Dust Models

This section details tengri’s dust attenuation and emission models.

Attenuation Framework

tengri provides a configurable dust attenuation framework with three modes of increasing physical fidelity, selectable at configuration time. All modes share the same attenuation curve registry (Attenuation Curve Library) and support the clumpy geometry of Lower et al. (2022).

Single-Component (Uniform Screen)

The simplest mode applies a single optical depth \(\tau_V\) and attenuation curve \(k(\lambda)\) identically to all stellar ages: $$T(\lambda) = f_{\rm obs} + (1 - f_{\rm obs}) \cdot \exp!\bigl[-\tau_V \cdot k(\lambda)\bigr].

(1)\[\]

Because the transmission is independent of stellar age, it factors completely out of the CSP age sum, reducing the dust computation from \(\mathcal{O}(n_{\rm age} \times n_\lambda)\) to \(\mathcal{O}(n_\lambda)\).

Two-Component Framework

The default mode follows Charlot and Fall (2000): stars are attenuated by two physically distinct dust components:

  • Birth cloud (BC): affects young stars (\(t_{\rm age} < t_{\rm birth}\)), optical depth \(\tau_{\rm BC}\).

  • Diffuse ISM: affects all stars, optical depth \(\tau_{\rm ISM}\).

Birth cloud and diffuse ISM can use different attenuation curves \(k_{\rm BC}(\lambda)\) and \(k_{\rm ISM}(\lambda)\).

Two evaluation modes are available. The fast mode (default) uses a hard age threshold at \(t_{\rm birth} = 10\) Myr and exploits a two-CSP decomposition: $$\begin{aligned}

F(\lambda) = {} & f_{\rm obs}, F_{\rm total}

  • (1 - f_{\rm obs}), T_{\rm ISM} \nonumber \ & \times \bigl[T_{\rm BC}, F_{\rm young}

  • F_{\rm old}\bigr], \end{aligned}

(2)\[\]

where \(T_{\rm BC} = \exp(-\tau_{\rm BC}\cdot k_{\rm BC})\) and \(T_{\rm ISM} = \exp(-\tau_{\rm ISM}\cdot k_{\rm ISM})\) are both 1D vectors of shape \((n_\lambda)\), and \(F_{\rm young}\)/\(F_{\rm old}\) are separate CSP age sums. No \((n_{\rm age} \times n_\lambda)\) intermediate arrays are materialized, yielding the same \(\mathcal{O}(n_\lambda)\) cost as the single-component mode.

The exact mode replaces the hard threshold with a differentiable sigmoid age transition: $$w(t_{\rm age}) = \sigma!\left(-\frac{\log_{10} t_{\rm age} - \log_{10} t_{\rm birth}}{\Delta_{\rm trans}}\right),

(3)\[\]

where \(\sigma(x) = 1/(1 + e^{-x})\) is the logistic sigmoid, \(t_{\rm birth} = 10^7\) yr (default), and \(\Delta_{\rm trans} = 0.3\) dex (default) controls the transition width. The sigmoid spans approximately 5–20 Myr, consistent with the typical timescale for molecular cloud dispersal (Chevance et al. 2020). The total optical depth is then $$\tau(\lambda, t_{\rm age}) = w(t_{\rm age}) \cdot \tau_{\rm BC}\cdot k_{\rm BC}(\lambda) + \tau_{\rm ISM}\cdot k_{\rm ISM}(\lambda),

(4)\[\]

where \(k(\lambda)\) is normalized at \(\lambda = 5500\) Å. This requires the full \((n_{\rm age} \times n_\lambda)\) outer product for the exponential, but preserves smooth differentiability through the age boundary, which may matter for gradient-based inference when the data constrain the transition region.

Total depth and diffuse fraction

Under parameterization='tauv_mu' the two screens are set by the total V-band depth \(\hat\tau_V\) and the fraction \(\mu \in [0, 1]\) of that depth in the diffuse ISM, following Charlot and Fall (2000) in the BEAGLE and MAGPHYS convention:

(5)\[ \tau_{\rm BC} = (1 - \mu)\,\hat\tau_V, \qquad \tau_{\rm ISM} = \mu\,\hat\tau_V.\]

This is the same two-screen model as the native \((\tau_{\rm BC}, \tau_{\rm ISM})\) pair, reparametrized, so equivalent values give bit-identical predictions. The convention in BEAGLE and MAGPHYS places a log-uniform prior on \(\hat\tau_V\) and either fixes \(\mu\) at 0.3 to 0.4 or leaves it uniform on \([0, 1]\). The declared free priors are \(\hat\tau_V \sim\) Uniform(0, 4), default 1.0, and \(\mu \sim\) Uniform(0, 1), default 0.3; another prior on \(\hat\tau_V\) is passed explicitly. The energy-balance table of the fast path is built on the hull of the \((\hat\tau_V, \mu)\) box, so its \(\tau_{\rm BC}\) and \(\tau_{\rm ISM}\) axes span the range of Equation (5) over the box, and a free \(\mu\) widens them.

