Multi-wavelength Extensions¶
X-ray and radio emission each have contributions from both the AGN and the host galaxy’s stellar populations. We describe them as a separate section because their galaxy components (X-ray binaries, star-forming synchrotron) are independent of AGN activity.
X-ray Emission¶
AGN Corona¶
The \(\alpha_{\rm ox}\) bridge.¶
For empirical disc backends that do not model a corona explicitly, the X-ray luminosity is anchored to the disc UV output via the Just et al. (2007) relation: $$\alpha_{\rm ox} = -0.137,\log L_{\nu}(2500,\text{\AA}) + 2.638,
where \(L_\nu(2500\,\text{\AA})\) is in \(\mathrm{erg\,s^{-1}\,Hz^{-1}}\). The definition \(\alpha_{\rm ox} \equiv 0.384\,\log(L_{2\,\rm keV}/L_{2500})\) then gives the 2 keV monochromatic luminosity. An additive offset \(\Delta\alpha_{\rm ox}\) (prior \(|\Delta\alpha_{\rm ox}| < 0.2\)) absorbs scatter around the mean relation.
Reproducing CIGALE’s yang20 controls.¶
The xray={'type': 'yang20'} block exposes the three controls of CIGALE’s yang20 module, with the same formulas:
Absolute \(\alpha_{\rm ox}\).
xray={'type': 'yang20', 'alpha_ox': Uniform(-1.9, -1.1)}sets the slope directly, as CIGALE’salpha_oxlist does. The corona forms the offset \(\Delta\alpha_{\rm ox} = \alpha_{\rm ox} - \alpha_{\rm ox}^{\rm Just07}(L_{2500})\) on every draw, so the 2 keV anchor is exactly \(\alpha_{\rm ox}\). Givealpha_oxordelta_alpha_ox, never both: naming both raisesParameterError. Without either key the offset defaults to \(\Delta\alpha_{\rm ox} = 0\).Anisotropy coefficients.
xray_a1(default 0.5, prior Uniform(0, 1)) andxray_a2(default 0.0, prior Uniform(-0.5, 0.5)) are the coefficients of the anisotropy in Eq. (eq-xray-aniso).xray_a1=0, xray_a2=0is isotropic, CIGALE’sangle_coef“0 & 0”.max_dev_alpha_ox. CIGALE rejects a model whose \(\alpha_{\rm ox}\) lies more thanmax_dev_alpha_oxfrom the Just et al. (2007) relation. tengri offers the soft form: a Gaussian prior on the offset with \(\sigma\) equal tomax_dev_alpha_ox,xray={'type': 'yang20', 'delta_alpha_ox': Gaussian(0.0, 0.2)}formax_dev_alpha_ox = 0.2. The log-prior at \(\Delta\alpha_{\rm ox} = \pm 0.2\) is 0.5 below its value at \(\Delta\alpha_{\rm ox} = 0\); a model is penalized, not rejected.
The per-term X-ray luminosities that CIGALE reports are derived keys in \(\log_{10}\) [dex re erg s\(^{-1}\)], each the band integral of the term that enters the X-ray SED. The properties are log_l_x_hmxb_2_10 (HMXB, 2–10 keV), log_l_x_lmxb_2_10 (LMXB, 2–10 keV), log_l_x_hotgas_0p5_2 (hot gas, 0.5–2 keV), log_l_x_agn_2_10 (corona, 2–10 keV), log_l_x_agn_total (unabsorbed corona, 0.25–1240 keV, the band of CIGALE’s yang20 module), and log_lnu_2kev_30deg, the intrinsic 2 keV monochromatic \(L_\nu\) at \(\cos i = 30^\circ\) in \(\log_{10}\) [dex re erg s\(^{-1}\) Hz\(^{-1}\)], the \(\alpha_{\rm ox}\) anchor CIGALE prints. The two XRB terms add to log_l_x_xrb. Terms that are exactly zero report \(-\infty\). Under the lopez24 corona the 2 keV anchor is not defined, so log_lnu_2kev_30deg reports \(-\infty\) there.
