We analyze a simple model of deterministic diffusion. The model consists of
a one-dimensional array of scatterers with moving point particles. The par
ticles move from one scatterer to the next according to a piecewise linear,
expanding, deterministic map on unit intervals. The microscopic chaotic sc
attering process of the map can be changed by a control parameter. The macr
oscopic diffusion coefficient for the moving particles is well defined and
depends upon the control parameter. We calculate the diffusion coefficent a
nd the largest eigenmodes of the system by using Markov partitions and by s
olving the eigenvalue problems of respective topological transition matrice
s. For different boundary conditions we find that the largest eigenmodes of
the map match the ones of the simple phenomenological diffusion equation.
Our main result is that the diffusion coefficient exhibits a fractal struct
ure as a function of the control parameter. We provide qualitative and quan
titative arguments to explain features of this fractal structure. [S1063-65
1X(99)15105-5].