Chevallard et al. (2013) diffuse-ISM law

The chevallard13 law is an optical-depth-dependent curve for the diffuse screen. It is accepted as law_diff only, and the diffuse screen supplies its own depth \(\tau_{\rm ISM}\) to the curve. With \(\tau_V \equiv \tau_{\rm ISM}\), the curve is

(6)\[ n_V(\tau_V) = \frac{2.8}{1 + 3\sqrt{\tau_V}}, \qquad b(\tau_V) = 0.3 - 0.05\,\tau_V\]
(7)\[ n(\lambda, \tau_V) = n_V(\tau_V) + b(\tau_V)\,\left(\frac{\lambda}{\mu\text{m}} - 0.55\right), \qquad k(\lambda, \tau_V) = \left(\frac{\lambda}{5500\,\text{\AA}}\right)^{-n(\lambda, \tau_V)}\]

and the diffuse transmission is \(\exp[-\tau_{\rm ISM}\, k(\lambda, \tau_{\rm ISM})]\), with \(k = 1\) at 5500 Å for every depth. The slope flattens as the depth grows. The fit of Chevallard et al. (2013) covers 0.1 to 2.5 μm. Outside that range the index is held at its 0.1 μm and 2.5 μm values, which is an approximation and not an extrapolation of the fit.

Age-Binned (\(N\)-Screen) Framework

The age-binned mode generalizes the two-component model to \(N\) independent screens, each with its own attenuation curve \(k_i(\lambda)\) and a window \((l_i, h_i)\) in \(\log_{10}(t_{\rm age}/{\rm yr})\), either edge of which may be unbounded. Windows need not partition the age axis. The optical depth seen by an SSP of age \(t_{\rm age}\) is $$\tau(\lambda, t_{\rm age}) = \sum_{i} w_i(t_{\rm age}), \tau_i, k_i(\lambda), \qquad w_i(t_{\rm age}) = \sigma!\left(\frac{\log_{10} t_{\rm age} - l_i}{\Delta_{\rm trans}}\right) \sigma!\left(\frac{h_i - \log_{10} t_{\rm age}}{\Delta_{\rm trans}}\right),

(8)\[\]

with \(\sigma\) the logistic sigmoid of Equation (3) and \(\Delta_{\rm trans}\) the shared transition width (default 0.3 dex); an unbounded window side contributes a factor of 1. The two-component model is the \(N = 2\) case: the screens \(\{(k_{\rm BC}, (-\infty, \log_{10} t_{\rm birth})),\, (k_{\rm ISM}, (-\infty, \infty))\}\) reduce \(w_0(t_{\rm age})\) to the \(w(t_{\rm age})\) of Equation (3) and \(w_1(t_{\rm age}) \equiv 1\), recovering Equation (4) exactly.

Nebular continuum and the discrete emission-line catalog are attenuated at the youngest age, the \(t_{\rm age} \to 0\) limit of Equation (8): screens with a finite lower edge vanish there, leaving $$\tau_{\rm neb}(\lambda) = \sum_{i:; l_i = -\infty} \tau_i, k_i(\lambda).

(9)\[\]

Each screen carries its own optical depth \(\tau_i\) and the shape parameters of its curve; the clumpy geometry of the next section is not applied in this mode.

Because only screens that reach \(t_{\rm age} \to 0\) enter Equation (9), a finite lower edge close to the youngest SSP node would give that node a non-negligible weight from screen \(i\) in Equation (8) while the nebular and line path saw none of it. Construction therefore refuses \(l_i < \log_{10} t_{\rm youngest} + 5\,\Delta_{\rm trans}\), the threshold below which the stellar-path weight exceeds \(\sigma(-5) \approx 0.7\%\); the edge must be unbounded or raised to the threshold.

When three or more screens tile the age axis with shared finite edges, \(\sum_i w_i(t_{\rm age})\) is not exactly 1 at a shared edge: each neighboring sigmoid reaches 0.5 there independently, so the sum runs slightly above 1 (1.0033 at a shared edge and up to 1.0058 at the center of a 1.5-dex screen for \(\Delta_{\rm trans} = 0.3\) dex). The partition is exact only for \(N = 2\), where the single shared edge has a half-infinite screen on each side; far from every edge it returns to 1 to machine precision for any \(N\).

Clumpy Geometry

All three modes support the clumpy-screen geometry of Lower et al. (2022), in which a fraction \(f_{\rm obs}\) of sightlines are unobscured: $$T(\lambda, t_{\rm age}) = f_{\rm obs} + (1 - f_{\rm obs}) \cdot \exp!\bigl[-\tau(\lambda, t_{\rm age})\bigr].

(10)\[\]

Setting \(f_{\rm obs} = 0\) recovers the standard uniform screen geometry.