X-ray spectral shape.¶
The corona spectrum is a power law with exponential cutoff: $$L_\nu \propto E^{1 - \Gamma},\exp!\left(-\frac{E}{E_{\rm cut}}\right),
with photon index \(\Gamma = 1.8\) and cutoff energy \(E_{\rm cut} = 300\) keV as defaults (free when fitting X-ray data).
X-ray anisotropy.¶
Following Yang et al. (2022), the observed X-ray luminosity depends on viewing angle: $$L_X(\theta) = L_X(0),\bigl[a_1\cos\theta + a_2\cos^2!\theta + (1 - a_1 - a_2)\bigr],
with default coefficients \(a_1 = 0.5\), \(a_2 = 0\). This produces a factor \({\sim}\,2\) suppression for edge-on (Type 2) sightlines relative to face-on, consistent with X-ray surveys.
Physical corona mode.¶
When a physical disc backend (qsosed or RELAGN) is selected, the corona photon index is derived self-consistently from the Beloborodov (1999) Comptonization formula: $$\Gamma_{\rm hot} = \frac{7}{3}\left(\frac{L_{\rm diss}}{L_{\rm seed}}\right)^{!-0.1}!,
where \(L_{\rm diss}\) is the power dissipated in the corona and \(L_{\rm seed}\) is the seed photon luminosity intercepted from the disc. This bypasses the \(\alpha_{\rm ox}\) bridge entirely, since \(L_X\) and \(\Gamma\) emerge from the same accretion physics as the disc SED.
Galaxy X-ray Binaries¶
X-ray binary (XRB) emission from the host galaxy is independent of AGN activity and arises from accretion onto compact remnants in binary systems. Two populations contribute:
High-mass X-ray binaries (HMXBs)¶
scale with star-formation rate because the massive donor stars that power them are short-lived (\(\lesssim 30\,\)Myr).
Low-mass X-ray binaries (LMXBs)¶
scale with stellar mass because their low-mass donors evolve on Gyr timescales.
The total XRB luminosity (Grimm et al. 2003; Gilfanov 2004) is: $$L_X^{\rm XRB} = 2.6 \times 10^{39},\frac{\mathrm{SFR}}{M_\odot,\mathrm{yr}^{-1}} ;+; 8.3 \times 10^{28},\frac{M_\star}{M_\odot} \quad [\mathrm{erg,s^{-1}}],
with logarithmic offset parameters \(\delta_{\rm HMXB}\) and \(\delta_{\rm LMXB}\) (in dex) to absorb population-level scatter (Fragos et al. 2013). Future work will replace these empirical scaling relations with self-consistent XRB population synthesis from X-BPASS (Briel et al. 2025), which models accreting compact-object spectra alongside their stellar progenitors.
Radio Emission¶
The radio module decomposes the total luminosity into three physically distinct contributions: $$L_\nu^{\rm radio} = L_\nu^{\rm synch} + L_\nu^{\rm ff} + L_\nu^{\rm AGN},
where \(L_\nu^{\rm synch}\) is the star-forming synchrotron component (anchored to the FIRRC), \(L_\nu^{\rm ff}\) is the thermal free-free contribution from HII regions, and \(L_\nu^{\rm AGN}\) is the AGN jet component. Each term can be inspected independently via radio_components().
Star-Forming Synchrotron and FIRRC Modes¶
Synchrotron radiation from supernova remnants and cosmic-ray electrons is the dominant radio continuum at \(\lesssim 10\,\)GHz. Its normalization is set by the FIR–radio correlation (FIRRC): $$q_{\rm IR} = \log!\left(\frac{L_{\rm IR}}{3.75 \times 10^{12},{\rm Hz}\cdot L_{\nu}(1.4,{\rm GHz})}\right).
