Abstract:
Dynamical systems which switch between several di erent branches or modes of evolution
via a Markov process are simple mathematical models for irreversible systems. The
averaged evolution for such a dynamics can be obtained by a compression of the corresponding
reversible dynamics onto a coinvariant subspace in the sense of the Lax/Phillips
Scattering Scheme.
If the dynamics switches between some stable modes of evolution and some unstable
modes, still the averaged or expectation evolution might be stable. From the theory
of random evolutions the generator A of the averaged evolution is obtained, and a
de nition of stability in average is suggested. With regard to this context the generator
A is investigated and conditions for stability in average are given for certain special
situations. Based on these results, a conjecture is made about su cient and necessary
conditions for stability in average for some more general cases.
In order to nd hints how to verify or how to modify the conjecture three qualitatively
di erent solvable models are studied. Here the spectral properties of the generators of
all three models are studied, and the results are put in relation to the conjecture.