Per-Source Screen Choice

The two-component framework’s birth-cloud and diffuse-ISM screens attenuate the stars directly; three other keys on the dust_attenuation group pick which of those two screens (or neither) attenuates a non-stellar emission source, using the same clumpy-geometry transmission above:

Key

Default

Governs

nebular_screen

"birth_cloud"

The nebular continuum, the discrete line catalog, and the fast-nebular fallback grid

shock_screen

"diffuse"

The MAPPINGS V shock SED (Shock Emission Backend)

agn_screen

"none"

AGN light

Each accepts "birth_cloud", "diffuse", "none", or the synonym "off". When agn_screen is set to "birth_cloud" or "diffuse", the corresponding screen’s transmission (Equation (10)) is applied to AGN light to model galaxy dust screening of the AGN SED. However, agn_screen cannot be combined with agn_norm="cigale_joint" because both depend on reading the dust budget (L_ir, L_absorbed); use agn_norm="independent" or "conserving" instead. single_component dust has no birth-cloud/diffuse distinction, so it accepts only "none"/"off" or a source’s own default; wg00 and dust-off models refuse these keys outright.

Attenuation Curve Library

Table 1 lists the attenuation curves registered in tengri. All curves return \(k(\lambda)\) normalized such that \(k(5500\,\text{\AA}) = 1\).

Name

Reference

UV Bump

Free Parameters

power_law

Charlot & Fall 2000

No

\(n\) (slope)

calzetti

Calzetti+2000

No

None (\(R_V = 4.05\) fixed)

leitherer02

Leitherer+2002

No

None (UV extension 970–1800 Å)

kriek_conroy

Kriek & Conroy 2013

Yes

\(\delta\), \(E_b\)

noll09

Noll+2009

Yes

\(\delta\), \(E_b\)

salim_sbl18

Salim+2018 (modified Calzetti+L02)

Yes

\(\delta\), \(E_b\)

smc

Pei 1992

No

None

gordon03_smcbar

Gordon et al. 2003

No

None

lmc

Pei 1992

Weak

None

cardelli

Cardelli+1989

Yes

\(R_V\)

li08

Li et al. (2008)

Yes

\(c_1\), \(c_2\), \(c_3\), \(c_4\)

salim

Salim+2018 (DSPS default)

Yes

\(\delta\), \(E_b\)

tea

Haskell et al. (2024)

Yes

\(p\), \(b_{\rm UV}\)

narayanan_z

Narayanan+2018 (\(z\)-dependent)

Yes

\(z\) (redshift)

conroy2010

Conroy+2010 (CCM89, scaled 2175 Å bump)

Yes

\(R_V\), \(B\) (bump strength)

: Attenuation curves available in tengri. Three dust geometry models (wg00_shell, wg00_cloudy, wg00_dusty; Witt and Gordon (2000)) are also available.

In addition to the curves listed above, tengri also registers older-generation grain-composition laws, vw07_bc and vw07_diff (Wild et al. (2007) birth-cloud and diffuse-ISM power-laws), prevot_smc (the original Prevot+1984 SMC parameterization prior to Pei 1992), and the wd01_smcbar, wd01_mwrv31, d03_mwrv31, and hd23_mwrv31 grain-model laws, which provide curves tied directly to specific dust-grain populations. These are exposed through the same registry interface as the curves in Table 1.

Several curves incorporate the 2175 Å UV bump via a Drude profile: $$D(\lambda; \lambda_0, \gamma_D) = \frac{(\lambda, \gamma_D)^2}{(\lambda^2 - \lambda_0^2)^2 + (\lambda, \gamma_D)^2},

(11)\[\]

with default \(\lambda_0 = 0.2175\,\mu{\rm m}\) and \(\gamma_D = 0.035\,\mu{\rm m}\). The Kriek–Conroy family of curves (Kriek and Conroy 2013; Salim et al. 2018) modifies the Calzetti base by adding a power-law slope deviation \(\delta\) and UV bump amplitude \(E_b\): $$k(\lambda) = k_{\rm base}(\lambda) \cdot \left(\frac{\lambda}{5500,\text{\AA}}\right)^{\delta} + E_b\cdot D(\lambda).

(12)\[\]

Several variants differ in operator ordering: Noll09 applies \((\text{base} + \text{bump}) \times \text{slope}\), whereas Kriek–Conroy and SBL18 apply \(\text{base} \times \text{slope} + \text{bump}\); the SBL18 variant (Salim et al. 2018) uses the far-UV Leitherer02 extension. The TEA curve (Haskell et al. 2024) imposes a physically motivated \(E_b(\delta) = 2.5\,\exp(3.5\,\delta)\times 10^s\) correlation calibrated on NIHAO-SKIRT radiative transfer, reducing the free parameters to two.

