Between *r* ~ 1 kpc and the black hole's accretion disk proper
(*r* 10^{3}
*GM*_{h} / *c*^{2} ~ 0.03 *M*_{8}
pc for a black hole of mass *M*_{h} = 10^{8}
*M*_{8} *M*_{}) lies the
transition disk. Unlike the accretion disk, this disk will be heated
primarily by external radiation rather than by the local release of
gravitational binding energy. Beyond *r* ~ 20 _{2}^{-2}
*M*_{8} pc the stars in the
galaxy, of velocity dispersion 10^{2} _{2} km s^{-1}
determine the potential. A
photoionized or molecular disk will be self-gravitating and Jeans
unstable if its column density exceeds

where *h* is the scale height of the disk and *T* =
10^{3} *T*_{3} K the gas
temperature. If this gas flows in on a timescale *t*_{in}
to supply the black hole with an accretion rate
= *M*_{}
yr^{-1}, then ~ 1
*r*_{pc}^{-1} *v*_{300}^{-1}
(*t*_{in} / *t*_{orb})g cm^{-2} where
the orbital speed of the gas is 300 *v*_{300} km
s^{-1} and
*t*_{orb} is the orbital time of gas at radius
*r*. Purely local viscosity
results in enormous inflow times, and Begelman, Frank and Shlosman
(papers in these proceedings, and references therein) have argued that
the angular momentum transport is determined by self-gravitation and
global bar instabilities, so that the mean column density always
exceeds _{sg}. Dust in
such a column has an enormous optical depth to UV
photons: 10^{4}
*T*_{3}^{1/2} _{2} *r*_{pc}^{-1}. Provided
the disk is warped (by any of the
mechanisms described in section 2) or
flared, infrared reradiation is inevitable. A disk warped through angle
will intercept and reradiate
a fraction ~ / 3 of the
luminosity of the central source. For
idealized dust of constant , if
*C* = *d*(covering factor) / *dln r*
*r*^{q},
then in the `inertial range' (*kT*_{min} << *h*< << *kT*_{max}) the superposition of
dust emission from all radii produces a reradiated spectrum with
*L*_{}
^{-s}, where

Since substantial warps are to be expected on all scales, we expect in
some *average* sense *q* 0 and hence *s* ~ 1. A diversity of bumps and
wiggles is to be expected in individual objects, depending on the
radii (and hence, via figure 1,
wavelength) where their warps are most
pronounced, the size and composition of their dust grains, and whether
they are viewed from an angle where dust at a large radius absorbs
re-emission from the interior. This seems in accord with the infrared
spectral energy distributions of AGN: a great variety, with a median
*s* ~ 1
(Sanders et al. 1989,
1988b).
As discussed in section 2, the
properties of more realistic galactic dust modify slightly the simple
result of equation (4), and by flattening the *T(r)* relation introduces
a propensity for 3-5 *µ*m ``bumps'' (see
figure 2).