The spectral shape is a power law \(L_\nu \propto \nu^{-\alpha_{\rm SF}}\) (default \(\alpha_{\rm SF} = 0.8\)). Three FIRRC calibrations are available via sfr_mode:
bell2003(Bell 2003): constant \(q_{\rm IR} = 2.64\), the calibration of the total 1.4 GHz luminosity (see the free-free section below).bell2003_split(AGNfitter-rX): the same Bell total split 90% / 10% into a non-thermal (\(\alpha = 0.75\)) and a thermal (\(\alpha = 0.1\)) power law; it takes no separate free-free term. Beside a nebular backend that carries free-free (cue,cloudy_grid) the split keeps its 90 % synchrotron and drops its own 10 % thermal law, so the nebular continuum is the only thermal term; the 1.4 GHz total is then the synchrotron share of Bell’s calibration plus that backend’s free-free, as forbell2003above. Because the nebular free-free is modeled rather than fixed at Bell’s empirical 10 %, that total differs from the calibration by a few per cent (0.96 for a delayed-tau galaxy of 1 Gyr at 5 Gyr, 1.02 for constant star formation).delvecchio2021(Delvecchio et al. 2021, Eq. 5): the multi-parametric total-radio fit \(q_{\rm IR} = q_0(1+z)^{z_s} - m_s(\log M_\star - 10)\) at 1.4 GHz, with defaults \(q_0 = 2.646\), \(m_s = 0.148\), \(z_s = -0.023\) (the paper’s best fit, \(\pm 0.024\), \(\pm 0.013\), \(\pm 0.008\)).mccheyne2022(McCheyne et al. 2022, A&A 662, A100, Sect. 5.2): the joint redshift and mass fit at 150 MHz, \(q_{\rm TIR} = q_0(1+z)^{z_s} + m_s(\log M_\star - 10.45)\) with \(q_0 = 1.98 \pm 0.02\), \(z_s = 0.02 \pm 0.04\), \(m_s = -0.22 \pm 0.03\), valid for \(z < 0.4\) and \(M_\star > 10^{10.45}\). The IR input is the 8–1000 \(\mu\)m luminosity (the paper’s Eq. 4 definition), and the 150 MHz synchrotron uses the paper’s \(\alpha = 0.60\) (Table 3). The constants are from the paper’s accepted-manuscript text layer (Leiden repository, hdl 1887/3561784).
For both mass/redshift modes \(q_{\rm IR}\) calibrates the total luminosity at the mode’s own reference frequency, \(L^{\rm tot}_{\nu_{\rm ref}} = L_{\rm IR}/(3.75\times10^{12}\,{\rm Hz}\cdot10^{q_{\rm IR}})\), exactly as for bell2003: the synchrotron term is the total minus the Murphy et al. (2011) free-free term at \(\nu_{\rm ref}\), and no Bell \(n(L)\) factor multiplies it (the mass and luminosity trend is already in \(q_{\rm IR}\)). The build evaluates the worst corner of the declared box, over \(q_0\), the mass and redshift slopes, the redshift, and the stellar mass the radio block reads, and refuses the box if \(q\) there exceeds \(q_\ast\) at \(\nu_{\rm ref}\) (the same limit as for bell2003). The lowest reachable \(\log M_\star\) is the SFH log_total_mass floor plus \(\log_{10}\) of the smallest SSP mass-remaining fraction (a conservative bound).
Calibrated domain of the mass and redshift relations.¶
Each relation is an empirical linear fit, valid only over the sample it was fitted to. Outside that sample it is not extrapolated: its mass and redshift arguments are held at the nearest edge, so the relation is constant beyond the edge and its gradient with respect to those arguments is zero there.
delvecchio2021: \(0.1 < z < 4.5\) and \(10^8 < M_\star/M_\odot < 10^{12}\) (Delvecchio et al. 2021, Sect. 2).mccheyne2022: \(M_\star > 10^{10.45}\,M_\odot\) and \(z < 0.4\) (McCheyne et al. 2022, Sect. 5.2, the joint fit). No upper mass bound is stated.
If a declared prior on the stellar mass or the redshift reaches outside these ranges, the build issues a warning that names the range and the calibration. It is not an error: the relation is held at the edge there. A default galaxy at \(z = 0\) therefore warns in delvecchio2021, because \(z = 0\) lies below the calibrated range, and its \(q_{\rm IR}\) is the value at \(z = 0.1\).
The build refuses a declared box only when its worst corner makes the synchrotron term negative. For the default free priors of both modes that worst corner stays below \(q_\ast\), so those boxes build.