The Li et al. (2008) analytical curve provides a four-parameter family (\(c_1\)–\(c_4\)) that reproduces MW, LMC, SMC, and Calzetti-like curves within a single functional form (see also Markov et al. 2023, 2025): $$\begin{aligned}

\frac{A_\lambda}{A_V} = {} & \frac{c_1}{(\lambda/0.08)^{c_2} + (0.08/\lambda)^{c_2} + c_3} \nonumber \ & + \frac{233,[1 - c_1/D_0 - c_4/4.60]}{(\lambda/0.046)^2 + (0.046/\lambda)^2 + 90} \nonumber \ & + \frac{c_4}{(\lambda/0.2175)^2 + (0.2175/\lambda)^2 - 1.95}, \end{aligned}

(13)\[\]

where \(\lambda\) is in \(\mu\)m, \(D_0 = 6.88^{c_2} + 0.145^{c_2} + c_3\), and the three terms represent the UV/optical continuum, far-UV rise, and 2175 Å bump respectively. This is the fiducial attenuation model in recent high-redshift JWST analyses (Markov et al. 2025).

The remaining curves in Table 1, power law (Charlot and Fall 2000), Calzetti (Calzetti et al. 2000), Cardelli (Cardelli et al. 1989), SMC/LMC (Pei 1992), and Leitherer02 (Leitherer et al. 2002), follow their published functional forms without modification.

narayanan_z is the exception, because Narayanan et al. (2018) publish no functional form for the redshift dependence. They report median attenuation curves at integer \(z = 0\) to \(6\) from a 25 Mpc MUFASA radiative-transfer run and distribute those curves as a table. tengri fits the Kriek and Conroy (2013) form to each published median over 1250 Å to 1 \(\mu\)m and interpolates the fitted slope and bump linearly in redshift, holding the end value outside \(0 \le z \le 6\) rather than extrapolating. The residual root-mean-square of that fit is 0.010 to 0.024 on curves of order unity, so it reproduces the published medians to a few percent and no better. Redshift is the only free parameter: the law is the median curve at \(z\), and a fit that wants a slope or bump of its own uses kriek_conroy, which is that model. scripts/fit_narayanan2018_medians.py reproduces the fit from the repackaged table in data/attenuation/.

UV crossover. The Leitherer et al. (2002) far-UV extension, used by leitherer02, noll09 and salim_sbl18 below 1800 Å, joins the Calzetti curve at 1800 Å.

Dust Emission Models

Dust emission models span parametric (modified blackbody, Casey (2012)) and tabulated (Dale et al. (2014), Draine and Li (2007), Draine and Li (2014), Hensley and Draine (2023), Jones et al. (2017), da Cunha et al. (2008), Boquien and Salim (2021)) approaches. Tabulated models are loaded lazily from pre-computed template grids at first use; the templates are interpolated in the model parameters via JAX-compatible routines and are fully differentiable. All models are normalized by an energy-balance constraint.

Energy Balance

The absorbed luminosity is computed as $$L_{\rm abs} = \int_0^{\infty} \bigl[1 - T(\lambda)\bigr], L_{\nu,\rm intrinsic}(\lambda), d\nu,

(14)\[\]

where \(T(\lambda)\) is the SFH-weighted effective dust transmission from Equation (10). Each emission model is then normalized so that \(\int L_{\nu,\rm emit}\, d\nu = \eta \, L_{\rm abs}\), where \(\eta\) is the energy-balance relaxation parameter (dust_eta_balance, default \(\eta = 1\)). Setting \(\eta = 1\) enforces strict energy conservation; \(\eta > 1\) accommodates additional infrared luminosity from sources not in the line of sight (e.g., embedded AGN); \(\eta < 1\) accounts for geometric configurations where some absorbed energy is re-emitted out of the observed aperture.

Modified Blackbody

The optically thin modified blackbody (2–3 parameters: \(T_{\rm dust}\), \(\beta_{\rm IR}\), and optionally redshift for CMB correction): $\(L_\nu^{\rm dust} \propto \left(\frac{\nu}{\nu_{\rm ref}}\right)^{\beta_{\rm IR}} B_\nu(T_{\rm eff}),\)\( where \)B_\nu(T)\( is the Planck function, \)\nu_{\rm ref}\( corresponds to \)250,\mu\(m, and \)T_{\rm eff}$ is the CMB-corrected dust temperature.

CMB temperature correction.

At high redshift the CMB sets a temperature floor on dust grains. Following da Cunha et al. (2013): $$T_{\rm eff} = \bigl(T_{\rm dust}^{,n} + T_{\rm CMB}(z)^{,n} - T_{\rm CMB,0}^{,n}\bigr)^{1/n},

(15)\[\]

where \(n \equiv 4 + \beta_{\rm IR}\), \(T_{\rm CMB}(z) = T_{\rm CMB,0}\,(1+z)\), and \(T_{\rm CMB,0} = 2.7255\) K (Fixsen 2009, T_CMB_0, #2806).

CMB luminosity boost.

The dust absorbs the CMB as well as the starlight, so its emitted total grows with the heated temperature. Following da Cunha et al. (2013, Sect. 2.2, Eqs. 11 and 17), the spectrum before the contrast integrates to \(\int L_\nu^{\rm emit}\,d\nu = \eta\,L_{\rm abs}\,[T_{\rm eff}/T_{\rm dust}]^{n}\) at \(z > 0\), and to \(\eta\,L_{\rm abs}\) at \(z = 0\). The excess is CMB power the dust absorbs (#2806).

