWRV31, MWRV40, MWRV55, LMCAvg, LMC2, SMCBar D03: MWRV31, MWRV40, MWRV55 ZDA04: MWRV31 C11: MWRV31 J13: MWRV31 HD23: MWRV31 Y24: MWRV31

Type:

str

get_tau(lam, interp='slinear')[source]

Calculate V-band normalised optical depth.

Parameters:

  • lam (float/np.ndarray of float) – An array of wavelengths or a single wavlength at which to calculate optical depths (in AA, global unit).

  • interp (str) – The type of interpolation to use. Can be ‘linear’, ‘nearest’, ‘nearest-up’, ‘zero’, ‘slinear’, ‘quadratic’, ‘cubic’, ‘previous’, or ‘next’. ‘zero’, ‘slinear’, ‘quadratic’ and ‘cubic’ refer to a spline interpolation of zeroth, first, second or third order. Uses scipy.interpolate.interp1d.

Returns:

The optical depth.

Return type:

float/np.ndarray of float

get_tau_at_lam(lam, interp='slinear')[source]

Calculate optical depth at a wavelength.

Parameters:

  • lam (float/array-like, float) – An array of wavelengths or a single wavlength at which to calculate optical depths (in AA, global unit).

  • interp (str) – The type of interpolation to use. Can be ‘linear’, ‘nearest’, ‘nearest-up’, ‘zero’, ‘slinear’, ‘quadratic’, ‘cubic’, ‘previous’, or ‘next’. ‘zero’, ‘slinear’, ‘quadratic’ and ‘cubic’ refer to a spline interpolation of zeroth, first, second or third order. Uses scipy.interpolate.interp1d.

Returns:

float/array-like, float

The optical depth.

class synthesizer.emission_models.transformers.dust_attenuation.MWN18[source]

Milky Way attenuation curve used in Narayanan+2018.

data

The data describing the dust curve, loaded from MW_N18.npz.

Type:

np.ndarray of float

tau_lam_v

The V band optical depth.

Type:

float

get_tau(lam, interp='cubic')[source]

Calculate V-band normalised optical depth.

The result is returned at the dtype of lam. Note that the spline interpolation kinds are evaluated in double precision inside SciPy whatever they are given, so a float32 request is computed in double and narrowed on return rather than being computed in float32 throughout. The answer is therefore slightly more accurate than a true float32 evaluation would be, not less, and the tabulated curve is bounded well inside the float32 range so narrowing cannot overflow.

Parameters:

  • lam (float/array, float) – An array of wavelengths or a single wavlength at which to calculate optical depths (in AA, global unit).

  • interp (str) – The type of interpolation to use. Can be ‘linear’, ‘nearest’, ‘nearest-up’, ‘zero’, ‘slinear’, ‘quadratic’, ‘cubic’, ‘previous’, or ‘next’. ‘zero’, ‘slinear’, ‘quadratic’ and ‘cubic’ refer to a spline interpolation of zeroth, first, second or third order. Uses scipy.interpolate.interp1d.

Returns:

The optical depth, at lam’s dtype.

Return type:

float/array, float

get_tau_at_lam(lam, interp='cubic')[source]

Calculate the optical depth at a wavelength.

Parameters:

  • lam (float/array, float) – An array of wavelengths or a single wavlength at which to calculate optical depths (in AA, global unit).

  • interp (str) – The type of interpolation to use. Can be ‘linear’, ‘nearest’, ‘nearest-up’, ‘zero’, ‘slinear’, ‘quadratic’, ‘cubic’, ‘previous’, or ‘next’. ‘zero’, ‘slinear’, ‘quadratic’ and ‘cubic’ refer to a spline interpolation of zeroth, first, second or third order. Uses scipy.interpolate.interp1d.

Returns:

float/array, float

The optical depth.

class synthesizer.emission_models.transformers.dust_attenuation.ParametricLi08(UV_slope=44.9, OPT_NIR_slope=7.56, FUV_slope=61.2, bump=0.0, model='Calzetti')[source]

Parametric, empirical attenuation curve.

Implemented in Li+08, Evolution of the parameters up to high-z (z=12) studied in: Markov+23a,b

UV_slope

Dimensionless parameter describing the UV-FUV slope (0, 50)

Type:

float

OPT_NIR_slope

Dimensionless parameter describing the optical/NIR slope (0, 10)

Type:

float

FUV_slope

Dimensionless parameter describing the FUV slope (-1, 75)

Type:

float

bump

Dimensionless parameter describing the UV bump strength (-0.005< bump <0.06)

Type:

float

model

Fixing attenuation/extinction curve to one of the known templates: MW, SMC, LMC, Calzetti

Type:

str

get_tau(lam)[source]

Calculate V-band normalised optical depth.

(Uses the Li_08 function defined above.)

Parameters:

lam (float/np.ndarray of float) – An array of wavelengths or a single wavlength at which to calculate optical depths (in AA, global unit).

Returns:

The V-band normalised optical depth.

Return type:

float/array-like, float

get_tau_at_lam(lam)[source]

Calculate optical depth at a wavelength.

(Uses the Li_08 function defined above.)

Parameters:

lam (float/array-like, float) – An array of wavelengths or a single wavlength at which to calculate optical depths (in AA, global unit).

Returns:

float/array-like, float

The optical depth.

class synthesizer.emission_models.transformers.dust_attenuation.PowerLaw(slope=-1.0)[source]

Custom power law dust curve.

slope

The slope of the power law.

