Two-Phase Torus SED Model

Updated 13 February 2026

Two-phase torus SED model is a framework that combines smooth and clumpy dust components to simulate infrared emission in AGNs.
It integrates 3D radiative transfer with realistic dust distributions to capture near-IR excess and mid-IR silicate features.
The model improves AGN parameter estimation by reconciling observed SED diversity with dynamic toroidal dust structures.

A two-phase torus SED model is a physically motivated radiative transfer framework developed to describe the infrared emission of dust surrounding active galactic nuclei (AGN). In these models, the circumnuclear obscurer is not treated as either purely homogeneous or purely made of discrete clouds, but as a mixture of a “smooth” (interclump) component and a population of “clumpy” (cloud) structures. This approach aims to reconcile observed broadband SED features—especially the near-IR excess and the diversity of the mid-IR silicate band—with a geometrically and physically realistic torus structure, accounting for radiative transfer and dust physics in three dimensions [(Siebenmorgen et al., 2015); (Stalevski et al., 2013); (Stalevski et al., 2011); (González-Martín et al., 2023); (Varnava et al., 19 Feb 2025)].

1. Torus Geometry and Dust Distribution

The two-phase torus model prescribes a spatially complex, toroidal geometry surrounding the AGN:

Smooth Disk/Interclump Component: The foundation is a smooth, typically isothermal disk-like density distribution in the equatorial plane. The canonical profile is

$\rho_\text{disk}(r, z) = \rho_0 \left(\frac{r_\text{out}}{r}\right) \exp\left[-\pi \left(\frac{z}{2h}\right)^2\right]$

with $h(r)$ scaling mildly with radius and $\rho_0$ set by the equatorial optical depth.

Clumpy (Cloud) Population: Discrete, high-density clouds are embedded within the disk. Their parameters include total number $N_\text{cl}$ , size $d$ , optical depth per cloud $\tau_{V, cl}$ , and a volume-filling factor $f$ (ranging from less than $2\%$ to over $70\%$ depending on the configuration). Clouds are distributed stochastically in azimuth and radius, with weighting in polar angle to ensure the torus geometry.
Critical Dimensions: The inner radius $r_\text{in}$ is set by dust sublimation ( $T_\text{sub} \sim 800-1800\,$ K), scaling as $r_\text{in} \propto L_\text{AGN}^{1/2}$ . Typical $r_\text{in}$ is $0.1-1\,\text{pc}$ for $L_\text{AGN} = 10^{11} L_\odot$ . The outer-to-inner radius ratio is typically $r_\text{out}/r_\text{in} \sim 100-200$ (Siebenmorgen et al., 2015, González-Martín et al., 2023).
Parameter Space: The key free parameters—the inner radius, the cloud filling factor $\eta$ , cloud optical depth $\tau_{V,cl}$ , disk midplane optical depth $\tau_{V,mid}$ , and viewing angle—establish a five-dimensional grid sampled in precomputed SED libraries (Siebenmorgen et al., 2015).

2. Dust Properties and Radiative Transfer Formalism

Dust Composition: Most models assume a standard Galactic mixture with a power-law size distribution, $n(a) \propto a^{-3.5}$ , over $a = 0.005 - 0.25\,\mu$ m (ISM-type), although "fluffy" aggregates (with 50% vacuum volume, $a = 16-260$ nm) are also used to boost far-IR and submm emissivity compared to interstellar medium (ISM) grains (Siebenmorgen et al., 2015).
Grain Size Effects: Some libraries introduce the maximum grain size $P_\text{size}$ as a parameter, with values up to $10\,\mu$ m to reflect evidence of grain growth and to match weak or broad silicate features in observed SEDs (González-Martín et al., 2023). The mass-weighted mean radius, $\langle P \rangle$ , critically modulates SED shape in the $2-30\,\mu$ m regime.
Radiative Transfer: Three-dimensional Monte Carlo radiative transfer is solved on adaptive grids. Photon packets from an ionizing central SED are tracked as they scatter (anisotropically, using Henyey-Greenstein phase functions), are absorbed, and re-emitted. Self-consistent dust temperatures are calculated by energy balance (often using the Bjorkman & Wood method), and local emission is set by Planck-weighted opacities.
Governing Equation:

$\frac{dI_\nu}{ds} = -\kappa_{\text{ext}}\,\rho\,I_\nu + j_\nu$

with $j_\nu = \kappa_{\text{abs}}\,\rho\,B_\nu(T_\text{dust})$ , and scattering terms handled separately [(Stalevski et al., 2013); (González-Martín et al., 2023)]. Both clumps and the smooth component follow the same microphysics, differing only in density.

Library Construction: SEDs are computed across large parameter grids for statistical SED fitting. For example, the GoMar23 library includes $>690,000$ SEDs varying viewing angle, torus opening angle, radial/polar gradients, clump properties, and grain size (González-Martín et al., 2023).

3. SED Features, Diagnostics, and Parameter Dependence

The emergent SED from two-phase torus models exhibits key observational diagnostics:

Near-IR Excess: Two-phase models naturally produce a $2-6\,\mu$ m continuum in type 1 (unobscured) orientations, attributed to hot interclump dust illuminated close to the sublimation radius. This alleviates the "near-IR deficit" present in purely clumpy torus models, which lack optically thin hot dust [(Siebenmorgen et al., 2015); (Stalevski et al., 2013); (Stalevski et al., 2011)].
Mid-IR Silicate Feature: The $9.7\,\mu$ m silicate feature exhibits emission in type 1 and absorption in type 2 geometries, with strengths governed by the disk and cloud optical depths and dust grain properties. Fluffy grains broaden and shift the peak to longer wavelength ( $11-11.5\,\mu$ m) (Siebenmorgen et al., 2015). The observed range of feature depth (emission $|\tau_{Si}| < 0$ , absorption up to $\tau_{Si} \sim 1.4$ ) is reproduced.
SED Modulation by Parameters:

| Parameter | Effect on SED | Reference | |---------------------|-----------------------------------------------|------------------| | $r_\text{in}$ | Larger $r_\text{in}$ : cooler dust, FIR shift | (Siebenmorgen et al., 2015) | | $\tau_{V,mid}$ | NIR bump (face-on), silicate absorption | (Siebenmorgen et al., 2015) | | $\eta$ , $\tau_{V,cl}$ | Steeper MIR, stronger silicate absorption | (Siebenmorgen et al., 2015) | | $P_\text{size}$ | Larger grains: flatter NIR, weaker silicate | (González-Martín et al., 2023) |

Anisotropy Correction: The directional dependence of IR emission is quantified as $\epsilon(\theta) = F_\text{dust}(\theta)/F_\text{no-dust}(\theta)$ . For face-on views, $\epsilon > 1$ , and for edge-on, $\epsilon < 1$ . Correction is essential when inferring intrinsic AGN luminosity from IR observations (Siebenmorgen et al., 2015).

4. Empirical Validation and Applications

SED Libraries: The principal application is rapid AGN parameter estimation by SED fitting. The SED libraries span viewing angles, filling factors, optical depths, and dust properties to match observed broad- and mid-band SEDs for varied AGN types.
AGN/Starburst Decomposition: Two-phase models often account for the entire $1\,\mu\text{m}$ –submm SED of luminous AGN without invoking a major starburst component; any starburst (when needed) typically constitutes $<20\%$ of the total IR luminosity (Siebenmorgen et al., 2015, Varnava et al., 19 Feb 2025).
Empirical Success: SEDs of type 1/2 Seyferts, quasars, radio galaxies, and hyperluminous IR galaxies are fit with these libraries (Siebenmorgen et al., 2015), including NGC 1365 and NGC 4151 (Swain et al., 2023). Notably, the near-IR bump in Seyfert 1s is resolved without invoking separate hot dust rings (Siebenmorgen et al., 2015). For $\sim 85-90\%$ of Spitzer/IRS AGN, two-phase models with tunable grain size match both continuum slope and silicate feature strength (González-Martín et al., 2023).
Comparison to Other Models: While smooth (tapered disc) geometries can fit some datasets equally well, they predict more pronounced angular anisotropy and typically deeper silicate features, potentially biasing intrinsic AGN luminosity estimates by as much as an order of magnitude relative to two-phase fits (Varnava et al., 19 Feb 2025).