CMB contrast factor.

The observed dust flux is suppressed relative to the intrinsic emission when observed against the CMB: $$\frac{S_{\rm obs}}{S_{\rm int}} = 1 - \frac{B_\nu(T_{\rm CMB}(z))}{B_\nu(T_{\rm eff})}.

(16)\[\]

Casey (2012) MBB + Mid-IR Power Law

The Casey (2012) model adds a mid-IR power-law component to the modified blackbody, capturing the warm dust continuum that a single MBB underestimates at \(\lambda \lesssim 50\,\mu\)m. Three free parameters: \(T_{\rm dust}\), \(\beta_{\rm IR}\), and \(\alpha_{\rm MIR}\) (mid-IR slope). An internal optically_thin flag suppresses the power-law component when set; in that limit the model reduces to a pure modified blackbody with two free parameters.

Dale et al. (2014) Templates

A 1-parameter family controlled by the power-law slope \(\alpha\) of the radiation field intensity distribution \(dM_{\rm dust}/dU \propto U^{-\alpha}\) (Dale et al. 2014). Low \(\alpha\) (\(\mathord{\sim}1\)–\(1.5\)) produces warm, peaked SEDs characteristic of luminous IR galaxies; high \(\alpha\) (\(\mathord{\sim}3\)–\(4\)) produces cooler SEDs. Templates are loaded from a pre-computed grid of 64 spectra and interpolated in \(\alpha\); CMB corrections (Equation (15)) are applied at \(z > 0\).

Draine & Li (2007)

A 3-parameter model (\(U_{\min}\), \(\gamma_{\rm PDR}\), \(q_{\rm PAH}\)) following Draine and Li (2007). Templates are loaded from a pre-computed grid and interpolated in the three parameters. The radiation field intensity \(U_{\min}\) sets the diffuse ISM dust temperature, \(\gamma_{\rm PDR}\) controls the fraction arising from photo-dissociation regions, and \(q_{\rm PAH}\) modulates mid-IR features.

Draine & Li (2014) Update

The Draine and Li (2014) update adds a fourth parameter controlling the high-\(U\) dust mass fraction independently of \(\gamma_{\rm PDR}\), with revised emissivities from updated laboratory data. Templates are loaded from a separate pre-computed grid.

MAGPHYS (da Cunha et al. 2008), planned

This backend is planned but not yet registered in DUST_EMISSION_MODELS; the description below corresponds to the model that will be exposed once the implementation lands. The four-component model of da Cunha et al. (2008) decomposes the IR SED into PAH features (Drude profiles following Smith et al. (2007)), a hot MIR continuum (\(T \sim 180\) K), warm grains (\(T_W \sim 30\)–\(60\) K), and cold grains (\(T_C \sim 15\)–\(25\) K), each as a modified blackbody with \(\beta = 1.5\) (birth cloud) or \(\beta = 2.0\) (ISM). The energy fractions \(\xi_{\rm PAH}\), \(\xi_{\rm MIR}\), \(\xi_W\) are free parameters; \(\xi_C = 1 - \xi_{\rm PAH} - \xi_{\rm MIR} - \xi_W\).

Astrodust+PAH (Hensley & Draine 2023)

The Hensley and Draine (2023) model replaces the silicate-graphite-PAH composition of DL07 with a single composite grain material (“astrodust”) plus separate PAH nanoparticles, using laboratory-measured optical properties. The interface is identical to DL07 (\(q_{\rm PAH}\), \(U_{\min}\), \(\gamma_{\rm PDR}\)); the revised grain composition shifts \(q_{\rm PAH}\) to 5.91% (vs. DL07’s 4.6%).

THEMIS (Jones et al. 2017)

The Jones et al. (2017) THEMIS framework models amorphous hydrogenated carbon a-C(:H) nanoparticles and amorphous silicate grains with physically motivated size distributions and UV-processing-dependent optical properties. The small-grain fraction \(q_{\rm hac}\) (analogous to \(q_{\rm PAH}\)) and \(U_{\min}\) are the primary free parameters; templates are pre-tabulated via DustEM.

Boquien & Salim (2021) Empirical Templates

The Boquien and Salim (2021) templates add sSFR as a second axis alongside \(L_{\rm TIR}\), capturing the observation that galaxies at fixed IR luminosity but higher sSFR have warmer dust and stronger MIR features. Templates are interpolated from a grid of 2584 star-forming galaxies; \(L_{\rm TIR}\) is set by the energy balance and sSFR is computed from the SFH.

