Dusty discs in active galactic nuclei

We use a modified version of the code of Efstathiou & Rowan-Robinson, which solves accurately the axially symmetric radiative-transfer problem in dust clouds, to model the infrared emission from dust in active galactic nuclei. The method takes into account a distribution of grain species and sizes, and includes treatment of multiple scattering from grains. We consider three different geometries (flared discs, tapered discs and anisotropic spheres) for the torus that is believed to obscure the central engine and broad-line region, and compare predicted spectra with the observations. We find that very thick tapered discs (discs whose height increases with distance from the central source, but tapers off to a constant height in their outer part) with an opening angle of ~=45 deg and following an r^-1 density distribution are the most successful in satisfying the observational constraints. Tapered discs with different parameters, as well as flared discs and anisotropic spheres, predict spectra that are inconsistent with the presently available data, either because they exhibit strong emission or absorption 10-mu features when viewed face-on, or because they predict too narrow an infrared continuum. Because of the geometry and the density distribution we assumed for the tapered discs, the 10-μm emission features are eliminated with a relatively low equatorial optical depth. The resulting spectra are therefore much broader than those predicted by the very thick and compact cylinders of uniform density suggested by Pier & Krolik. We also see no need to invoke lower silicate efficiencies or silicate depletion by shocks.