The three FIRRC parameters (\(q_0\), \(m_s\), \(z_s\)) are exposed as free parameters in all mass/redshift-dependent modes, enabling hierarchical inference over galaxy populations.
Normalization relative to CIGALE.¶
CIGALE’s radio module normalizes the star-forming synchrotron at the 21 cm wavelength (\(\nu = 1.42758\,\)GHz) with a default \(q_{\rm IR} = 2.58\). tengri anchors at \(1.4\,\)GHz, the frequency at which Bell (2003) defines \(q_{\rm IR}\), and its default is \(q_{\rm IR} = 2.64\). At equal \(L_{\rm IR}\) and \(\alpha_{\rm SF}\) the tengri luminosity is the CIGALE one times the frequency-independent factor \(10^{\,q_{\rm CIGALE} - q_{\rm IR}}\,(1.4/1.42758)^{\alpha_{\rm SF}}\), which is 0.857 at the defaults (\(\alpha_{\rm SF} = 0.8\)). Setting radio_q_ir = 2.58 leaves the anchor factor alone: 0.985 at \(\alpha_{\rm SF} = 0.8\), 0.988 at 0.6 and 0.981 at 1.0. The anchor is a model-grammar key: radio={'sf': {'type': 'bell2003', 'nu_ref': '21cm', 'freefree': False}} normalizes at CIGALE’s 1.42758 GHz (a number in Hz is also accepted; '1.4GHz', the default, is Bell’s). With that spelling and radio_q_ir set to CIGALE’s value, tengri and the installed pcigale 2025.1 radio module agree to 2e-5 at 0.15, 1.4, 5 and 30 GHz (tests/regression/radio/test_radio_21cm_anchor_2762.py); without it the ratio is the constant 0.98451 at \(\alpha_{\rm SF} = 0.8\). In \(q_{\rm IR}\) the frequency is a property of the calibration: Bell (2003, Eq. 1) and Yun et al. (2001) measure \(L_\nu\) at 1.4 GHz, Helou et al. (1985) at 1.49 GHz, so the default keeps Bell’s. The key applies to bell2003 only.
Synchrotron suppression at low SFR.¶
At low star-formation rates, cosmic-ray electrons lose energy through inverse Compton scattering off the CMB and infrared radiation field before producing significant synchrotron emission (Klein et al. 1984; Price and Duric 1992; Bell 2003). Bell (2003) parameterizes this as a luminosity-dependent non-thermal fraction (Eq. 3): $$n = \begin{cases} 0.9 & L > L_\star, \ 0.9,(L/L_\star)^{0.3} & L \leq L_\star, \end{cases}
where \(L_\star\) corresponds to \(M_V = -21\) and the total radio luminosity is \(L_\nu = L_\nu^{\rm thermal} + n\,L_\nu^{\rm non\text{-}thermal}\). This correction is significant for dwarf galaxies with \(\mathrm{SFR} \lesssim 0.1\,M_\odot\,{\rm yr}^{-1}\). The bell2003 mode applies it nowhere (its \(q_{\rm IR}\) is a total calibration). The delvecchio2021 and mccheyne2022 modes do not apply it either: their \(q_{\rm IR}(M_\star, z)\) already carries the mass and luminosity dependence that \(n(L)\) would correct for. The apply_suppression argument of the functions is an opt-in kept for callers that want the legacy multiplier.
Thermal Free-Free Emission¶
Thermal bremsstrahlung (free-free) from HII regions contributes at the \(\sim\)5–15% level at GHz frequencies and dominates above \(\sim\)30 GHz, tracing the instantaneous SFR on \(\lesssim 10\,\)Myr timescales. Following Murphy et al. (2011) (Eq. 11): $$\begin{aligned}
L_\nu^{\rm ff} = {} & \frac{1}{4.6\times10^{-28}} \left(\frac{T_e}{10^4,{\rm K}}\right)^{!0.45} \nonumber \ & \times \left(\frac{\nu}{{\rm GHz}}\right)^{!\alpha_{\rm ff}} ,{\rm SFR}, \qquad {\rm SFR} = 3.88\times10^{-44},L_{\rm IR}\ [{\rm erg,s^{-1}}], \end{aligned}
where the SFR [\(M_\odot\,{\rm yr}^{-1}\)] is derived from \(L_{\rm IR}\) via Eq. 4 of Murphy et al. (2011) and \(L_\nu^{\rm ff}\) is in erg s\(^{-1}\) Hz\(^{-1}\). At \(1.4\,\)GHz and \(T_e = 10^4\,\)K this gives \(2.10 \times 10^{27}\) erg s\(^{-1}\) Hz\(^{-1}\) (5.5\(\times10^{-7}\,L_\odot\,{\rm Hz}^{-1}\)) per \({\rm M_\odot\,yr^{-1}}\).