Schreiber et al. (2018) Dust Library

dust_emission={'type': 'schreiber2016'} evaluates the tabulated dust-continuum and PAH template library of Schreiber et al. (2018), as packaged by CIGALE under its schreiber2016 module name. The library holds, for each dust temperature from 15 to 99 K in 1 K steps, the spectrum of one kilogram of dust as a dust continuum plus a PAH component. The two free parameters are \(T_{\rm dust}\) and the PAH mass fraction \(f_{\rm PAH}\) (Schreiber et al. 2018, Sect. 3.2). The two per-kilogram templates are mixed as \((1 - f_{\rm PAH})\,S^{\rm cont}_\nu + f_{\rm PAH}\,S^{\rm PAH}_\nu\) and the mixture is renormalized to \(L_{\rm IR}\). Over the full tabulated range (1 to 3000 \(\mu\)m) the PAH template carries \(R = 3.063\) to \(3.065\) times the power of the continuum template per kilogram at every temperature, so a mass fraction \(f_{\rm PAH}\) carries a power share \(f R/(1 - f + f R)\); for \(f_{\rm PAH} \gtrsim 0.2\) the \(\nu L_\nu\) peak sits on the 7.7 \(\mu\)m PAH complex. Over 8 to 1000 \(\mu\)m alone the ratio is smaller (1.59 at 15 K, falling to 1.33 at 99 K) because part of the PAH power lies below 8 \(\mu\)m. At the three nodes (20 K, 0.05), (35 K, 0.2) and (50 K, 0.5) the six band powers between 3 and 1000 \(\mu\)m agree with CIGALE 2025.1 to better than \(10^{-4}\).

The template file data/schreiber2016_templates.h5 is required: the model raises if it is missing and has no analytic substitute. The defaults are \(T_{\rm dust} = 35\) K and \(f_{\rm PAH} = 0.05\), the declared dust_T and dust_f_pah defaults this type resolves to (other types resolve their own: 30 K for modified_blackbody, 25 K for schreiber2018; CIGALE’s own default is 20 K).

Temperature interpolation. The templates are interpolated linearly between the 1 K nodes: the model is exact at a node, and its derivative with respect to dust_T is discontinuous at every node. For the 8 to 24 \(\mu\)m band power the slope jumps by a median of 42% across the 83 interior nodes, so a gradient-based sampler sees a kink at each integer temperature.

CMB heating and contrast. Opt in with dust_emission={'type': 'schreiber2016', 'cmb': True}; see CMB heating and contrast for the tabulated models. Without the key the redshift is not used. The rest-frame contrast against the CMB is \(0.94\), \(0.83\) and \(0.73\) at rest-frame 250, 500 and 1000 \(\mu\)m for \(T_{\rm dust} = 35\) K at \(z = 4\), and \(0.79\), \(0.64\) and \(0.55\) at \(z = 6\) (0.87, 0.71 and 0.59 at 25 K, \(z = 4\)), so the sub-millimeter flux of a \(z \gtrsim 4\) galaxy is overestimated by these factors when the key is left off.

CMB Heating and Contrast for the Tabulated Models

modified_blackbody, graybody and casey2012 apply the CMB corrections of da Cunha et al. (2013) unconditionally. The tabulated models (schreiber2016, schreiber2018, draine_li2007, draine_li2014, dale2014, dale2014_cigale, themis, astrodust, dh02_ce01, bosa) take them as an opt-in, dust_emission={'type': ..., 'cmb': True}, off by default (the output is then bit-identical to a build without the key). The redshift is the model’s own, so a free redshift is differentiated through and WavePrecomp tabulates the per-filter response over the redshift range.

Contrast, all tabulated models. The emission is multiplied by Eq. 18 of da Cunha et al. (2013), $\(C_\nu = 1 - \frac{B_\nu\bigl(T_{\rm CMB}(z)\bigr)}{B_\nu\bigl(T_{\rm d}(z)\bigr)}, \qquad T_{\rm CMB}(z) = T_{\rm CMB,0}(1+z),\)\( (eq-cmb-contrast-tabulated) with \)T_{\rm CMB,0} = 2.7255$ K (Fixsen 2009), the one CMB temperature every CMB formula uses (T_CMB_0 in tengri.utils.physics_constants, #2806). It is evaluated at rest-frame frequency.

Heating, single-temperature libraries only. schreiber2016 and schreiber2018 have a dust temperature parameter, so the template is read at the heated temperature of Eq. 12, $\(T_{\rm d}(z) = \Bigl[T_{\rm d,0}^{\,4+\beta} + T_{\rm CMB,0}^{\,4+\beta}\bigl((1+z)^{4+\beta} - 1\bigr)\Bigr]^{1/(4+\beta)},\)\( (eq-cmb-heating-tabulated) with \)\beta = 1.8\( (these libraries carry no emissivity index of their own; it is the analytic models' default), and scaled by the luminosity boost the paper gives in Sect. 2.2, \)\(\frac{L_{\rm dust}(z)}{L_{\rm dust}(0)} = \Bigl[\frac{T_{\rm d}(z)}{T_{\rm d,0}}\Bigr]^{4+\beta},\)\( before the contrast. The boost is the CMB energy the dust absorbs; the contrast removes it again from what is detected, so the detected power stays near \)L_{\rm IR}\(. A 35 K galaxy at \)z = 4\( has \)T_{\rm d}(z) = 35.03\( K and a boost of \)1.004\(; at \)z = 6\( the boost is \)1.03$.

