MEASURES DESCRIBING A TURBULENT FLOW

INTRODUCTION The motion of a fluid in a region R of R' or R' is defined by a function t : , ( t ) , where u ( t ) belongs to some functional space, %, of velocity fields in R. In a turbulent regime, one expects r , ( t ) to be distributed according to some probability law. This probability law is defined by a measure p on 7f . invariant under the deterministic time evolution of the system. We have reached some understanding of the invariant measures for a time evolution in a finite-dimensional space 7 f . The notions of strange attractor, sensitive dependence on initial condition, and characteristic exponents have been useful in this respect. Also, it is possible to define stable and unstable manifolds almost everywhere, and one can, in a number of cases, identify those measures which are stable under small stochastic perturbations. In this paper, we discuss the extension of results obtained for finite-dimensional dynamical systems to the more realistic case of the time evolution defined in a Hilbert space by the Navier-Stokes equation.