The free-free spectral index \(\alpha_{\rm ff} \approx -0.1\) is nearly flat (default), in contrast to the steep synchrotron slope \(\alpha_{\rm SF} \approx -0.8\). Both \(T_e\) (default \(10^4\,\)K) and \(\alpha_{\rm ff}\) are exposed as free parameters for hierarchical inference.
Controlling free-free inclusion¶
The free-free component’s default depends on where the model already carries thermal emission. It is off when the nebular backend carries a free-free continuum (cue or cloudy_grid): those backends extend their continuum beyond 1 cm as optically thin free-free with slope \(\nu^{-0.1}\), so a second term via Murphy et al. (2011) would double-count it. It is on from the SSP grid edge upward only when the SSP includes nebular emission (a wNE grid, whatever neb type is declared on it): the SSP flux already carries the thermal continuum up to its last wavelength (1 cm for ssp_prsc_miles_chabrier_wNE), so the radio term starts there and the model has one thermal term at every wavelength. It is on over its whole range (from 1 mm) otherwise, for example a bare-stellar SSP with none or cb19 (which publishes no continuum). The first rule is keyed on the declared nebular backend name (nebular_backend_carries_freefree in tengri.components.nebular._models), the second on the SSP’s nebular stamp (ssp_data.nebular == "included", set by the nebular_included file attribute or the wNE filename). An SSP file with neither keeps the term over its whole range; stamp it with tools/stamp_ssp_nebular_attrs.py --included if it includes nebular emission.
To override the default, pass freefree: True or freefree: False explicitly in the radio.sf sub-dict:
radio={
"sf": {"type": "bell2003", "freefree": False},
"agn": {"type": "powerlaw"},
}
Omitting the freefree key applies the auto rule above. Exception: bell2003_split forces freefree: False (that mode already holds a thermal share of its own total), and passing freefree: True with it raises an error.
What radio_q_ir calibrates¶
radio_q_ir is the calibration of Bell (2003): the ratio of the 8–1000 \(\mu\)m luminosity to the total 1.4 GHz radio luminosity, about 10% of which is thermal. The \(L_{\rm IR}\) used is the total absorbed dust power \(L_{\rm absorbed}\,\eta\); the 8–1000 \(\mu\)m band holds 0.944 of it for the default dust emission model (DL14) and 0.958 for Casey (2012), for the delayed-\(\tau\) galaxy used in the checks (integral of the dust-emission SED over the model grid). The three spellings:
freefreeomitted (default). \(q_{\rm IR}\) is the total. The synchrotron term is \((1 - f_{\rm th})\) of \(L_{\rm IR}/(3.75\times10^{12}\,{\rm Hz}\times10^{q_{\rm IR}})\), with \(f_{\rm th}\) the free-free share at 1.4 GHz from the Murphy et al. (2011) term (0.1335 at \(q_{\rm IR} = 2.64\), \(T_e = 10^4\,\)K), and the thermal term is added by this block, or by the nebular continuum when the nebular backend carries free-free. Without a nebular backend the 1.4 GHz total equals the calibration. With one it equals the calibration to within the difference between that backend’s free-free share and \(f_{\rm th}\): with Cue the total is 1.7% below the calibration for a constant star formation history and 7.3% below for a delayed-\(\tau = 1\) Gyr galaxy at 5 Gyr, and the difference follows the ionizing photon rate per unit \(L_{\rm IR}\).freefree: False. \(q_{\rm IR}\) calibrates the non-thermal term alone, as in CIGALE: the synchrotron term carries the whole \(L_{\rm IR}/(3.75\times10^{12}\,{\rm Hz}\times10^{q_{\rm IR}})\), this block adds no thermal term, and a nebular free-free continuum adds to it.freefree: True. As the default, with the thermal term from this block even beside a nebular backend that carries free-free (two thermal terms).
