X-CIGALE: Modular SED Fitting

Updated 3 March 2026

X-CIGALE is a modular spectral energy distribution fitting tool that integrates multiple physically motivated modules to model galaxy and AGN emissions across a wide wavelength range.
It combines advanced methods for AGN accretion, torus geometry, polar dust, and host-galaxy components under a global energy balance, enabling robust inference of parameters like SFR and stellar mass.
Utilizing Bayesian analysis and extensive parameter grids, X-CIGALE accurately decomposes AGN and host contributions, enhancing detection efficiency and providing actionable insights for cosmic evolution studies.

X-CIGALE is an advanced spectral energy distribution (SED) fitting framework extending the CIGALE platform to self-consistently model the emission of galaxies and active galactic nuclei (AGNs) from X-ray to radio wavelengths. By integrating physically-motivated modules for AGN accretion, torus geometry, polar dust, and host-galaxy components—alongside the statistical machinery of Bayesian parameter estimation—X-CIGALE enables robust inference of physical parameters such as star formation rates (SFR), stellar mass, AGN luminosity, accretion rate, Eddington ratio, and dust properties for diverse extragalactic populations (Yang et al., 2020, Mountrichas et al., 2020, Yang et al., 2022, López et al., 2024).

1. Architectural Principles and Core Framework

X-CIGALE adopts the modular and energy-balance based philosophy of CIGALE, incorporating distinct physical processes into separate modules, each parameterized through grids sampled in the likelihood calculation. The principal modules comprise:

Star formation history (SFH): e.g. delayed SFR(t)∝t exp(−t/τ), with optional recent bursts.
Stellar populations: Bruzual & Charlot (2003) or similar, Chabrier or Salpeter IMF, solar metallicity.
Dust attenuation: Empirical (Calzetti), two-component (Charlot & Fall), or SMC-like laws.
Dust emission: Dale et al. (2014) templates for reprocessing of UV/optical into IR.
AGN emission: Modern torus models (SKIRTOR; with density ρ∝r⁻ᵖ e^−q|cosθ|), viewing-angle parameterization, and a power-law/empirical disk spectrum.
Polar dust: SMC-type extinction and greybody re-emission at characteristic T∼100 K, β∼1.6.
X-ray emission: Power-law AGN corona (Γ configurable) and (optionally) host X-ray binaries.
Radio: Addition of AGN radio jets/loudness; star-formation radio emission.

Modules interact through global energy balance, with the UV/optical extinction matched to IR emission. For AGN, X-ray, disk, and torus emission are treated consistently, with key physical relations (notably the αₒₓ relation, L_X–L_2500Å or L_X–L_12μm) constraining the normalization across bands (Yang et al., 2020, Yang et al., 2022, López et al., 2024).

2. X-ray and AGN Modules: Physical Methods and Innovations

Early X-CIGALE incarnations imposed the ultraviolet–X-ray correlation via the empirical $\alpha_{ox} = -0.3838\log[L_\nu(2\,\mathrm{keV})/L_\nu(2500\mathrm{\AA})]$ and the Just et al. (2007) relation, discarding SED models whose AGN disk and X-ray fluxes violated $\Delta\alpha_{ox}$ bounds of typically ±0.2 (Yang et al., 2020, Mountrichas et al., 2020). This enables consistent anchoring of the AGN corona and UV disk emission, breaking degeneracies between AGN and host-star contributions.

Subsequent developments extended the module capabilities:

AGN X-ray anisotropy: X-CIGALE now allows $L_X(\theta)/L_X(0^\circ) = a_1\cos\theta + a_2\cos^2\theta + (1-a_1-a_2)$ , capturing different geometries (isotropic, thin-disk, or custom), where θ is the viewing angle (Yang et al., 2022).
X-ray binaries and host-galaxy emission: Models for high-mass XRBs (HMXB), low-mass XRBs (LMXB), and hot gas, with scaling following Fragos et al. (2013b) and Mezcua et al. (2018), enable decomposition of galaxy X-ray emission (Yang et al., 2022, Yang et al., 2020).
Bolometric corrections and LLAGN regime: Low-luminosity AGN require an alternate prior, using the empirical L_X–L_12μm relation (Asmus et al. 2015) via $\alpha_{IRX}$ and explicit ADAF+disk engine mixing (δ_AGN):

$\lambda L_\lambda^\text{total} \propto \delta_{AGN}\,\lambda L_\lambda^{ADAF} + (1-\delta_{AGN})\lambda L_\lambda^{Disk}$

yielding self-consistent bolometric corrections across $L_{Bol} \sim10^{39}$ – $10^{46}$ erg s⁻¹ (López et al., 2024).

3. SED Fitting Workflow and Statistical Framework

Parameter inference in X-CIGALE proceeds via grid sampling. For each parameter set and observation:

Model SEDs are synthesized, convolved with the filter set (including X-ray and IR boxcar bands).
$\chi^2=\sum_i[(F_{obs,i} - F_{model,i})/\sigma_i]^2$ is computed, incorporating upper limits via likelihood integration for censored data (Yang et al., 2020).
Bayesian-posteriors are constructed: $P(\theta) \propto \sum_{models} \exp(-\chi^2/2)\, \delta(\theta-\theta_{model})$ .
Physical parameters are reported as the marginal mean/1σ of the posterior, alongside best-fit single-model values.
Model selection or complexity control leverages the Bayesian Information Criterion (BIC) or Akaike (AIC, ΔAIC).

This architecture delivers robust host-galaxy and AGN decompositions. Systematic uncertainties and degeneracies are evaluated through mock-catalog recovery tests, exploring the impact of parameter grid density and photometric coverage (Yang et al., 2022, Mountrichas et al., 2020).