5. Theoretical Foundations and Physical Interpretation

Physical Motivation: Hydrodynamic simulations, turbulence theory, and magnetically elevated disk models all generically yield multiphase, clumpy structures in AGN disks and tori (Hopkins, 2024). Supersonic and turbulent conditions lead to log-normal density probabilities, with a significant fraction of mass in dense clouds embedded in a lower-density medium.
Multi-Phase Implementation: The clumpiness is parameterized with a volume-filling factor and density contrast. In many implementations, $10-30\%$ of the volume is occupied by high-density clumps with a density contrast $C \sim 10^2-10^3$ relative to the interclump medium [(Stalevski et al., 2013); (Stalevski et al., 2011)].
SED Synthesis: Local radiative transfer equations and distributions are integrated across the toroidal volume to calculate emergent SEDs, both for the diffuse and clump phases. Analytic limits reproduce modified blackbody spectra, with the clumpy component flattening the mid-IR and steepening the far-IR slope compared to smooth models (Hopkins, 2024).

6. Observational Diagnostics and Limitations

Isotropy Wavelength: Beyond $\lambda \sim 100-200\,\mu$ m, two-phase tori become optically thin, and emission is almost isotropic; below $\lambda \sim 50\,\mu$ m strong viewing-angle dependence persists (Siebenmorgen et al., 2015).
Silicate Feature Ambiguity: The presence of both emission and absorption in the silicate feature can be mimicked by both smooth and two-phase models; thus, spectroscopically, there is no unique signature distinguishing two-phase versus smooth or clumpy-only scenarios [(Siebenmorgen et al., 2015); (Stalevski et al., 2013)].
Parameter Degeneracy: SED shapes are degenerate with respect to changes in orientation, filling factor, individual cloud properties, and especially the dust grain size distribution. Variations in dust properties (especially large $P_\text{size}$ ) are essential to fitting the full diversity of observed AGN tori (González-Martín et al., 2023).
Unaccounted Far-IR/Submm Deficit: In some systems, the model underpredicts far-IR/submm flux, implying that starburst or extended torus components may contribute at the longest wavelengths (Siebenmorgen et al., 2015).

7. Extensions, Controversies, and Current Research Directions

Magneto-hydrodynamical Integration: Newer work integrates two-phase SED modeling explicitly into magnetically-dominated disk theories, with clump properties tied to the turbulence and large-scale field structure. Analytical similarity solutions can reproduce observed covering factors, cloud columns, and SED scalings over dynamical ranges of mass and accretion rate (Hopkins, 2024).
Ring-and-Torus and Additional Structures: For some AGN, especially those with high-resolution data, extensions to "ring+torus+polar wind" models are employed, requiring additional inner graphite structures and outflow cones (RAT models) to match the UV-to-IR SED (Swain et al., 2023).
Comparative SED Fitting: Bayesian SED fitting codes now routinely compare smooth, clumpy, and two-phase torus templates against composite starburst+AGN+stellar-population models for entire galaxy samples. Choice of torus model affects inferred intrinsic AGN powers by up to an order of magnitude but has less impact on global star formation rates (Varnava et al., 19 Feb 2025).
Grain Growth and Large Grains: There is mounting spectroscopic evidence that micron-sized and even larger dust grains are common in AGN tori, affecting feature strengths and SED slopes, and requiring the explicit addition of dust grain size as a parameter in current and future SED libraries (González-Martín et al., 2023).

A plausible implication is that further advances in spatial resolution, spectral coverage, and physically motivated parametric frameworks will shift consensus away from single-phase and purely geometric AGN torus models in favor of dynamic, multiphase structures attuned to local physical conditions and AGN feedback processes.