For the default spelling the synchrotron term would turn negative for \(q_{\rm IR}\) above \(q_\ast = -\log_{10}[3.75\times10^{12}\,(3.88\times10^{-44}/4.6\times10^{-28})(T_e/10^4\,{\rm K})^{0.45}\,1.4^{\alpha_{\rm ff}}]\) (3.5145 at \(10^4\,\)K, 3.379 at \(2\times10^4\,\)K). SEDModel.build raises if the support of radio_q_ir (a fixed value or the upper end of its prior) reaches that limit for the largest \(T_e\) and \(\alpha_{\rm ff}\) the model allows; the declared free prior of radio_q_ir ends at 3.37.
Passing freefree: True explicitly keeps the term over its whole range on a nebular-included SSP (it then overlaps the SSP’s own continuum between 1 mm and the SSP edge); freefree: False removes it, leaving no thermal emission beyond the SSP edge.
Which infrared window L_IR is integrated over: ir_window¶
Without the key, each model uses the window its own paper defines \(q\) on: bell2003 and bell2003_split (Bell 2003, total infrared 8–1000 \(\mu\)m), delvecchio2021 (Delvecchio et al. 2021, \(q_{\rm IR}\) on the total rest 8–1000 \(\mu\)m luminosity) and mccheyne2022 (McCheyne et al. 2022, \(q_{\rm TIR}\) with \(L_{\rm TIR}\) the 8–1000 \(\mu\)m MAGPHYS luminosity, their Sect. 5 and Eq. 4) all default to 'tir'. An explicit ir_window always wins, so 'total' gives CIGALE’s convention.
The q relations are written for different infrared luminosities. Helou et al. (1985) define \(q\) on the far-infrared flux between 42.5 and 122.5 \(\mu\)m; Bell (2003) and the recent calibrations (for example Delvecchio et al. 2021, whose abstract defines \(q_{\rm IR}\) on the total rest 8–1000 \(\mu\)m luminosity) use the total infrared, 8–1000 \(\mu\)m. The radio.sf key ir_window selects the rest-wavelength window of the dust-emission SED that is integrated to form \(L_{\rm IR}\), for every q-based star-formation model (bell2003, bell2003_split, delvecchio2021, mccheyne2022) and for the Murphy et al. (2011) free-free term, which is scaled by the same \(L_{\rm IR}\):
'total'. \(L_{\rm IR}\) is the dust power as the dust components publish it (L_ir), CIGALE’s convention. No band integral is taken. It is never the default of a q-based model; name it to reproduce CIGALE.'tir'(the default of every q-based model). \(L_{\rm IR} = \int_{8\,\mu{\rm m}}^{1000\,\mu{\rm m}} L_\nu\,d\nu\) of the re-emitted dust SED (sed_dust_ir).'fir'. The same integral over 42.5–122.5 \(\mu\)m. Helou et al. (1985) define their FIR from the IRAS 60 and 100 \(\mu\)m fluxes, so 42.5–122.5 \(\mu\)m is the nominal IRAS band; the integral over it is what is evaluated here.
The window edges need not fall on grid nodes: the SED is taken linear in frequency between nodes and each grid cell is clipped to the window, so the band is the requested one to the accuracy of the grid trapezoid. For the default delayed-\(\tau\) galaxy the 8–1000 \(\mu\)m band holds 0.944 (DL14) and 0.958 (Casey 2012) of the dust power, so 'tir' lowers the radio flux by 0.025 and 0.019 dex relative to 'total'; this is the shift in the default output.
radio={"sf": {"type": "bell2003", "ir_window": "tir"}, "agn": {"type": "powerlaw"}}
A window named explicitly needs a dust_emission block (it reads sed_dust_ir) and raises a ConfigError without one; a model with no dust-emission SED and no ir_window key uses the published dust power. A rest wavelength grid that does not span the window (for 'tir', 8–1000 \(\mu\)m) also raises a ConfigError. Under diffuse_screen the window is applied to the emergent (screened) dust SED. The derived property q_ir always reports the ratio of the total dust power L_ir to the 1.4 GHz luminosity, whatever window calibrated the model, so with 'tir' it differs from the input radio_q_ir by the band fraction.