4. Key Physical Parameters and Diagnostic Indices

X-CIGALE infers a suite of host and AGN properties. Principal parameters include:

AGN fraction ( $\mathrm{frac_{AGN}}$ ): The fraction of total 5–1000μm IR luminosity from the AGN torus (Mountrichas et al., 2020).
$\alpha_{ox}$ index: Measures X-ray/UV slope; crucial for constraining the energy budget in luminous AGN, but requiring modification (to $\alpha_{IRX}$ ) in the LLAGN regime (López et al., 2024).
Bolometric corrections ( $k_X$ ): Empirically calibrated for LLAGN as $k_X \sim 9$ –11, lower than quasars, reflecting the different SED contributions.
SFR, $M_*$ : Derived from joint fitting of SFH, SSP, dust modules; parameter reliability spans typical σ≈0.1–0.3 dex (Padilla et al., 2021, Koutoulidis et al., 2021).
AGN luminosity ( $L_{AGN}$ ), BHAR, $\lambda_{Edd}$ : From AGN disk/torus normalization and X-ray constraints, with Eddington ratios recoverable and cross-checkable against proxies (e.g., $L_X/M_*$ ) (López et al., 2023).

For AGN type classification, X-CIGALE uses combinations of torus viewing angle (e.g., $i_{best}=30^\circ$ for type 1, $70^\circ$ for type 2), polar dust extinction, and $\mathrm{frac_{AGN}}$ thresholds, outperforming simple single-parameter cuts in completeness and reliability (Mountrichas et al., 2021, Padilla et al., 2021).

5. Impact on Observational Studies and Practical Methodology

X-CIGALE underpins multiwavelength SED analyses across large extragalactic surveys:

AGN/host decomposition: Robustly separates AGN from star-formation even in the presence of degeneracy between young dust-enshrouded SF and red AGN tori (Mountrichas et al., 2020, Mountrichas et al., 2021).
Improved AGN detection: X-ray and polar-dust modules increase detection efficiency for Compton-thin and heavily obscured AGN, revealing populations missed by classical mid-IR color criteria (Mountrichas et al., 2020, Mountrichas et al., 2021).
Type-1/type-2 classification: The inclusion of inclination and polar dust recovers $\sim85$ % of spectroscopic type-1 and $>95$ % of type-2 AGN, provided adequate multiwavelength coverage (Mountrichas et al., 2021, Padilla et al., 2021).
Cosmic evolution and co-evolution: Instantaneous SFR and BHAR derived from X-CIGALE enable forward modeling of $M_{BH}$ – $M_*$ relations, supporting self-regulated AGN feedback scenarios (López et al., 2023).
LLAGN census: Empirical L_X–L_12μm-based modules deliver meaningful AGN power estimates in LINERs, Seyferts, and other weakly accreting sources, extending SED modeling below $L_X <10^{42}$ erg s⁻¹ (López et al., 2024).

X-CIGALE efficiency depends critically on SED coverage (UV to FIR/radio), photometric calibration, and precise intrinsic X-ray fluxes (absorption-corrected via N_H or hardness ratios). For statistical robustness, quality cuts (e.g., $\chi^2_{red}<5$ ) and parameter consistency checks are recommended (Koutoulidis et al., 2021).

6. Software Distribution, Evolution, and Usage

All X-CIGALE functionality has been merged into the mainstream CIGALE codebase (v2022.0+), with continued enhancements for AGN anisotropy, radio, and LLAGN-specific modeling (Yang et al., 2022, López et al., 2024). The package provides:

Python-based modular design: Users specify a configuration file enumerating modules and parameter grids.
Integrated statistical outputs: Full posteriors, best-fit SEDs, component fluxes, and diagnostics.
Public codebase and documentation: Downloadable via pip or GitLab; full documentation and parameter file templates online.

Example parameter file segments:

[parameters]
sfhdelayed     # SFH module
population     # SSP
dustatt_calzetti # Dust attenuation
dustem_dale2014 # Dust emission
skirtor2016    # AGN clumpy torus
polar_dust     # Polar dust
xray           # X-ray AGN/host
radio          # AGN jet/radio module

7. Limitations, Systematics, and Future Prospects

Current limitations reflect boundaries in physical modeling and data quality:

For heavily Compton-thick AGN, the SED modeling is limited by the accuracy of absorption corrections and photometric constraints (López et al., 2023, Mountrichas et al., 2020).
The classical $\alpha_{ox}$ prior is not universally applicable; ADAF+disk or $\alpha_{IRX}$ models are required in the low-luminosity regime (López et al., 2024).
Degeneracies persist between torus inclination and polar dust for moderate $E(B-V)$ , only partially mitigated by mock-catalog recovery techniques (Yang et al., 2020).
Optimal performance requires full UV–radio photometric coverage plus robust X-ray measurements, which can be a limiting factor in highly obscured systems at $z>1.5$ (Padilla et al., 2021).

Ongoing code development focuses on enhancing LLAGN sensitivity, refining AGN/star formation decomposition with machine-learning classification of outputs, and integrating direct adaptivity to instrument-specific data footprints (Yang et al., 2022, López et al., 2024).

X-CIGALE has become a community standard for disentangling AGN and host properties in cosmic evolution studies, AGN/galaxy feedback scenarios, and the optimization of legacy and future X-ray/IR/radio surveys (e.g., eROSITA, ATHENA) (Mountrichas et al., 2020, Yang et al., 2022, López et al., 2023, López et al., 2024).