High dimensional chaotic systems which behave like random walks in state space

Abstract: By analysing an n-dimensional generalisation of Thomas's cyclically symmetric attractor we find that this chaotic dynamical system behaves like a random walk constrained onto the surface of a hypersphere. The growth of error is limited, with qualitatively different behaviour depending on a control parameter. For moderate values of the control parameter, linear growth of error is seen. For low values of the control parameter, the error is limited by the random walk behaviour. Finally, we link this to the predictability of homogeneous isotropic turbulence, which we find here also behaves like a constrained random walk.