MUTUAL INFORMATION AND GLOBAL STRANGE ATTRACTORS IN TAYLOR-COUETTE FLOW

We apply a new method for the estimation of the mutual information M to experimental time series from different locations in weakly turbulent Taylor-Couette flow. We observe a dependence of M on the partition scale epsilon and find M (epsilon) is-proportional-to D(M) log (1/epsilon) for small epsilon and sufficient embedding dimension m. Our results confirm, that the overall dynamics of the Taylor-Couette flow is dominated by a global strange attractor of fractal information dimension D1 = D(M) at moderate Reynolds numbers. The dimension estimates obtained by our method are found to be more reliable than results from dimension algorithms based on single measurements alone, a fact which makes our method most attractive for dimension analysis in extended systems.