Why only contrast for the radiation-field libraries. Draine and Li, Dale et al., THEMIS, Astrodust, Dale and Helou/Chary and Elbaz and BOSA have no dust temperature: a spectrum is a mixture of grains whose temperatures are set by a starlight intensity, plus stochastically heated PAH and small-grain emission that the CMB does not heat at all. There is no single shift of such a template that is the CMB’s effect, so the spectrum is left as tabulated and only the contrast is applied. Its dust temperature is read off the template, as the temperature of the \(\nu^{\beta}B_\nu\) spectrum with the same mean frequency over \(\lambda \ge 50\,\mu\)m (a few K above the temperature of the cold grains, because the warm emission longward of 50 \(\mu\)m is included), and heated by Eq. 12 so that it stays above \(T_{\rm CMB}(z)\). This is an approximation: it is exact for a single-temperature modified blackbody, but for a two-temperature mixture the mean-frequency temperature leans to the hot component while the CMB penalty is set by the cold one, so the detected flux is overestimated by about 5–15 % at \(z = 4\) and 20–100 % at \(z = 6\) when a 20–25 K cold component carries a large share (2–8 % when it is a 5 % hot component). It also does not reproduce the heating of the cold grains. Where the cold grains matter at \(z \gtrsim 6\), prefer schreiber2016 or an analytic model.

Luminosity boost. The Schreiber libraries follow Sect. 2.2 of da Cunha et al. (2013) and include the heated-dust luminosity boost \([T_{\rm d}(z)/T_{\rm d,0}]^{4+\beta}\). The analytic models (modified_blackbody, graybody, casey2012) apply the same boost whenever \(z > 0\), since their CMB heating is always on (#2806; before that they omitted it and under-predicted their emission by up to about 1.8 times for 20 K dust at \(z = 6\)). For casey2012, \(\beta\) is the emissivity index of its graybody term; the boost multiplies the whole normalized spectrum, mid-IR power law included.

Normalization. Every model with CMB heating emits \(L_{\rm IR}\,[T_{\rm d}(z)/T_{\rm d,0}]^{4+\beta}\) before the contrast, where \(L_{\rm IR}\) is the stellar-absorbed (energy-balance) luminosity. The published sed_dust_ir is that spectrum times the contrast, so its frequency integral is not \(L_{\rm IR}\) at \(z > 0\): the contrast removes the undetectable part by construction, and the rest of the boost is the CMB energy the dust absorbs.

References

Boquien, Médéric, and Samir Salim. 2021. “New dust emission templates for star-forming galaxies.” 653: A149. https://doi.org/10.1051/0004-6361/202140734.

Calzetti, Daniela, Lee Armus, Ralph C. Bohlin, Anne L. Kinney, Jan Koornneef, and Thaisa Storchi-Bergmann. 2000. “The Dust Content and Opacity of Actively Star-forming Galaxies.” 533 (2): 682–95. https://doi.org/10.1086/308692.

Cardelli, Jason A., Geoffrey C. Clayton, and John S. Mathis. 1989. “The Relationship between Infrared, Optical, and Ultraviolet Extinction.” 345 (October): 245. https://doi.org/10.1086/167900.

Casey, Caitlin M. 2012. “Far-infrared spectral energy distribution fitting for galaxies near and far.” 425 (4): 3094–103. https://doi.org/10.1111/j.1365-2966.2012.21455.x.

Charlot, Stéphane, and S. Michael Fall. 2000. “A Simple Model for the Absorption of Starlight by Dust in Galaxies.” 539 (2): 718–31. https://doi.org/10.1086/309250.

Chevallard, Jacopo, Stéphane Charlot, Benjamin Wandelt, and Vivienne Wild. 2013. “Insights into the content and spatial distribution of dust from the integrated spectral properties of galaxies.” 432: 2061–91. https://doi.org/10.1093/mnras/stt523.

Chevance, Mélanie, J. M. Diederik Kruijssen, Alexander P. S. Hygate, et al. 2020. “The lifecycle of molecular clouds in nearby star-forming disc galaxies.” 493 (2): 2872–909. https://doi.org/10.1093/mnras/stz3525.

da Cunha, Elisabete, Stéphane Charlot, and David Elbaz. 2008. “A simple model to interpret the ultraviolet, optical and infrared emission from galaxies.” 388 (4): 1595–617. https://doi.org/10.1111/j.1365-2966.2008.13535.x.

da Cunha, Elisabete, Brent Groves, Fabian Walter, et al. 2013. “On the Effect of the Cosmic Microwave Background in High-redshift (Sub-)millimeter Observations.” 766: 13. https://doi.org/10.1088/0004-637X/766/1/13.

