Note
Go to the end to download the full example code.
Custom AGN torus model via SEDModelComponent and direct integration¶
A toy single-temperature blackbody torus implemented as a modern
SEDModelComponent subclass, discoverable through SEDModel.build
and composable with other AGN blocks. The SEDModelComponent pattern is
the recommended path for any new SED physics — AGN, dust, or stellar.
The toy curve is a graybody (smooth, no features); the SKIRTOR curve carries the silicate 9.7 μm feature and inclination-dependent geometry the toy model elides. Both are normalized to the same bolometric luminosity and plotted in νL_ν space.
References¶

import os
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2" # suppress XLA/PjRt C++ INFO+WARNING logs
import warnings
import jax
import jax.numpy as jnp
import matplotlib.pyplot as plt
import numpy as np
import tengri
from tengri.agn import register_agn_block
from tengri.plot import setup_style
setup_style()
warnings.filterwarnings("ignore", message=".*BakedInBackend.*")
C_AA_PER_S = 2.998e18
LSUN_ERG = 3.828e33
@register_agn_block(
"torus",
"demo_graybody",
citation="gallery demo (replace with your reference)",
status="experimental",
short_doc="Single-temperature graybody torus — demonstration only",
)
def demo_graybody_torus(
wavelength: jnp.ndarray,
agn_log_lbol: float,
agn_T_torus: float = 300.0,
agn_torus_frac: float = 0.5,
**_kwargs,
) -> jnp.ndarray:
"""Graybody torus emission at a single dust temperature.
Parameters
----------
wavelength : array_like, shape (n_wave,)
Rest-frame wavelength [Angstrom].
agn_log_lbol : float
log10(L_bol / L_sun).
agn_T_torus : float
Dust temperature [K]. Default 300 K.
agn_torus_frac : float
Torus covering / luminosity fraction of L_bol. Default 0.5.
Returns
-------
ndarray, shape (n_wave,)
Torus flux L_nu [erg/s/Hz].
"""
h = 6.626e-27
k_B = 1.381e-16
c_cgs = 2.998e10
nu = c_cgs / (wavelength * 1.0e-8)
B_nu = (2 * h * nu**3 / c_cgs**2) / (jnp.exp(h * nu / (k_B * agn_T_torus)) - 1.0)
L_bol_erg = 10.0**agn_log_lbol * LSUN_ERG
nu_ref = c_cgs / 1.0e-4
B_ref = (2 * h * nu_ref**3 / c_cgs**2) / (jnp.exp(h * nu_ref / (k_B * agn_T_torus)) - 1.0)
L_nu = (B_nu / B_ref) * (L_bol_erg / 1.0e10) * agn_torus_frac
return L_nu
# Negligible host SFH: total mass ~1e-10 Msun, completely subdominant
# to the AGN luminosity below. ``log_sfr`` was the legacy kwarg; current
# ``const`` SFH parametrizes by total mass over [start_gyr, end_gyr].
SFH = {"type": "const", "all_params": tengri.Fixed(tengri.DEFAULT), "log_total_mass": -10.0}
DUST = {
"law": "power_law",
"type": "two_component",
"all_params": tengri.Fixed(tengri.DEFAULT),
"tau_diff": 0.0,
"tau_bc": 0.0,
}
LOG_LBOL = 12.0
ssp = tengri.load_ssp()
model_skirtor = tengri.SEDModel.build(
ssp,
sfh=SFH,
dust_attenuation=DUST,
agn={
"all_params": tengri.Fixed(tengri.DEFAULT),
"log_lbol": LOG_LBOL,
"lum_ratio": 1.0,
"disc": {"type": "multicolor", "all_params": tengri.Fixed(tengri.DEFAULT)},
"torus": {"type": "skirtor", "all_params": tengri.Fixed(tengri.DEFAULT)},
},
redshift=tengri.Fixed(0.0),
)
p_skirtor = dict(model_skirtor.spec.sample(jax.random.PRNGKey(0)))
out_skirtor = model_skirtor.predict(p_skirtor)
wave_um = np.asarray(model_skirtor.wavelengths) * 1.0e-4
nu_lnu_skirtor = (
C_AA_PER_S / np.asarray(model_skirtor.wavelengths) * np.asarray(out_skirtor.rest_sed())
)
L_nu_toy = np.asarray(
demo_graybody_torus(
jnp.asarray(model_skirtor.wavelengths),
agn_log_lbol=LOG_LBOL,
agn_lum_ratio=1.0,
agn_T_torus=300.0,
agn_torus_frac=0.5,
)
)
nu_lnu_toy = C_AA_PER_S / np.asarray(model_skirtor.wavelengths) * L_nu_toy
fig, ax = plt.subplots(figsize=(7.5, 4.6))
ax.loglog(wave_um, nu_lnu_skirtor, color="C0", lw=1.6, label="SKIRTOR (production)")
ax.loglog(wave_um, nu_lnu_toy, color="C3", lw=1.6, label="demo graybody (T=300 K)")
ax.set(
xlim=(0.1, 1.0e3),
ylim=(1.0e41, 1.0e47),
xlabel=r"Rest-frame wavelength [$\mu$m]",
ylabel=r"$\nu L_\nu$ [erg s$^{-1}$]",
)
ax.legend(frameon=False, fontsize=9, loc="lower center")
fig.tight_layout()
plt.savefig("plot_custom_torus_extension.png", dpi=150, bbox_inches="tight")