Defaults against the literature: the four models default to the window of their own paper (all 8–1000 \(\mu\)m), none is calibrated on the 42.5–122.5 \(\mu\)m window of Helou et al. (1985), and CIGALE’s total-dust-power convention needs ir_window: 'total'.
AGN Radio Jets¶
For radio-quiet AGN, a simple power law \(L_\nu \propto \nu^{-\alpha_{\rm AGN}}\) (\(\alpha_{\rm AGN} = 0.7\)) is sufficient. It is normalized through the radio loudness radio_loudness \(= \log_{10}[L_\nu(5\,{\rm GHz})/L_\nu(4400\,\text{\AA})]\), the B-band ratio. \(L_\nu(4400\,\text{\AA})\) is the intrinsic accretion-disc luminosity that a composable AGN model publishes at its 30° reference inclination; the other AGN models publish no disc luminosity, and the radio block then uses \(L_{\rm bol}/(5.15\,\nu_B)\), a bolometric correction of 5.15 at 4400 Å. The power law is multiplied by \(\exp(-\nu/\nu_{\rm cut})\) with radio_log_nu_cut \(= \log_{10}(\nu_{\rm cut}/{\rm Hz}) = 13\) by default, the synchrotron-aging cutoff of AGNfitter-rX; radio_log_nu_cut = 40 removes it. At the default the cutoff lowers the jet by the factors 0.997, 0.990 and 0.970 at 30, 100 and 300 GHz, and by 0.9995 at the 5 GHz anchor of the loudness.
CIGALE defines the loudness on the 2500 Å anchor, linearly, as \(R = L_\nu(5\,{\rm GHz})/L_\nu(2500\,\text{\AA})\) with the intrinsic disc luminosity at a 30° line of sight, and applies no cutoff. The anchor is a model-grammar key on the AGN radio block: radio={'agn': {'type': 'powerlaw', 'loudness_ref': '2500_30deg'}} takes the loudness on the composable disc’s 2500 Å luminosity at 30°, and radio_loudness stays \(\log_{10}R\) for either anchor, so CIGALE’s \(R\) is radio_loudness \(= \log_{10}R\) and its linear value is \(10^{\rm radio\_loudness}\). The default '4400' keeps the B-band anchor. The 2500 Å option needs a composable AGN, because the monolithic models publish no disc luminosity; the build raises otherwise rather than fall back to the bolometric correction. With radio_loudness = 1 (CIGALE’s \(R = 10\)) the 5 GHz luminosity is \(8.35\times10^{29}\,{\rm erg\,s^{-1}\,Hz^{-1}}\) for the Schartmann (2005) disc at \(\log_{10}L_{\rm agn,bol} = 11\) and \(30°\), CIGALE’s value to 2 per mille; the 4400 Å anchor gives \(1.11\times10^{30}\) for the same disc. The jet’s \(\exp(-\nu/10^{13}\,{\rm Hz})\) cutoff still applies in the 2500 Å case, and removes 0.05 per cent at 5 GHz; radio_log_nu_cut = 40 takes it away.
For radio-loud AGN with spectral curvature, the double power-law (DPL) model from AGNfitter-rx (Martı́nez-Ramı́rez et al. 2024) captures the optically thick/thin transition: $$L_\nu = L_{5{\rm GHz}} \left(\frac{\nu}{\nu_t}\right)^{!\alpha_1}! \left[1 - \exp!\left(-!\left(\frac{\nu_t}{\nu}\right)^{!\alpha_1-\alpha_2}\right)\right] \exp!\left(-\frac{\nu}{\nu_{\rm cut}}\right),
where \(\alpha_1\) is the steep (optically thin) slope (default \(-0.75\)), \(\alpha_2\) is the flat (optically thick) slope (default \(-0.1\)), \(\nu_t = 10^{\log\nu_t}\) is the transition frequency, and \(\nu_{\rm cut} = 10^{13}\,\)Hz is the synchrotron aging cutoff. The cutoff frequency is radio_log_nu_cut, shared with the power-law model.