Dale, Daniel A., George Helou, Georgios E. Magdis, Lee Armus, Tanio Dı́az-Santos, and Yong Shi. 2014. “A Two-parameter Model for the Infrared/Submillimeter/Radio Spectral Energy Distributions of Galaxies and Active Galactic Nuclei.” 784: 83. https://doi.org/10.1088/0004-637X/784/1/83.

Draine, B. T., and Aigen Li. 2007. “Infrared Emission from Interstellar Dust. IV. The Silicate-Graphite-PAH Model in the Post-Spitzer Era.” 657 (2): 810–37. https://doi.org/10.1086/511055.

Draine, Bruce T., and Aigen Li. 2014. “Interstellar Dust Models with Updated Optical Properties.” 785: 159. https://doi.org/10.1088/0004-637X/785/2/159.

Haskell, P., S. Das, D. J. B. Smith, R. K. Cochrane, C. C. Hayward, and D. Anglés-Alcázar. 2024. “Beware the recent past: a bias in spectral energy distribution modelling due to bursty star formation.” 530 (1): L7–12. https://doi.org/10.1093/mnrasl/slae019.

Hensley, Brandon S., and Bruce T. Draine. 2023. “An Updated Model for Interstellar Grain Alignment and the Astrodust+PAH Grain Model.” 948: 55. https://doi.org/10.3847/1538-4357/acc4c2.

Jones, A. P., M. Köhler, N. Ysard, M. Bocchio, and L. Verstraete. 2017. “The global dust modelling framework THEMIS.” 602: A46. https://doi.org/10.1051/0004-6361/201630225.

Kriek, Mariska, and Charlie Conroy. 2013. “The Dust Attenuation Law in Distant Galaxies: Evidence for Variation with Spectral Type.” 775 (1): L16. https://doi.org/10.1088/2041-8205/775/1/L16.

Leitherer, Claus, I. -Hui Li, Daniela Calzetti, and Timothy M. Heckman. 2002. “Global Far-Ultraviolet (912-1800 Å) Properties of Star-forming Galaxies.” 140 (2): 303–29. https://doi.org/10.1086/342486.

Li, Aigen, S. L. Liang, D. A. Kann, D. M. Wei, S. Klose, and Y. J. Wang. 2008. “On Dust Extinction of Gamma-Ray Burst Host Galaxies.” 685: 1046–54. https://doi.org/10.1086/591049.

Lower, Sidney, Desika Narayanan, Joel Leja, Benjamin D. Johnson, Charlie Conroy, and Romeel Davé. 2022. “How Well Can We Measure Galaxy Dust Attenuation Curves? The Impact of the Assumed Star-dust Geometry Model in Spectral Energy Distribution Fitting.” 931 (1): 14. https://doi.org/10.3847/1538-4357/ac6959.

Markov, V., S. Gallerani, A. Pallottini, et al. 2023. “Dust attenuation at high redshifts: the JWST perspective.” 679: A12. https://doi.org/10.1051/0004-6361/202346983.

Markov, V., S. Gallerani, A. Pallottini, et al. 2025. “Unveiling the trends between dust attenuation and galaxy properties at \(z \sim 2\)–\(12\) with JWST.”

Narayanan, Desika, Romeel Davé, Benjamin D. Johnson, Robert Thompson, Charlie Conroy, and James Geach. 2018. “The IRX-\(\beta\) dust attenuation relation in cosmological galaxy formation simulations.” 474 (2): 1718–36. https://doi.org/10.1093/mnras/stx2860.

Pei, Yichuan C. 1992. “Interstellar Dust from the Milky Way to the Magellanic Clouds.” 395 (August): 130. https://doi.org/10.1086/171637.

Salim, Samir, Médéric Boquien, and Janice C. Lee. 2018. “Dust Attenuation Curves in the Local Universe: Demographics and New Laws for Star-forming Galaxies and High-redshift Analogs.” 859 (1): 11. https://doi.org/10.3847/1538-4357/aabf3c.

Schreiber, Corentin, David Elbaz, Maurilio Pannella, Laure Ciesla, Tao Wang, and Maximilien Franco. 2018. “Dust temperature and mid-to-total infrared color distributions for star-forming galaxies at 0<z<4.” 609: A30. https://doi.org/10.1051/0004-6361/201731506.

Smith, J. D. T., B. T. Draine, D. A. Dale, et al. 2007. “The Mid-Infrared Spectrum of Star-forming Galaxies: Global Properties of Polycyclic Aromatic Hydrocarbon Emission.” 656: 770–91. https://doi.org/10.1086/510549.

Wild, Vivienne, Guinevere Kauffmann, Tim Heckman, et al. 2007. “Bursty stellar populations and obscured active galactic nuclei in galaxy bulges.” 381 (2): 543–72. https://doi.org/10.1111/j.1365-2966.2007.12256.x.

Witt, Adolf N., and Karl D. Gordon. 2000. “Multiple Scattering in Clumpy Media. II. Galactic Environments.” 528 (2): 799–816. https://doi.org/10.1086/308197.