Type:

float

get_tau(lam)[source]

Calculate V-band normalised optical depth.

Parameters:

lam (float/np.ndarray of float) – An array of wavelengths or a single wavlength at which to calculate optical depths.

Returns:

The optical depth.

Return type:

float/np.ndarray of float

get_tau_at_lam(lam)[source]

Calculate optical depth at a wavelength.

Parameters:

lam (float/np.ndarray of float) – An array of wavelengths or a single wavlength at which to calculate optical depths.

Returns:

The optical depth.

Return type:

float/np.ndarray of float

class synthesizer.emission_models.transformers.dust_attenuation.SommovigoBartlett2026(B_0='B_0', B_1s='B_1s', B_2s='B_2s', B_3='B_3')[source]

Four-parameter attenuation curve from Sommovigo & Bartlett 2026.

Attenuation curve model calibrated on TNG50/TNG100 SKIRT radiative transfer simulations (snapnum93, z=0.07) across three dust grain models:

  • MW: Weingartner & Draine 2001 (WD01), Milky Way (R_V=3.1)

  • SMC: Weingartner & Draine 2001 (WD01), SMC Bar

  • stellar: Hirashita & Aoyama 2019, log-normal grain size distribution (a0=0.1 um, sigma=0.47)

The model consists of three additive components: a UV bump term, a UV slope with exponential decay, and a curvature/far-UV term.

The curve parameters can be predicted from galaxy physical properties using the companion function SommovigoBartlett2026.predict.

References

Sommovigo, Bartlett et al. 2026 (https://arxiv.org/abs/2606.10027) Weingartner & Draine 2001, ApJ, 548, 296 Hirashita & Aoyama 2019, MNRAS, 491, 3844

B_0

UV bump amplitude. Typical range (0, 2.5); strongly dust-model dependent: ~1.2 for MW, ~0 for SMC/stellar.

Type:

float

B_1s

Linear slope term, scaled by 1e-3. Typical range (-0.15, 0.10).

Type:

float

B_2s

Slope modulation term, scaled by 1e-3. Typical range (-2.5, 1.1).

Type:

float

B_3

Exponential/curvature parameter. Typical range (0.8, 5.1).

Type:

float

get_tau(lam)[source]

Calculate V-band normalised optical depth.

Parameters:

lam (float/np.ndarray of float) – An array of wavelengths or a single wavelength at which to calculate optical depths (in AA, global unit).

Returns:

The V-band normalised optical depth (A_lambda / A_V).

Return type:

float/np.ndarray of float

get_tau_at_lam(lam)[source]

Calculate optical depth at a wavelength.

Since the SB26 parametrisation is normalised to A_V by construction (A_lam/A_V = 1 at 5542 AA), this is identical to get_tau.

Parameters:

lam (float/np.ndarray of float) – An array of wavelengths or a single wavelength at which to calculate optical depths (in AA, global unit).

Returns:

The optical depth.

Return type:

float/np.ndarray of float

classmethod predict(sigma_sfr, inclination, z_gas, log10_mstar, ssfr, dust_model='MW', add_noise=False)[source]

Predict attenuation curve parameters from galaxy properties.

Sequential prediction pipeline mapping galaxy physical properties to the SommovigoBartlett2026 attenuation curve parameters. The prediction chain is:

A_V -> B_1s -> B_3 -> B_0 -> B_2s

Each parameter depends on galaxy properties and/or previously predicted curve parameters. Calibrated on TNG50/TNG100 SKIRT radiative transfer simulations.

References

Sommovigo, Bartlett et al. 2026 (https://arxiv.org/abs/2606.10027)

Parameters:

  • sigma_sfr (unyt_quantity/unyt_array) – Star formation rate surface density [Msun yr^-1 kpc^-2].

  • inclination (unyt_quantity/unyt_array) – Galaxy inclination angle [degrees].

  • z_gas (float/np.ndarray of float) – Gas-phase metallicity as absolute mass fraction (e.g. Zsun = 0.0142, Asplund et al. 2009).

  • log10_mstar (float/np.ndarray of float) – Log10 stellar mass [log10(Msun)].

  • ssfr (unyt_quantity/unyt_array) – Specific star formation rate [Gyr^-1].

  • dust_model (str) – Dust mixture model. One of ‘MW’, ‘SMC’, or ‘stellar’. Selects the small-grain fraction and the B_0 prediction formula.

  • add_noise (bool) – If True, add Gaussian scatter matching the intrinsic dispersion of the calibration sample. Default is False.

Returns:

Dictionary with keys ‘A_V’, ‘B_0’, ‘B_1s’, ‘B_2s’, ‘B_3’, and ‘tau_v’ (= A_V / 1.086).

Return type:

dict

Examples

Get predicted parameters for a single galaxy:

params = SommovigoBartlett2026.predict(
    sigma_sfr=0.01 * Msun / yr / kpc**2,
    inclination=60.0 * degree,
    z_gas=0.02, log10_mstar=10.5,
    ssfr=0.5 / Gyr, dust_model='MW',
)
curve = SommovigoBartlett2026(
    B_0=params['B_0'],
    B_1s=params['B_1s'],
    B_2s=params['B_2s'],
    B_3=params['B_3'],
)
transmission = curve.get_transmission(
    params['tau_v'], lam,
)

Examples using synthesizer.emission_models.transformers.dust_attenuation.SommovigoBartlett2026

Sommovigo & Bartlett 2026 attenuation

Sommovigo & Bartlett 2026 attenuation

Dust curves example

Dust curves example