The most conspicuous feature of the large-scale magnetic field in the
Andromeda nebula (M31) is a magnetic belt centred at a galacto-centric
radius of about 10 kpc. We suggest a nonlinear dynamo model for M31,
which associates this belt with the gas density profile as well as wit
h the kinematic parameters of M31. Our model is based on the observed
density profile and vertical structure of the gas disc of M31, and on
a recent rotation curve. The traditional explanation of the belt in ki
nematic dynamo theory was connected with a double-peaked form of the r
otation curve of M31. We show that this explanation fails with recent
determinations of the rotation curve of M31. Furthermore, it is unsati
sfactory because it underestimates, even in the kinematic (linear) reg
ime, the interplay of magnetic fields generated at different galacto-c
entric radii. Our model predicts an extended belt of magnetic field be
tween 7 and 12 kpc radius, in accordance with observations. However, o
ur models typically also have another maximum of the regular magnetic
field between 2 and 6 kpc radius which is not obvious in the synchrotr
on emission. We discuss possible reasons why it could avoid detection
if it is real, and suggest a search for the field in this region, e.g.
via Faraday rotation of polarized radio sources behind M31. A further
generic feature of our models is that the scale height of the regular
magnetic field is significantly larger than that of the gas, and has
a steeper increase with galacto-centric radius. This can be important,
e.g., for cosmic ray confinement and vertical gas balance.