Pm. Drysdale et Pa. Robinson, LEVY RANDOM-WALKS IN FINITE SYSTEMS, Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics, 58(5), 1998, pp. 5382-5394
Levy walks on finite intervals with absorbing boundaries are studied u
sing analytic and Monte Carlo techniques. The integral equations for L
evy walks in infinite 1D systems are generalized to treat the evolutio
n of the probability distribution on finite and semi-infinite interval
s. In particular the near-boundary behavior of the probability distrib
ution and also its properties at asymptotically large times are studie
d. The probability distribution is found to be discontinuous near the
boundary for Levy walks in finite and semi-infinite systems. Previous
results for infinite systems, and a previous scaling for semi-infinite
systems, are reproduced. The use of linear operator theory to solve t
he integral equations governing the evolution of the Livy walk implies
that the probability distribution decays exponentially at large times
. For a jump distribution that satisfies psi(x)similar to\x\(-alpha) f
or large \x\, the decay constant for the exponential decay is estimate
d and found to scale at large L as L1-alpha for 2<alpha<3 and L-1 for
1<alpha<2, in contrast to L-2 for normal diffusion. For 2<alpha<3, the
ratio of the decay constants of the first and second eigenfunctions i
s less than 4 for large L, so that the second eigenfunction is relativ
ely more important in describing the system's large time behavior than
the corresponding eigenfunction for normal diffusion. For 1<alpha<2 t
he ratio of the decay constants may be greater or less than 4. The sha
pes of the eigenfunctions for the Levy processes are obtained numerica
lly and the strong similarity between the first eigenfunction and its
normal diffusion counterpart for 2 less than or similar to alpha<3 ind
icate that it would be difficult experimentally to distinguish such a
Levy process on a finite interval from a normal diffusive system by co
nsidering only the asymptotic shape of the probability distribution. F
or alpha less than or similar to 2 we observe significant differences
between the first and second eigenfunctions and their normal diffusion
counterparts. For moderately large intervals, the first eigenfunction
is flatter with large boundary discontinuities while the second eigen
function can differ from its normal diffusion counterpart in both its
symmetry properties and number of nodes. [S1063-651X(98)09210-1].