References¶
Bell, E. F. 2003. “Estimating Star Formation Rates from Infrared and Radio Luminosities: The Origin of the Radio-Infrared Correlation.” 586 (April): 794–813. https://doi.org/10.1086/367829.
Beloborodov, Andrei M. 1999. “On the Number of Active Galactic Nuclei at High Accretion Rates.” 510: L123–26. https://doi.org/10.1086/311810.
Briel, M. M., J. J. Eldridge, E. R. Stanway, and H. F. Stevance. 2025. “X-BPASS: self-consistent X-ray binary evolution in stellar population synthesis.” arXiv e-Prints. https://arxiv.org/abs/2508.18628.
Delvecchio, I., E. Daddi, M. T. Sargent, et al. 2021. “The infrared-radio correlation of star-forming galaxies is strongly \(M_{\star}\)-dependent but nearly redshift-invariant since \(z \sim 4\).” 647 (March): A123. https://doi.org/10.1051/0004-6361/202039647.
Fragos, T., B. D. Lehmer, S. Naoz, A. Zezas, and A. Basu-Zych. 2013. “Energy Feedback from X-Ray Binaries in the Early Universe.” 776: L31. https://doi.org/10.1088/2041-8205/776/2/L31.
Gilfanov, M. 2004. “Low-mass X-ray binaries as a stellar mass indicator for the host galaxy.” 349: 146–68. https://doi.org/10.1111/j.1365-2966.2004.07473.x.
Grimm, H.-J., M. Gilfanov, and R. Sunyaev. 2003. “High-mass X-ray binaries as a star formation rate indicator in distant galaxies.” 339: 793–809. https://doi.org/10.1046/j.1365-8711.2003.06224.x.
Just, Darren W., W. N. Brandt, Ohad Shemmer, et al. 2007. “The X-Ray Properties of the Most Luminous Quasars from the Sloan Digital Sky Survey.” 665: 1004–22. https://doi.org/10.1086/519990.
Kennicutt, Jr., Robert C. 1998. “Star Formation in Galaxies Along the Hubble Sequence.” 36 (January): 189–232. https://doi.org/10.1146/annurev.astro.36.1.189.
Klein, U., R. Wielebinski, and R. Beck. 1984. “A radio continuum study of the Magellanic Clouds. III. The radio continuum spectra of the bright H II regions in the Large Magellanic Cloud.” 133: 19–26.
Martı́nez-Ramı́rez, Gabriela C., Gabriela Calistro Rivera, Elisabeta Lusso, and Francesco Shankar. 2024. “AGNfitter-rx: modelling AGN and galaxy SEDs from radio to X-rays.” 535: 2961–85. https://doi.org/10.1093/mnras/stae2437.
McCheyne, I., S. Oliver, M. Sargent, et al. 2022. “The LOFAR Two-metre Sky Survey Deep fields. The mass dependence of the far-infrared radio correlation at 150 MHz using deblended Herschel fluxes.” 662 (June): A100. https://doi.org/10.1051/0004-6361/202141307.
Murphy, E. J., J. J. Condon, E. Schinnerer, et al. 2011. “Calibrating Extinction-free Star Formation Rate Diagnostics with 33 GHz Free-free Emission in NGC 6946.” 737 (2): 67. https://doi.org/10.1088/0004-637X/737/2/67.
Price, Richard, and Nebojsa Duric. 1992. “New Results on the Radio–Far-Infrared Relation for Galaxies.” 401: 81. https://doi.org/10.1086/172041.
Yang, Guang, Médéric Boquien, W. N. Brandt, et al. 2022. “Fitting AGN/galaxy X-ray-to-radio SEDs with CIGALE and improvement of the code.” 927: 192. https://doi.org/10.3847/1538-4357/ac4971.